Description
Adam Smith hypothesized that impersonal exchange was necessary for a society to develop
specialized division of labor and create wealth. Douglass North and Vernon Smith argue
that successful developed economies are the result of institutions. We hypothesize and
provide evidence from ethnographic data that the basic accounting technology of recording
transactions is associated with more extensive impersonal exchange and increased specialization
in the division of labor. Our intuition is that extensive impersonal exchange
requires reliable memory of trading partners’ past behavior to sustain trust and encourage
reciprocity when a group expands beyond the size of traditional hunter-gatherer groups.
Our findings are consistent with the hypothesis that transaction records are necessary
for the emergence of complex economies as suggested by the archaeological evidence of
recordkeeping in Mesopotamian societies 10,000 years ago.
Memory, transaction records, and The Wealth of Nations
Sudipta Basu
a
, Marcus Kirk
b
, Greg Waymire
c,
*
a
Fox School of Business, Temple University, Philadelphia, PA 19122, USA
b
Fisher School of Accounting, University of Florida, Gainesville, FL 32611, USA
c
Goizueta Business School, Emory University, Atlanta, GA 30322, USA
a b s t r a c t
Adam Smith hypothesized that impersonal exchange was necessary for a society to develop
specialized division of labor and create wealth. Douglass North and Vernon Smith argue
that successful developed economies are the result of institutions. We hypothesize and
provide evidence from ethnographic data that the basic accounting technology of recording
transactions is associated with more extensive impersonal exchange and increased special-
ization in the division of labor. Our intuition is that extensive impersonal exchange
requires reliable memory of trading partners’ past behavior to sustain trust and encourage
reciprocity when a group expands beyond the size of traditional hunter-gatherer groups.
Our ?ndings are consistent with the hypothesis that transaction records are necessary
for the emergence of complex economies as suggested by the archaeological evidence of
recordkeeping in Mesopotamian societies 10,000 years ago.
Ó 2009 Elsevier Ltd. All rights reserved.
Introduction
The palest ink is better than the best memory.
Chinese Proverb
(K)nowledge of the past, the record of truths revealed
by experience, is eminently practical, as an instrument
of action and a power that goes to making the future.
Lord Acton
1
The past had not merely been altered, it had actually
been destroyed. For how could one establish even the
most obvious fact when there existed no record outside
your own memory?
George Orwell
2
Human economies vary considerably in scale, com-
plexity, and performance; some generate great wealth
while others remain mired in poverty. In The Wealth of
Nations, Smith (1776/1976, Book I, Chapter II, p. 17) ar-
gued that the growth of economies derived from exten-
sive impersonal exchange, which grew out of a human
‘‘propensity to truck, barter, and exchange one thing for
another.” Humans sustain cooperation better than other
primate species in part because we can remember and
communicate information about the cooperative acts of
others, which is a prerequisite for reciprocity and reputa-
tion formation (Axelrod & Hamilton, 1981; Nowak & Sig-
mund, 2005). Nevertheless, the evolved proclivities and
abilities of the brain that favor exchange and cooperation
can account for human groups only up to a size of about
200 persons (Dunbar, 1992). What role did the institution
of recordkeeping play in allowing some economies to cir-
cumvent the biological constraints of memory and there-
by expand impersonal exchange and produce great
wealth?
The institutions that societies use to govern economic
and social interaction have been suggested as necessary
0361-3682/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.aos.2009.07.002
* Corresponding author. Tel.: +1 404 727 6589; fax: +1 404 727 6313.
E-mail address: [email protected] (G. Waymire).
1
Inaugural lecture on the study of history, Cambridge University, June
1895, as cited in Collini, Winch, and Burrow (1983).
2
Orwell (1949/1961, p. 36).
Accounting, Organizations and Society 34 (2009) 895–917
Contents lists available at ScienceDirect
Accounting, Organizations and Society
j our nal homepage: www. el sevi er. com/ l ocat e/ aos
for economic development (North, 2005; Smith, 2008).
3
These institutions include legal codes that support property
rights and money that relaxes constraints inherent to barter
exchange, both of which date back thousands of years (Sag-
gs, 1989, pp. 156–175; Redish, 2003). Recordkeeping for ex-
change transactions is an even older institution found in the
?rst human settlements of Mesopotamia circa 8000 BCE
(Schmandt-Besserat, 1992). Humans ?rst invented writing
to keep records (Nissen, Damerow, & Englund, 1993), which
coincided in time (circa 3000 BCE) with the emergence of
the ?rst cities, underscoring the central importance of trans-
action records that objectively document the history of ex-
change. Accounting scholars have long recognized the
presence of recordkeeping in ancient societies (Carmona &
Ezzamel, 2007, provide an overview; see also Macve,
2002).
4
Despite this evidence, we have no parsimonious sci-
enti?c explanation and little broad evidence for the emer-
gence of transaction records and their role in economic
and social development.
Basu and Waymire (2006) hypothesize that transaction
records emerge to symbolically represent the history of ex-
change in a more permanent manner external to the hu-
man brain. External records lessen the chances of
individual memory failure and are valuable in tracking a
trading partner’s past behavior as a basis for current deci-
sions. Records can also establish reliable social memory
and common knowledge useful to two or more parties in
structuring an exchange. For example, ‘‘hard” records that
are veri?ed by witnesses can make it ‘‘dif?cult for people
to disagree” later about whether past promises have been
ful?lled (Ijiri, 1975, p. 36). As recordkeeping evolves to en-
code more information, it enables drafting and enforcing
contracts that govern complex exchange transactions
across time and geographical boundaries – e.g., reliable re-
cords are needed to secure the property rights that facili-
tate modern capitalism (De Soto, 2000).
Basu and Waymire’s hypothesis parallels the hypothesis
that double-entry bookkeeping enabled modern capitalist
organization, which was advanced by Sombart (1919), We-
ber (1927), Joseph Schumpeter (1942), and Ludwig von
Mises (1949).
5
Both hypotheses assert a foundational role
for basic accounting institutions in enabling the emergence
of more complex forms of economic and social interaction.
The main distinguishing feature of Basu and Waymire’s
hypothesis is that basic accounting institutions like record-
keeping play this role early in the development of complex
societies rather than in emergent business organizations
within already developed societies.
Basu and Waymire (2006) predict that (1) recordkeep-
ing emerges because increasing exchange complexity
taxes the brain’s memory resources, and (2) accounting
records work in tandem with other fundamental institu-
tions to promote economic development. These predic-
tions can be investigated in several complementary
ways. One is to conduct experiments using neuroscienti?c
methods to investigate whether the human brain’s evalu-
ation of exchange parallels culturally evolved accounting
practices (Dickhaut, Basu, McCabe, & Waymire, 2009a,
2009b). A second avenue is to investigate whether the
causal links inherent in the Basu and Waymire (2006)
story are observed in a controlled experimental setting
(Basu, Dickhaut, Hecht, Towry, & Waymire, 2009). An-
other possibility, which we explore in this paper, is to
use naturally occurring data to test whether institutional
and economic development is greater in those societies
that have developed technologies for recording
transactions.
We present evidence on whether the association be-
tween recordkeeping technology and societal size and
complexity in ancient Mesopotamia generalizes to other
human societies, and how strongly the prevalence of
recordkeeping is associated with the presence of other
fundamental economic institutions. We explore these is-
sues using ?eld data collected by ethnographers and
archaeologists from a broad cross-section of human socie-
ties, and subsequently coded into machine-readable data
by Murdock and White (1969). Murdock and White’s
Standard Cross-Cultural Sample (SCCS) provides extensive
coded data for a variety of cultural variables – presently
over 2000 – for 186 societies selected to maximize the
cross-society independence of observations. SCCS societies
are ‘‘pinpointed” to speci?c dates that vary across socie-
ties. The SCCS data include a variable that measures a
society’s recordkeeping technology as well as numerous
measures of economic, social, and institutional
development.
We use the SCCS data to empirically evaluate whether
recordkeeping is a necessary institution for unleashing
the economic forces of impersonal exchange and division
of labor hypothesized by Smith to be the ultimate source
of economic wealth. We document that recordkeeping use
and sophistication is greater in societies that have sur-
passed the modest levels suggested by Dunbar (1992),
and also that recordkeeping is present as early as or ear-
lier in economic development than other basic institu-
tions such as money, property rights, hierarchical
organizations, a judiciary, and the use of credit. This evi-
dence suggests that recordkeeping is a precursor to rather
than the result of economic complexity. Our analyses also
demonstrate that the extent of impersonal exchange is
positively associated with the use of recordkeeping, and
that specialized division of labor and the level of capital
accumulation are more extensive in societies with greater
opportunities for market exchange. Collectively, our ?nd-
ings are consistent with the hypothesis that basic
accounting institutions are necessary but not suf?cient
to foster the extensive impersonal exchange and complex
3
We use ‘‘institution” in the broad sense of Douglass North (1991, p. 97):
‘‘Institutions are the humanly devised constraints that structure political,
economic and social interaction. They consist of both informal constraints
(sanctions, taboos, customs, traditions, and codes of conduct) and formal
rules (constitutions, laws, property rights). Throughout history, institutions
have been devised by human beings to create order and reduce uncertainty
in exchange. They evolve incrementally, connecting the past with the
present and the future; history in consequence is largely a story of
institutional evolution in which the historical performance of economies
can only be understood as part of a sequential story.”
4
Earlier examples include Littleton (1927), Robert (1952), Keister (1963),
Keister (1964), Mattessich (1987) and Baxter (1989).
5
This hypothesis and its empirical basis are discussed and debated by
Yamey (1949), Yamey (1964, Yamey (2005), Most (1972), Most (1973), and
Carruthers and Espeland (1991).
896 S. Basu et al. / Accounting, Organizations and Society 34 (2009) 895–917
economic interactions that characterize modern devel-
oped economies.
As with all studies, ours is subject to important cave-
ats. Because we examine naturally occurring data that
are subject to measurement error and are not designed
to measure intertemporal change within economies, our
analysis does not speak directly to causality in the rela-
tions between recordkeeping, exchange, and economic
development. Rather, our paper is an attempt to investi-
gate cross-cultural statistical associations between
recordkeeping and patterns of economic interaction ob-
served in societies at early stages of their development.
The evidence we present here complements the causal
evidence on the origins and consequences of basic
accounting institutions presented in Basu, Dickhaut,
Hecht, Towry, and Waymire (2009) and Dickhaut, Basu,
McCabe, and Waymire (2009a), Dickhaut, Basu, McCabe,
and Waymire (2009b). In combination, all these studies
suggest that veri?able historical transaction records play
a foundational role in the development of market econo-
mies. This study viewed in isolation is a necessary, but
not suf?cient, condition to support such a claim insofar
as it only provides evidence on the external validity of
the hypothesis.
We ?rst discuss the history of recordkeeping and de-
scribe how recordkeeping systems vary across the SCCS
societies. We then present our hypotheses. We next offer
evidence on the emergence of recordkeeping in the course
of social and economic development. Following that, we
present evidence on the association between recordkeep-
ing, impersonal exchange, and the division of labor. Con-
clusions from and implications of our ?ndings are
summarized in the paper’s ?nal section.
The history of transaction records and recordkeeping in
SCCS societies
Historical background on the origins and nature of transaction
records
Kohler’s (1952, p. 356) A Dictionary for Accountants de-
?nes a ‘‘record” as ‘‘A book or document containing or evi-
dencing some or all of the activities of an enterprise or
containing or supporting a transaction, entry, or account.”
Examples include ‘‘a book of account; subsidiary ledger;
invoice; voucher; contract; correspondence; internal re-
port; minute book.” This de?nition and examples evoke
images of a ‘‘paper trail” where writing symbolically por-
trays the types and quantities of goods exchanged, the per-
sons buying and selling the goods, and their obligations
under the exchange. Implicit in this characterization is that
a written language, numbers, and a system of weights and
measures already exist that shape the speci?cs of a trans-
action record.
Yet, this assumes more than existed in the earliest
recordkeeping systems. Indeed, the earliest transaction
records precede writing by thousands of years (Sch-
mandt-Besserat, 1992). The0 sophistication of records is
directly tied to how well various properties of exchange
goods can be categorized, described and measured. A
transaction record is merely an artifact that does not
per se require written language or the use of re?ned
weights and measures, although such things are useful
(Hutchins, 1999; Schmandt-Besserat, 1999). A record is
valuable because it allows a person to remember impor-
tant attributes of a transaction; two or more persons can
also use a transaction record to state their common
knowledge about the nature of the transaction. Preserv-
ing exchange information on records outside the brain
increases the life of common knowledge – that is, shared
understandings can be carried forward in time to a
greater extent than would be the case with the spoken
word.
The oldest-known recordkeeping systems were used in
Mesopotamian villages in 8000 BCE, with baked clay to-
kens and stones symbolizing individual agricultural com-
modities transferred between two parties (see the left-
hand side of Panel A in Fig. 1). By 4000 BCE, transfers of dif-
ferent manufactured goods were recorded with newer
complex tokens with various shapes and incisions (Nissen
et al., 1993; Schmandt-Besserat, 1995; Schmandt-Besserat,
1996). Tokens were sealed within baked hollow clay balls
(‘‘bullae”) by 3500 BCE. By 3200 BCE, personal seals of
the transacting parties and witnesses, along with indenta-
tions of the enclosed tokens, were impressed on the exte-
rior of the bulla before it was baked (see the right-hand
side in Panel A of Fig. 1).
Schmandt-Besserat (1995, p. 2100) sees the Mesopota-
mian token as a ‘‘mnemonic device by which to handle
and store an unlimited quantity of data without risking
the damages of memory failure.” Tokens and bullae
stored data on several transaction attributes long before
writing was invented. First, the unique token shapes al-
lowed the type and quantity of commodities exchanged
to be clearly identi?ed. Second, personal seals enabled
identi?cation of the exchange parties and witnesses from
the af?xed ‘‘signatures” on a bulla. Third, the external to-
ken impressions let everyone know the commodities in-
volved without breaking open the bulla. This was
advantageous since the bulla need then be broken only
at the transaction settlement, thereby facilitating com-
plex intertemporal exchange. Fourth, the baking of bullae
and their storage in a secure location rendered the infor-
mation ‘‘hard” both literally and also metaphorically in
that it made it ‘‘dif?cult for people to disagree” later (Ijiri
1975, p. 36).
Another non-written recordkeeping system was the In-
can quipu, ?rst documented after the Spanish conquered
that large and wealthy civilization around 1500 AD. The In-
can quipu (shown in panel B of Fig. 1) was a ‘‘knotted
string” on which transaction attributes were symbolized
by thread colors and length as well as the number of knots
and their location on the cord (Urton, 2002). Like the early
Mesopotamian tokens, the quipus kept track of tribute paid
by citizens to the Incan government (Assadourian 2002, p.
120; Pollock 1999, pp. 78–116).
The ‘‘tally stick” (see Panel C of Fig. 1), relied upon by
the English Treasury as recently as the 19th century, was
a receipt for tax collections by Royal agents (Robert,
1952; Baxter, 1989). The tally stick differed from tokens
and quipus in that some writing was employed (Robert,
S. Basu et al. / Accounting, Organizations and Society 34 (2009) 895–917 897
1952, p. 76). Tally sticks have been used all over the world,
at least as early as 500 BC by the Chinese (Goetzmann &
Williams, 2005), and as recently as the 1970s in rural
France (Ifrah, 2001). These pre-literate recordkeeping
systems are similar to the vouchers, contracts and other
written material that are used to generate ‘‘journal entries”
Fig. 1. Examples of recordkeeping technologies present in human history.
Picture sources:
Left hand side of Panel A in this ?gure: Schøyen Collection, Oslo and London.
.
Right hand side of Panel A in this ?gure: Louvre Museum, Paris, France.http://www.humanitiesinteractive.
org/ancient/mideast/index.html?collectionVar=AncientCulturesStop&pageVar=2.
Panel B in Fig. 1: Larco Museum, Lima, Peru.
.
Panel C in Fig. 1: Science Museum, London, UK.
.
Panel D in Fig. 1: Schøyen Collection, Oslo and London.
.
898 S. Basu et al. / Accounting, Organizations and Society 34 (2009) 895–917
in double-entry bookkeeping systems developed in 13th
century Italy. An ancient direct analog to the receipt is
the Cuneiform tablet (see Panel D of Fig. 1), which
Schmandt-Besserat (1992) has hypothesized to have
evolved from the earlier token system.
Academic accountants have long recognized the ancient
origins of accounting records (Keister, 1963; Keister, 1964;
Littleton, 1927; Mattessich, 1987; Mattessich, 2000). Inter-
est in the subject of ancient accounting has accelerated in
the past few decades, both within the accounting literature
and the archaeological literature (Carmona & Ezzamel,
2007; Hudson & Wunsch, 2004). The picture that emerges
from these literatures is that transaction records and the
accountability they render have played an important role
in ordering the economies of ancient societies in Egypt
and Mesopotamia (Hudson, 2004). Our study complements
these literatures by offering broader statistical evidence on
the emergence of recordkeeping and its role in economic
development.
Recordkeeping in SCCS societies
We use the Standard Cross-Cultural Sample (SCCS) con-
structed by Murdock and White (1969) in all of our empir-
ical analyses. Anthropologists have used the SCCS
extensively in ethnographic research; recently it has been
used by economists in studies of cross-cultural differences
in the structure and performance of economies in varying
stages of development (e.g., Baker, 2008; Baker & Miceli,
2005).
The SCCS comprises 186 societies sampled from a wide
range of time periods (including two from before the
Common Era) and geographical locations. The SCCS socie-
ties include contemporary hunter-gatherers, early historic
states, and contemporary industrial societies. This wide
coverage re?ects Murdock and White’s (1969) conscious
decision to mitigate selection biases that favored societies
with English language ethnographic sources. The sam-
pling unit is typically the local community studied in
the primary ethnography used to code a given society
(Murdock & White, 1969, p. 30). Appendix A provides
an in-depth description of how the SCCS database was
constructed.
A major issue with a sample like the SCCS is whether it
is representative of the population of societies present
throughout the world. Some have argued that the SCCS
societies constitute a sample biased in favor of those stud-
ied more intensively by Murdock, his colleagues, and their
students (Otterbein, 1976). Subsequent research demon-
strates that such bias is not present in the SCCS (Gray,
1996). Rather, the predominant bias in the SCCS is towards
societies that have been more intensively studied by eth-
nographers, as previously noted by Murdock and White
(1969, p. 332).
The SCCS presently provides data on more than 2000
categorical variables coded nominally or ordinally. Mur-
dock and Morrow (1970) coded 22 variables related to sub-
sistence economy and related practices, and these
variables are the ?rst listed in SCCS. Additional variables
are added to the free online electronic database as
researchers encode new variables by reading the pre-spec-
i?ed primary and secondary ethnographic studies.
6
Our primary variable of interest is Recordkeeping (SCCS
variable #149, entitled ‘‘Writing and Records”), which is
coded on an ordinal scale from 1 to 5 (Murdock & Provost,
1973, pp. 379–380).
7
Panel A of Fig. 2 describes how this
variable is categorized and shows the number of SCCS soci-
eties classi?ed into each category (see also Panel A of Appen-
dix B). We defer the introduction of several other SCCS
variables until the later sections in which we analyze them.
A value of 1 for Recordkeeping signi?es ‘‘writing, re-
cords, and mnemonic devices in any form are lacking or
unreported.” Seventy-three (39.3%) of the SCCS societies
are coded as completely lacking in records. One such soci-
ety is that of the Mbuti, who are nomadic, gathering Pyg-
mies living in Central/East Africa. Panel B of Fig. 2 shows
that non-recordkeeping societies are less prevalent than
recordkeeping societies (those with a score of 2 or more)
in each sub-sample classi?ed by the society’s pinpointed
year. Panel C of Fig. 2 demonstrates that a greater number
of non-recordkeeping cultures are located in Sub-Saharan
Africa and South America. Conversely, the frequency of cul-
tures possessing records is highest in the Circum-Mediter-
ranean and Eurasia.
8
The next Recordkeeping value of 2 indicates that ‘‘writ-
ing and signi?cant records are lacking but the people em-
ploy mnemonic devices, e.g., simple tallies.” Forty-nine
SCCS societies (26.3%) are coded as using mnemonic de-
vices. Examples include the ancient Mesopotamian tokens,
shown in Panel A of Fig. 1, or the shells used as wampum
by American Indians (Schmandt-Besserat, 1992; Szabo,
2005). A speci?c SCCS example is that of the Kapauku Pap-
uans of New Guinea, who use shell artifacts extensively in
exchange (Pospisil, 1963, pp. 291–293 and 300–311).
A value of 3 for Recordkeeping indicates that a society
‘‘lacks true writing but possesses signi?cant non-written
records in the form of picture writing, quipus, pictorial
inscriptions, or the like.” Twenty-one SCCS societies
(11.3%) are coded as having non-written records. One of
these societies is that of the Incas, who were pinpointed
in 1530 AD shortly after the Spanish invasion of the Amer-
icas. The Incan quipu shown in Panel B of Fig. 1 has long
been recognized as an accounting device to record transac-
tions (Keister, 1964; Urton, 2002).
6
The construction of data for a single variable in SCCS is a painstaking
process. Data coders typically read one or more primary sources and several
secondary sources to determine whether a given society follows a
particular practice. The speci?c coded values are then compiled and
included in SCCS after additional checks are made on their accuracy. SCCS
often includes additional code indicating when categorization of a speci?c
data point is more ambiguous. A quality-ordered bibliography of reference
sources is included in the database (White et al., 2005), and the coding
sheets indicate which particular sources were used for a particular society’s
code, so that subsequent researchers can assess data reliability (White,
1986). Murdock and Morrow (1970, pp. 312–330) describe this process in
detail for the ?rst variables included in SCCS and the individual sources
used in creating these data for each speci?c society.
7
Murdock and Provost (1973) originally coded it 0–4 but the SCCS
presently reports it from 1 to 5.
8
These ?ndings are consistent with Diamond’s (1997) theory that useful
technologies were more likely to spread along the same latitudes in Eurasia
than along the same longitudes in Africa and South America.
S. Basu et al. / Accounting, Organizations and Society 34 (2009) 895–917 899
A value of 4 for Recordkeeping indicates a society that
‘‘has an indigenous system of writing but lacks any signif-
icant accumulation of written records or alternatively has
long used the script of alien people.” Twelve SCCS societies
(6.5%) fall into this category.
9
The ethnographic texts on
which this code is based indicate that written records exist
but provide few speci?cs. For example, Longrigg (1953, pp.
21–25) describes book production and newspapers in Kurd-
istan around 1900, Barth (1960, p. 32) notes that marriage
contracts among the Basseri were drafted by specialists in
marriage rites, and Gamble (1967, pp. 22–26) remarks that
Wolof was the commercial language in the Wolof society
and that school books on this language existed as early as
1823.
The highest category for Recordkeeping (coded as 5) ap-
plies to 31 SCCS societies (16.7%) where the ‘‘society has an
indigenous system of true writing and possesses written
records of at least modest signi?cance.” This group in-
cludes the society (Babylonia) with the earliest pinpointing
Fig. 2. Description of recordkeeping variable from standard cross-cultural sample (SCCS).
9
Although these societies are ranked higher than those in the Record-
keeping category 3, their order could arguably be reversed. To mitigate such
ranking ambiguity, we redid all our analyses excluding all societies in
Recordkeeping category 4 and found very similar results.
900 S. Basu et al. / Accounting, Organizations and Society 34 (2009) 895–917
date (1750 BCE) in the SCCS sample. This date is at the end
of Hammurabi’s reign as Babylonian monarch in a period
when information on transactions and contracts was regu-
larly stored on Cuneiform clay tablets (Van De Mieroop,
2002; Van De Mieroop, 2004). Panel D of Fig. 1 shows an
example of a Cuneiform tablet.
We acknowledge that Recordkeeping is only an indirect
measure of the presence and use of transaction records
within a society. Thus, an SCCS society coded as having a
fully developed writing system could also use that system
for creating other documents – e.g., religious texts. We be-
lieve that the archaeological and historical evidence cited
in this section suggests strongly that Recordkeeping re?ects
the underlying construct of recording transactions. At the
same time, this variable, like many collected for other pur-
poses, is measured with some degree of error. Strictly
speaking, Recordkeeping re?ects the availability of a tech-
nology that could be used for recording economic
transactions.
Hypotheses
The simplest accounting systems provide a historical
record of economic transactions in which a one-way trans-
fer or bilateral exchange has occurred. Transactions gener-
ate a ‘‘paper trail” of receipts, vouchers and contracts that
can be used to verify transaction details in case of forget-
fulness or subsequent disputes.
10
Transaction records are
common to large-scale societies, even those that are pre-lit-
erate. The oldest-known accounting records using ‘‘tokens”
appear at the same time and place (circa 8000 BCE Mesopo-
tamia) as the emergence of agriculture and permanent hu-
man settlements (Schmandt-Besserat, 1992). The
Sumerians invented writing to keep records and accounts
(circa 3200 BCE), which occurred at the same time as sub-
stantial increases in group size and population density in
the earliest cities (Nissen et al., 1993). Thus, accounting
innovations in ancient Mesopotamia coincided with in-
creased societal and economic complexity, suggesting a po-
tential causal connection.
Our focus in this paper is on whether the tight coupling
between accounting advances and economic development
in the ancient Near East was unique, or a pattern that de-
scribes human societies more generally. We are also inter-
ested in the extent to which recordkeeping expands the
scope of impersonal exchange and promotes increasingly
specialized division of labor. We investigate three hypoth-
eses that are depicted in Fig. 3. Each hypothesis is stated in
the lower portion of the ?gure.
Our ?rst hypothesis concerns the emergence of record-
keeping as a society expands in size. We investigate two
implications of the hypothesis. First, recordkeeping is more
likely to emerge once a society has reached the size im-
plied by mental memory constraints. Second, recordkeep-
ing is a necessary institution for extensive impersonal
exchange. As such, recordkeeping will emerge as early as
or earlier than other economic institutions that support ex-
change. The arrow between the box labeled ‘‘Internal
Memory” and the box labeled ‘‘External Records” in Fig. 3
depicts our ?rst hypothesis.
Relying solely on mental memory, humans sustain
greater social exchange than other primates. This is largely
because our evolved brains remember past interactions
and analyze exchange opportunities more effectively than
other species (Cosmides & Tooby, 2005; Wilson, 2000).
That is, human brains are adapted for social exchange
and cooperation that improves our prospects for resource
acquisition and survival. Within a small kin-based group,
mental memory of past interactions and third-party gossip
helps actors identify trustworthy partners for a contem-
plated cooperative venture (Barkow, 1992; Demsetz,
2002). Hence, small groups have little need for external re-
cords because members can accurately track others’ repu-
tations even if they cannot perfectly recall the particulars
of all past interactions (Silk, 2004).
Keeping physical records outside the brain allows peo-
ple to reliably store greater amounts of information on past
interactions and better evaluate the desirability of ex-
change with a speci?c partner (Basu & Waymire, 2006;
Dickhaut et al., 2009a; Dickhaut et al., 2009b). Recordkeep-
ing expands human capacity to ‘‘recognize other individu-
als and keep score” (Ridley 1996, p. 83), which is a
prerequisite for sustaining repeated cooperative social ex-
change and reciprocity (Axelrod & Hamilton, 1981; Nowak
& Sigmund, 2005).
11
This suggests that sole reliance on
mental records will constrain societal expansion beyond a
modest group size.
The evolved human brain can sustain stable cooperative
groups to an upper limit of between 125 and 200 mem-
bers.
12
Dunbar (2001, p. 181) writes:
‘‘(T)here is indeed a characteristic group size of around
125–200 that reappears with surprising frequency in a
wide range of contemporary and Neolithic horticultural
societies. These groups . . . all share one crucial charac-
teristic: they consist of a set of individuals who know
one another intimately and interact on a regular basis. . .
Thus there seems to be quite strong evidence that at
least one component of human grouping patterns is as
much determined by relative neocortex size as are
groups of other primates. We have bigger, more com-
plexly organized groups than other species simply
because we have a larger onboard computer (the neo-
cortex) to allow us to do the calculations necessary to
10
Accounting scholars have long recognized the importance of basic
recordkeeping and its role in providing memory of past exchange trans-
actions (e.g., Demski, 1993; Hat?eld, 1924; Ijiri, 1975; Littleton, 1933,
1953; Potter, 1952).
11
An organism’s ability to recall past interactions with its environment
and adjust behavior in response is of ?rst-order importance to its survival.
This ability is important even for single cell organisms like the Escherichia
coli bacterium (Allman, 2000, pp. 3–8).
12
‘‘Dunbar’s Number” of 125–200 persons was calculated by correlating
troop size and (neocortical) brain size across different primates such as
monkey, baboons and chimpanzees, and extrapolating to expected human
group size using actual human brain size (Dunbar, 1992; Dunbar, 1998).
Dunbar validated his predicted number by studying the historical maxi-
mum sizes of hunter-gatherer tribes, Neolithic villages, Hutterite settle-
ments, Roman army units, and other human groups. Edney (1981, p. 27)
independently estimated that ‘‘the upper limit for a simple, self-contained,
sustaining, well-functioning commons may be as low as 150 people.”
S. Basu et al. / Accounting, Organizations and Society 34 (2009) 895–917 901
keep track of and manipulate the ever-changing world
of social relationships within which we live.”
Dunbar’s Number is the estimated limit to human
group size in the absence of institutions that store data
on past economic and social interactions outside the hu-
man brain. However, transaction records serve as an
(expandable) external hard drive for the mental computer.
This suggests that external recordkeeping should become
increasingly prevalent for groups that exceed 200 persons.
Thus, our ?rst hypothesis is that the extent of recordkeep-
ing and group size will be associated non-linearly, with lit-
tle relation for groups of 200 or fewer persons and a
positive relation for groups exceeding 200 persons.
As a group grows in size, repeated interaction with
familiar partners occurs less often. In addition, individual
cooperation with members of other groups cannot rely
on familiarity or repeated interaction. Recordkeeping helps
people successfully consummate impersonal exchange and
it subsequently enables the emergence of other exchange-
supporting institutions that rely on hard evidence of past
transactions. For example, hard transaction records often
provide a basis for establishing and enforcing property
rights subsequent to property transfers (de Soto, 2000;
VerSteeg, 2000, pp. 46–66). Records also provide the basis
for compiling an individual’s exchange and credit histories,
either in a speci?c market (e.g., eBay) or more generally
with credit ratings like Moody’s. Thus, our ?rst hypothesis
implies also that recordkeeping is a precursor to other ex-
change-supporting institutions. As such, recordkeeping is
expected to appear as early as or earlier than other ex-
change-supporting institutions as an economy expands.
Our second hypothesis is about how essential record-
keeping is for a market to expand and encompass larger
numbers of impersonal exchange transactions. This
hypothesis is depicted by the curved arrow in Fig. 3 con-
necting ‘‘External Records” to ‘‘Impersonal Exchange”
Within human families, many resource transfers are unidi-
rectional grants that can be motivated by love (by parents
for children), fear (by low status members of alpha males)
and ignorance (not recognizing that an object is valuable)
(Boulding, Pfaff, & Horvath, 1972).
13
Most primitive human
societies are extended kin groups with a norm of generalized
reciprocity, where help is expected from and is available to
all group members (Sahlins, 1972).
These societies begin to transcend the bounds of the
immediate group through gift exchanges with neighboring
groups. Such complex interactions inevitably entail rigid
norms of behavior that reduce cheating and any resultant
misinterpretation of intentions (Malinowski, 1922; Mauss,
Fig. 3. Hypotheses. Graphical depiction of hypotheses on the emergence of recordkeeping and its relation to exchange, division of labor, and economic
development.
13
Over lifetimes, such one-way resource transfers likely balance out, but
given high mortality rates in these groups, there is less expectation of stable
partnerships. In more egalitarian societies, transfers between spouses may
have more of an implicit exchange character than in less egalitarian
cultures. Thus, we do not mean to characterize these transfers as
necessarily excluding an exchange component, but rather want to empha-
size that they are not purely exchange transactions between equals.
902 S. Basu et al. / Accounting, Organizations and Society 34 (2009) 895–917
1950). As a by-product of formal gift exchange, trading in
less elaborate economic goods often develops. The scope
of exchange expands to include items that are not ‘‘gifts”
per se but rather are given outside the elaborate rituals of
gift exchange, with informal norms of balanced or sym-
metrical reciprocity in which a fair return is expected from
individual recipients in the future (Sahlins, 1972). Thus, as
the size of an exchange network grows, economic interac-
tion with less-closely-related acquaintances occurs more
often.
14
At some point in the recent human past, a new form of
economic interaction arose in the form of bilateral imper-
sonal exchange or barter between strangers (Seabright,
2004), with an associated norm of negative reciprocity
where each person gives up valuable things and expects
to be reciprocated immediately in a quid pro quo manner
(Sahlins, 1972). Basu’s et al. (2009) experimental results
suggest that recordkeeping may be crucial to this transfor-
mation towards market exchange, in that experimental
economies with recordkeeping exhibited stronger patterns
of negative reciprocity in impersonal exchange than non-
recordkeeping economies. Thus, our second hypothesis is
that recordkeeping will be associated with more extensive
and more complex exchange transactions within a given
society.
Our third hypothesis is that the expanded exchange en-
abled by recordkeeping is associated with increased spe-
cialization in division of labor. Smith (1776/1976, p. 17)
?rst articulated the relation between exchange and divi-
sion of labor in The Wealth of Nations:
This division of labour, from which so many advantages
are derived, is not originally the effect of any human
wisdom, which foresees and intends that general opu-
lence to which it gives occasion. It is the necessary,
though very slow and gradual, consequence of a certain
propensity in human nature which has in view no such
extensive utility; the propensity to truck, barter, and
exchange one thing for another.
Smith (1776, Book I, Chapter III, p. 21) then states his fa-
mous theorem ‘‘As it is the power of exchanging that gives
occasion to the division of labour, so the extent of this divi-
sion must always be limited by the extent of that power,
or, in other words, by the extent of the market” (see also
Stigler, 1951). We depict this hypothesis with the arrow
from ‘‘Impersonal Exchange” to ‘‘Division of Labor” in
Fig. 3. If recordkeeping promotes market expansion (our
second hypothesis), which in turn enables division of labor,
we can also expect a positive association between record-
keeping and division of labor when market extent is ex-
cluded from the regression.
Recordkeeping emerges as group size increases
Our ?rst hypothesis is that recordkeeping becomes
more prevalent after a group reaches the maximum size
achievable under biological constraints and that record-
keeping emerges as early as or earlier than other economic
institutions. A direct test of this hypothesis requires a mea-
sure of group size that re?ects the total number of people
in the group taking account of network ties with non-
group members.
15
The closest proxy within SCCS for this
construct is variable #63, Community Size, which is available
for all but one of the 186 SCCS societies. Community Size is
only a proxy for the overall size of a networked group since
it refers to the size of a typical community (i.e., city or vil-
lage) in the society being studied (Murdock & Wilson,
1972). This is an imperfect proxy to the extent that a com-
munity may have network links to individuals outside the
local community. Community Size takes on eight possible
categorical values. At the lowest are societies where ethnog-
raphers estimate the typical community size to be less than
50 persons and at the highest are communities that each
consist of more than 50,000 people. Panel B of Appendix B
shows the various categories of Community Size.
Panel A of Fig. 4 shows a plot of the frequency of each
Recordkeeping score for a given level of Community Size,
where bubble sizes are proportional to frequency. A line
connects the mean Recordkeeping score for each of the
eight Community Size categories. Consistent with our ?rst
hypothesis, the association between the Recordkeeping
and Community Size is positive and non-linear with a
monotonically increasing mean once Community Size
reaches the level of 200 persons or more. This relation is
reliably positive and signi?cantly different from zero
(Spearman q = 0.32; p < 0.01).
Visual inspection of Panel A in Fig. 4 indicates that most
of the SCCS societies without recordkeeping cluster in the
lower Community Size communities. However, this effect
is not uniform as 14 SCCS societies containing fewer than
200 persons in the typical community also have record-
keeping systems based on a written language (i.e., Record-
keeping = 4 or 5 and Community Size = 1, 2, or 3). These 14
societies include some cultures where the size of the total
population in the society is quite large. For example, these
SCCS societies include villages in both Korea and Japan,
which were subjected to ethnographic study in 1947 and
1950, respectively.
To reduce the impact of measurement error in Commu-
nity Size, we also estimated a more comprehensive mea-
sure of a society’s demographics. We performed a factor
analysis on four SCCS variables: (1) Community Size, (2) Set-
tlement Patterns (SCCS variable #234), (3) Population Size
(SCCS variable #1122), and (4) Population Density (SCCS
variable #1130). Settlement Patterns measures the extent
to which a society is nomadic versus sedentary, Population
Size is a measure of the society’s total size (based on census
data when available), and Population Density measures
the number of persons per square mile in the society.
De?nitions of these variables are provided in Panel B of
Appendix B.
14
In industrialized societies, generalized reciprocity typically character-
izes interactions between parents and children, while balanced reciprocity
characterizes transactions with cousins, neighbors and co-workers.
15
Group size is extensively used to measure the scale of sustained
cooperation within a given species; likewise it is used as a parsimonious
measure of the scale and development of human social complexity (Chick,
1997; Johnson & Earle, 2000; Wilson, 2000, pp. 131–138; Dunbar, 2001).
S. Basu et al. / Accounting, Organizations and Society 34 (2009) 895–917 903
Panel B of Fig. 4 shows results from an iterated principal
factor analysis using communalities among these four vari-
ables. A minimum Eigenvalue of one is used to choose fac-
tors for our subsequent analysis. A single factor with an
Eigenvalue of 2.15 explains 90% of the variation of the four
variables. Accordingly, we retain a single factor to specify a
variable that we refer to as Demographics. Factor loadings
for Demographics with the four variables are shown in Pa-
nel B of Fig. 4. The factor loading is the standardized coef-
?cient in a regression of the variable on the factor and
re?ects the strength of the relationship. Population Density
has the most positive factor loading while Community Size
has the least positive factor loading. The weaker associa-
tion between Demographics and Community Size suggests
that important features of a society’s demographic com-
plexity are not fully captured by Community Size.
Panel C of Fig. 4 shows a plot of the frequency of each
Recordkeeping score for each of eight levels of Demograph-
ics, where again bubble sizes are proportional to frequency.
To enable direct comparison with Panel A, we rank the
societies on the basis of Demographics, and then partition
them into eight sub-samples using the same number of
societies in each of the eight ordered categories of Commu-
nity Size used to produce Panel A. Inspection of Panel C sug-
gests that using Demographics in lieu of Community Size
modestly increases the strength of association between
Spearman ? = 0.32 (n=185, p < .01)
Factor Eigenvalue
Proportion
Explained
Cumulative
Explained Factor Loadings
1 2.15 0.90 0.90 Population Density (SCCS #1130) 0.88
2 0.19 0.08 0.98 Population Size (SCCS #1122) 0.69
3 0.04 0.02 1.00 Settlement Patterns (SCCS #234) 0.68
4 -0.00 -0.00 1.00 Community Size (SCCS #63) 0.62
True writing
with records
Community Size
None
Mnemonic
devices
Non-written
records
True writing;
no records
< 50 50 to
99
100 to
199
200 to
399
400 to
999
1,000 to
4,999
5,000 to
49,999
> 50,000
Spearman ? = 0.37 (n=185, p < .01)
True writing
with records
None
Mnemonic
devices
Non-written
records
True writing;
no records
Demographics factor groups
1 2 3 6 7 8 4 5
(A)
(B)
(C)
Fig. 4. Relation between Recordkeeping and group size
904 S. Basu et al. / Accounting, Organizations and Society 34 (2009) 895–917
Recordkeeping and the measure of group size (Spearman
q = 0.37, p < 0.01). In addition, the plotted relation between
mean Recordkeeping score and Demographics shows a gen-
erally increasing trend.
Panel D of Fig. 4 shows estimates from a regression
model where Recordkeeping is the dependent variable and
either Community Size or Demographics is the independent
variable. The model allows a kinked relation between
Recordkeeping and the independent variable, with the kink
located at group size or complexity value equal to ‘‘4.”
Intuitively, this means that we allow the relation between
Recordkeeping and Community Size to differ once groups
have reached a level of Community Size greater than or
equal to 200 persons (or a comparable level of complexity
based on Demographics). The results indicate a non-linear
positive relation between the extent of recordkeeping
and group size. In both models, the coef?cient on the inter-
action between group size and a 0–1 indicator for whether
the society is ‘‘large/complex” is positive and signi?cant
(p < 0.02 for both models). This evidence suggests that, as
a society surpasses the modest group size threshold of
200 suggested by Dunbar (1992), recordkeeping becomes
more complex.
A second implication of our ?rst hypothesis is that
recordkeeping emerges as early as or earlier than other ex-
change-supporting institutions. We identi?ed other ex-
change-supporting institutions using SCCS variables
re?ecting the use of money and credit in an economy,
the presence of a judiciary and property rights, and the
presence of administrative hierarchies. Money and prop-
erty rights are likely fundamental to expansion of ex-
change (Demsetz, 2002; Menger, 1892) and the demand
for accounting arises in part from the existence of complex
organizations and credit markets (Kimbrough, Smith, &
Wilson, 2008; Watts & Zimmerman, 1986). We use the fol-
lowing ?ve variables: Credit Source (SCCS variable #18),
Judiciary (SCCS variable #89), Administrative Hierarchy
(SCCS variable #91), Money (SCCS variable #155), and
Inheritance of Land (SCCS variable #278). These variables
are de?ned in Panel B of Appendix B.
Fig. 5 plots the cumulative percentage of societies in
which Recordkeeping or each of the other ?ve institutions
is present, where societies are ordered by Community Size
level. A speci?c institution is deemed to be present if the
society’s code for a given variable exceeds the minimum
possible value. Each point on a given line represents the to-
tal number of societies up to that size level where the insti-
tution is present divided by the total number of societies
with a code available for that variable.
The six SCCS variables’ cumulative frequency functions
cluster into three groups referenced by the capital letters
on the right-hand side of the ?gure. The ?rst cluster (la-
beled A) includes Recordkeeping along with Money and
Inheritance of Land. These institutions are present in
approximately 60% of the SCCS societies. The cumulative
frequencies for Inheritance of Land and Money are not sta-
tistically different from the cumulative frequency of
Recordkeeping (p > 0.10 based on a z-test of proportions).
This suggests that recordkeeping is a fundamental institu-
tion that, like the use of money in exchange and simple
property rights systems, emerges early as an economy
develops. This relation is consistent with the strong corre-
lation between property rights and recordkeeping in the
SCCS documented by Baker and Miceli (2005), which they
conjectured was due to the bene?cial role of recordkeeping
in allowing property transfers. That recordkeeping and
money emerge in similar fashion is likely not surprising.
Economists have noted that shells and similar artifacts
can serve as money to promote exchange and that these
monetary artifacts provide memory of past exchanges
(Kocherlakota, 1998; Townsend, 1989).
16
Recordkeeping
i
= ? + ?
1
Size
i
+ ?
2
Size
i
*Large_Comm
i
+ ?
3
Large_Comm
i
+ ?
i
Community Size Demographics
Variable Predicted sign Coefficient p-value Coefficient p-value
4 5 . 0 + e z i S 0.01 0.21 0.17
Size * Large_Comm + 0.52 0.02 1.15 0.00
Large_Comm – – 3.35 0.00 – 4.88 0.00
Pseudo R
2
0.07 0.10
N 185 185
Recordkeeping is SCCS variable #149 (Writing and Records) with categories defined on the y-axis in Panel A. In the model using Community
Size as an independent variable, Size is SCCS variable #63 (Community Size) with categories defined on the x-axis in Panel A. Large_Comm is
an indicator variable equal to 1 if Community Size is greater than 200 persons and equal to 0 if Community Size is less than 200 persons.
Demographics refers to the factor estimated in the lower portion of Panel B. For purposes of this test, the societies were ranked on
Demographics and placed into eight groups where the total number of societies assigned to each group was identical to the number of societies
in each of the eight groups corresponding to different values for Community Size. When Demographics is used as an independent variable, Size
equals a number from one to eight corresponding to which of the ranked groups that the society has been assigned (lowest = 1 and highest = 8).
Large_Comm in this model equals 0 if the society is in groups 1 – 3 and equals 1 if the society is in groups 4 – 8. The model is estimated using
Ordered Logit. p-Values are one-tailed when the si gn is directionally predicted and are based on heteroskedasticity-consistent standard errors
adjusted for residual correlation among observations belonging to the same major language family.
(D)
Fig. 4 (continued)
16
Also, the tally sticks used by the British Exchequer beginning several
centuries ago evolved into bills of exchange, which served a monetary
function in exchange (Robert, 1952, p. 80; Goetzmann & Williams, 2005).
S. Basu et al. / Accounting, Organizations and Society 34 (2009) 895–917 905
The other two clusters include Administrative Hierarchy
and Judiciary (cluster B) and Use of Credit (cluster C), which
are present in about 45% and 33% of the SCCS societies,
respectively. In each case, the frequency of these institu-
tions is statistically different from Recordkeeping at con-
ventional levels (p 6 0.01). These data indicate that more
advanced institutions like hierarchies, courts, and the
availability of credit beyond the family are less likely to ap-
pear in the earliest stages of an economy’s development
relative to recordkeeping, money and basic property rights.
The lower part of Fig. 5 shows the percentage of socie-
ties where a given institution is present for both societies
with and without recordkeeping. Each institution is pres-
ent more frequently in societies with recordkeeping than
societies without recordkeeping. Chi-square tests reject
the null of independence at p < 0.05 for each of the ?ve
comparisons. This pattern is consistent with institutional
co-evolution where multiple interdependent institutions
emerge as a society grows in size and complexity.
Overall, the evidence in Figs. 4 and 5 is consistent with
the hypothesis that recordkeeping is a foundational insti-
tution that emerges early as an economy expands. Further-
more, recordkeeping societies are more likely to develop
other exchange-supporting institutions.
The in?uence of recordkeeping on impersonal exchange
and division of labor
Recordkeeping promotes impersonal exchange
We require a proxy for the extent of impersonal ex-
change to test our second hypothesis, which is that record-
keeping promotes impersonal exchange. We develop a
multi-attribute measure based on the extent to which
impersonal exchange takes place in the society as well as
whether the society has other institutions that support
more extensive exchange. We use a broader measure of ex-
change because Smith (1776/1976, pp. 21–33) argues that
the extent of the market depends on an effective infra-
structure for storing and moving goods as well as a med-
ium of exchange.
Our measure of impersonal exchange is based on a fac-
tor analysis applied to Money and three other SCCS vari-
ables de?ned in Panel C of Appendix B. Intercommunity
Trade as a Food Source (SCCS variable #1) is a direct mea-
sure of impersonal exchange as re?ected in a society’s food
import levels. Similarly, Food Storage Via Surplus (SCCS var-
iable #21) measures the extent to which exchange is likely
to occur as a result of food production in excess of imme-
diate consumption needs. Land Transport (SCCS variable
#154) re?ects the extent to which a society has a transpor-
tation infrastructure necessary to support exchange with
more geographically distant areas. Money (SCCS variable
#155) was used in our previous analyses and measures
whether a medium of exchange that could support more
extensive impersonal exchange is available.
We perform a principal factor analysis using the com-
munalities among these four variables to extract underly-
ing dimensions, and we used a minimum Eigenvalue of
one to determine which factors to retain. Panel A of Table
1 shows that one factor with an Eigenvalue of 1.27 ac-
counts for 91% of the total variance of the four SCCS vari-
ables. Thus, we retain only one factor to specify Exchange.
The factor loadings in Panel B of Table 1 show that Money
is strongly associated with Exchange. Intercommunity Trade
Fig. 5. Cumulative % of SCCS societies where an institution is present at a given level of community size.
906 S. Basu et al. / Accounting, Organizations and Society 34 (2009) 895–917
as a Food Source, the most direct measure of exchange, and
Land Transport also exert a strong effect. Food Surplus Via
Storage shows the weakest association.
We estimate the following empirical model of the rela-
tion between Recordkeeping and Exchange:
Exchange
i
= a
0
+ b
1
RK_B
i
+ b
2
Large_Comm
i
+ b
3
RK_B
i
*
Large_
Comm
i
+ b
4
AgriculturalPotential
i
+ b
5
Climate
i
+ b
6
Region
i
+e
i
.
Exchange is the factor derived from the factor analysis in
Panels A and B of Table 1. RK_B is a transformed binary ver-
sion of SCCS #149 (Writing and Records) where 0 indicates
recordkeeping of any kind is absent and 1 indicates record-
keeping of any kind is present. We use a 0–1 indicator for
Recordkeeping since we expect a bi-directional causal rela-
tion between the quantity and complexity of recordkeep-
ing and exchange. Because Recordkeeping appears early in
development (see Fig. 5), we expect that these effects
(and any resultant endogeneity problems) will be lessened
with use of a 0–1 variable. Large_Comm is an indicator var-
iable that equals 0 when Community Size (SCCS #63) is less
than 200 persons and equals 1 when Community Size is
greater than or equal to 200 persons. Under our second
hypothesis, we expect that b
1
and b
3
will be positive.
The model includes three control variables (de?ned in
Panel C of Appendix B): (1) Agricultural Potential (SCCS
#921) of the society’s region, (2) Climate (SCCS #857) is
an ordinal variable re?ecting open access to rich ecological
resources, and (3) Region (SCCS #200) represents a series of
0–1 variables for a society’s location within one of six ma-
jor world regions. These variables are included to capture
cross-society differences in resource endowments, which
Diamond (1997) argues are prerequisites for economic
development.
The OLS models are estimated using the 182 SCCS soci-
eties where data are available for all variables. p-Values are
one-tailed when a signed prediction is present and are
based on heteroskedasticity-consistent standard errors ad-
justed for residual correlation among observations belong-
ing to the same major language family.
17
Estimation results are shown in Panel C of Table 1. The
?rst column shows results for the model when only RK_B
and the control variables are included. The estimated coef-
?cient, b
1
, equals 0.51, which is signi?cantly different from
zero at p < 0.00 (one-tailed). Recordkeeping increases the
adjusted R
2
of the OLS model including only control vari-
ables from 0.25 (untabulated) to 0.32. b
1
remains positive
(0.26) and signi?cant at p < 0.02 for the full model (shown
in the second column) and b
3
equals 0.57, which is signif-
icant at p < 0.02 (one-tailed). This latter effect suggests that
the presence of recordkeeping exerts a substantially larger
effect on the extent of impersonal exchange in larger soci-
eties. Results in the right-most column demonstrate that a
signi?cant relation between recordkeeping and exchange
persists when Demographics is included as an additional
control for social complexity. This evidence supports our
second hypothesis that recordkeeping supports an expan-
sion in impersonal exchange.
The in?uence of impersonal exchange on division of labor
Thus far we have seen that recordkeeping becomes
more prevalent once a society has reached modest size lev-
els, recordkeeping emerges early relative to other basic
Table 1
Tests of association between recordkeeping and impersonal exchange.
A: Exchange factor estimation (Iterated principal factors)
Factor Eigenvalue Proportion explained Cumulative explained
1 1.27 0.91 0.91
2 0.08 0.06 0.97
3 0.04 0.03 1.00
4 –0.00 –0.00 1.00
B: Factor loadings for exchange (One factor retained)
Money 0.84
Intercommunity trade
as a food source
0.51
Land transport 0.50
Food surplus via
storage
0.24
C: Association between recordkeeping and exchange (p-values)
Variable Predicted
sign
Model with
Demographics as a
control
RK_B + 0.51
(0.00)
0.26
(0.02)
0.13
(0.07)
Large_Comm –0.01
(0.93)
–0.41
(0.00)
RK_B
*
Large_Comm + 0.57
(0.01)
0.48
(0.01)
Demographics 0.56
(0.00)
Adj. R
2
0.32 0.38 0.63
N 182 182 182
Model: Exchange
i
= a
0
+ b
1
RK_B
i
+ b
2
Large_Comm
i
+ b
3
RK_B
*
Large_
Comm
i
+ Controls + e
i
.
The sample used in Panel C includes the 182 SCCS societies where data for
all variables were available. The models are estimated using OLS.
Exchange is the factor derived from the factor analysis in Panels A and B.
RK_B is a transformed binary version of SCCS #149 (Records and Writing)
where 0 indicates recordkeeping of any kind is absent and 1 indicates
recordkeeping of any kind is present. Large_Comm is an indicator variable
that equals 0 when Community Size (SCCS #63) is less than 200 persons
and equals 1 when Community Size is greater than 200 persons. Demo-
graphics refers to the factor estimated in Fig. 4, Panel C. Controls included
in all models are: (1) Agricultural Potential (SCCS #921) of the society’s
region, (2) Climate (SCCS #857), a 6-scale categorical variable ordered in
terms of open access to rich ecological resources, and (3) Region (SCCS
#200) dummies that represent 0–1 variables based on the society’s
location within one of six major world regions. p-Values are one-tailed
when a signed prediction is present and are based on heteroskedasticity-
consistent standard errors adjusted for residual correlation among
observations belonging to the same major language family.
17
An analysis of the SCCS data indicates that Recordkeeping is subject to
stronger patterns of cultural and historical diffusion than other SCCS
variables. Each SCCS culture is assigned a number from 1 to 186 where the
societies are ordered according to geographical proximity and cultural
similarity. Because societies with close geographical proximity are likely
ones where cultural diffusion may still be a prominent force, the correlation
between adjacent numbered societies within the SCCS measures the extent
to which the pinpointing process did not completely eliminate cross-
sectional dependence (Murdock & White, 1969). Recordkeeping displays
strong correlation (q = 0.30) when comparing adjacent neighbors within
the SCCS database. Thus, we estimate all models using heteroskedasticity-
consistent standard errors adjusted for residual correlation among obser-
vations belonging to the same major language family.
S. Basu et al. / Accounting, Organizations and Society 34 (2009) 895–917 907
institutions as societies grow larger, and that the emer-
gence of recordkeeping is associated with increased imper-
sonal exchange. Our ?nal hypothesis is, consistent with
Smith (1776/1976, p. 21), that expanded opportunities
for impersonal exchange are associated with increasingly
specialized division of labor within an economy.
Testing this hypothesis requires ?rst that we specify a
measure of a society’s division of labor. As with our tests
examining market exchange, we specify a measure of divi-
sion of labor using factor analysis applied to multiple SCCS
variables likely related to the underlying construct. Most
SCCS societies are heavily dependent on agriculture and
depend far less on industrial manufacturing and complex
services than modern Western economies. We thus con-
struct Division of Labor using measures of the extent to
which complex agriculture and other forms of occupa-
tional specialization are present in the society.
Five SCCS variables serve as inputs to a factor analysis to
specify Division of Labor, as de?ned in Panel C of Appendix B.
The two agricultural measures include: (1) Agriculture (SCCS
variable #151), which takes on ?ve possible values ranging
from 1 indicating no agriculture is present to 5 indicating
that intensive agriculture provides the primary foodsources
for the society, and (2) Intensity of Cultivation (SCCS variable
#232), which takes on six possible values ranging from 1
(‘‘no agriculture”) to 6 (‘‘intensive irrigated agriculture”).
Technological Specialization (SCCS variable #153) is a vari-
able re?ecting whether specialist potters, loom weavers,
or metalworkers are present in the society. Administrative
Hierarchy (SCCS variable #91) measures the extent to which
decision-making in the society is delegated to heads of sub-
groups and Class Strati?cation (SCCS variable #270) mea-
sures the extent to which social status arises fromresources
or power possessed by an individual or group.
Panel A of Table 2 shows that one factor with an Eigen-
value of 2.86 accounts for 81% of the total variance of the
?ve SCCS variables used; we therefore retain only one fac-
tor to specify Division of Labor. Panel B of Table 2 indicates
that Intensity of Cultivation and Agriculture exert more
in?uence on Division of Labor than do Technological Special-
ization, Administrative Hierarchy, and Class Strati?cation.
We estimate a model of the relation between Division of
Labor and Exchange as follows:
Division of Labor
i
= a
0
+ b
1
Exchange
i
+ Controls
i
+ e
i
Division of Labor is the factor obtained from the analysis
described in Panels A and B of Table 2 and Exchange is the
factor derived from the analysis in Panels A and B of Table
1. The model includes the same control variables used in
the regressions in Table 1: Agricultural Potential (SCCS
#921), Climate (SCCS #857), and Region (SCCS #200); de?-
nitions are in Appendix B. We hypothesize that b
1
will be
positive. The sample includes the 180 SCCS societies for
which data for all variables are available. Models are esti-
mated using OLS and p-values are one-tailed and are based
on heteroskedasticity-consistent standard errors adjusted
for residual correlation among observations belonging to
the same major language family.
The results in Panel C of Table 2 support the hypothesis
that Exchange exhibits a positive association with Division
of Labor. The coef?cient on Exchange (b
1
) equals 0.44
(p < 0.00, one-tailed), and remains signi?cantly positive
(b
1
= 0.29, p < 0.00) when Community Size is used as an
additional control. The relation between Division of Labor
and Exchange becomes weakly signi?cant (p = 0.06) when
Demographics is added as a control. This is the artifact of Ex-
change being strongly correlated with variables included in
Demographics other than Community Size. In particular,
variables capturing the dispersion and density of a popula-
tion such as Settlement Patterns (SCCS variable #234) and
Population Density (SCCS variable #1130) re?ect the loca-
tion of individuals within a society who can take advantage
of exchange opportunities. The ?nal two columns indicate
that the coef?cient on RK_B is positive and statistically sig-
ni?cant (coef?cient = 0.30, p = 0.04) when included with
Exchange as an independent variable. In contrast, RK_B is
Table 2
Tests of association tests between division of labor and impersonal
exchange.
A: Division of labor factor estimation (Iterated principal factors)
Factor Eigenvalue Proportion explained Cumulative explained
1 2.86 0.81 0.81
2 0.62 0.17 0.98
3 0.06 0.02 1.00
4 0.01 0.00 1.00
5 –0.00 –0.00 1.00
B: Factor loadings for division of labor (One factor retained)
Intensity of cultivation 0.87
Agriculture 0.82
Technological
specialization
0.71
Administrative
hierarchy
0.62
Class strati?cation 0.62
C: Association of division of labor with exchange (p-values)
Variable Predicted
sign
With
Community
Size as a
control
With
Demographics
as a
control
With RK_B
Exchange + 0.44
(0.00)
0.29
(0.00)
0.10
(0.06)
0.39
(0.00)
RK_B + 0.30
(0.04)
0.50
(0.00)
Adj. R
2
0.50 0.60 0.69 0.52 0.43
N 180 180 180 180 180
Model: Division of Labor
i
= a
0
+ b
1
Exchange
i
+ Controls + e
i
.
The sample used in Panel C includes only the 174 SCCS societies where
data for all variables were available. The models are estimated using OLS.
Division of Labor is the factor derived from the factor analysis in Panels A
and B. Exchange is the factor for the extent of market exchange derived
from the factor analysis in Panels A and B of Table 1. Controls include: (a)
Agricultural Potential (SCCS #921) of the society’s region, (b) Climate (SCCS
#857) is a 6-scale categorical variable ordered in terms of open access to
rich ecological resources, and (c) Region (SCCS #200) dummies represent
0–1 variables based on the society’s location within one of six major
world regions. Community Size is used as an additional control variable in
the second column and Demographics is used as the additional control
variable in the third column in lieu of Community Size. Community Size is
SCCS #63 and Demographics is the factor estimated in Panel C of Fig. 4.
RK_B is a transformed binary version of SCCS #149 (Writing and Records)
where 0 indicates recordkeeping of any kind is absent and 1 indicates
recordkeeping of any kind is present. RK_B is added as an additional
regressor in the fourth column. The ?nal column shows the results using
RK_B as an independent variable on a stand-alone basis. p-Values are one-
tailed and are based on heteroskedasticity-consistent standard errors
adjusted for residual correlation among observations belonging to the
same major language family.
908 S. Basu et al. / Accounting, Organizations and Society 34 (2009) 895–917
positive and highly signi?cant when included on a stand-
alone basis (coef?cient = 0.50, p < 0.00), and adds explana-
tory power to a baseline regression including only the con-
trol variables – adjusted R
2
increases from 0.37
(untabulated) to 0.43. The 40% decline from 0.50 to 0.30
for the coef?cient on RK_B suggests that the effect of
Recordkeeping on Division of Labor ?ows partially through
Exchange.
The causal force hypothesized to act upon division of la-
bor when impersonal exchange is possible is that individu-
als invest in human capital that affords thema comparative
advantage in producing specialized goods and services.
More generally, this is part of a broader investment pattern
within the society where the emergence of impersonal ex-
change leads to greater investment in all forms of capital –
e.g., tangible capital and improvements to land (Smith
1776/1976, pp. 351–371). Accordingly, we also examined
the relation between impersonal exchange and the level
of capital stock accumulated by a society.
We estimate a measure of a society’s accumulated capi-
tal stock using six SCCS variables that re?ect physical, hu-
man, and social capital. Three measures pertain to
accumulated physical capital: Large or Impressive Structures
(SCCS #66), Resource Base (SCCS #859), and Cropping Index
(SCCS #1128). Large or Impressive Structures re?ects the ex-
tent to which physical structures have been erected in the
society and ranges from 1 (‘‘none”) to 6 (‘‘economic or
industrial buildings”). Resource Base measures whether
complementary techniques or assets have been developed
that improve agricultural productivity. This variable takes
on three possible values with a code of 1 representing
‘‘low resources (ex, hunting, gathering, ?shing)” and a code
of 3representing ‘‘highresources (ex, advancedhorticulture
with metal hoes, intensive agriculture with plow, pastoral-
ism).” Cropping Index is a measure of land utilization in agri-
culture ranging from 1 (‘‘no agriculture or con?ned to non-
food crops”) to 6 (‘‘100% or more of land used per year”).
Education is a measure of accumulated human capital
and represents the sum of several variables (SCCS #’s
425, 426, 427, and 428) measuring the extent to which
children of the society receive training and education. Each
of these four variables is coded on a six-point scale from 1
(‘‘informal training, with minimal guidance”) to 6 (‘‘formal
schooling typical and frequent”). Political Autonomy (SCCS
#81) and Military Success (SCCS #908) are included as mea-
sures of the extent to which the society has developed a
government that protects against external threats. The
ability to repel external threats can enhance productivity
by reducing the extent to which another group expropri-
ates the fruits of labor by a society. Political Autonomy is
coded on a six-point scale that ranges from 1 (‘‘dependent
totally”) to 6 (‘‘fully autonomous”). Military Success is
coded on a scale that ranges from 1 (‘‘no – its bound-
aries/population are shrinking”) to 4 (‘‘yes – its bound-
aries/population are expanding”).
These six variables provide the inputs to a factor analy-
sis used to identify a variable we label Capital Stock. As be-
fore, we conduct a principal factor analysis using the
communalities among these variables, along with a mini-
mum Eigenvalue of one to determine which factors to re-
tain. Panel A of Table 3 shows that one factor with an
Eigenvalue of 1.85 accounts for 71% of the total variance
of the six SCCS variables used to identify Capital Stock. Pa-
nel B of Table 3 shows that Cropping Index is the most in?u-
ential variable and Education, Resource Base, and Large
Impressive Structures are also of importance. The political
variables, Political Autonomy and Military Success, are the
least important variables in specifying Capital Stock.
We estimate a model of the relation between Capital
Stock and Exchange as follows:
Capital Stock
i
= a
0
+ b
1
Exchange
i
+ Controls
i
+ e
i
Capital Stock is the variable identi?ed in the factor anal-
ysis in Panels A and B of Table 3 and all other variables are
as previously de?ned for purposes of our prior analyses.
Each model includes the control variables used previously
(Agricultural Potential, Climate, and Region). De?nitions of
all variables are provided in Panel E of Appendix B. The
Table 3
Tests of association between Capital Stock and Exchange.
A: Capital stock factor estimation (Iterated principal factors)
Factor Eigenvalue Proportion explained Cumulative explained
1 1.85 0.71 0.71
2 0.47 0.18 0.89
3 0.18 0.07 0.96
4 0.08 0.03 0.99
5 0.02 0.01 1.00
6 –0.00 –0.00 1.00
B: Factor loadings for capital stock (One factor retained)
Cropping index 0.75
Investment in
education
0.58
Resource base 0.55
Large impressive
structures
0.53
Political autonomy 0.38
Military success 0.29
C: Association of capital stock with exchange (p-values)
Variable With
RK_B
With
Community
Size as a
control
With
Demographics
as a control
Exchange 0.56
(0.00)
0.21
(0.00)
0.20
(0.00)
0.19
(0.00)
0.17
(0.01)
Division of
labor
0.65
(0.00)
0.64
(0.00)
0.60
(0.00)
0.58
(0.00)
RK_B 0.07
(0.24)
Adj. R
2
0.53 0.78 0.78 0.79 0.79
N 143 143 143 143 143
Model: Capital Stock
i
= a
0
+ b
1
Exchange
i
+ Controls + e
i
.
The sample used in Panel C includes the 143 SCCS societies where data for
all variables were available. The models are estimated using OLS. Capital
Stock is the factor derived from the factor analysis in Panels A and B,
Exchange is the factor for the extent of impersonal exchange derived from
the factor analysis in Panels A and B of Table 1, and Division of Labor is the
factor derived from the factor analysis in Panels A and B of Table 2. RK_B is
a transformed binary version of SCCS #149 (Records and Writing) where 0
indicates recordkeeping of any kind is absent and 1 indicates record-
keeping of any kind is present. Controls included in all models are: (1)
Agricultural Potential (SCCS #921) of the society’s region, (2) Climate (SCCS
#857) is a 6-scale categorical variable ordered in terms of open access to
rich ecological resources, and (3) Region dummies represent 0–1 variables
based on the society’s location within one of six major world regions.
Community Size equals SCCS variable #63 and Demographics is the factor
estimated in Panel C of Fig. 4. p-Values are one-tailed and based on het-
eroskedasticity-consistent standard errors adjusted for residual correla-
tion among observations belonging to the same major language family.
S. Basu et al. / Accounting, Organizations and Society 34 (2009) 895–917 909
sample used to estimate Models (3), (4), and (5) includes
the 143 SCCS societies where data for all variables are
available.
The hypothesis of interest is that Capital Stock is associ-
ated with the emergence of impersonal exchange after
controlling for other factors. Thus, we expect that b
1
will
be positive. As with the previous models, OLS estimation
is used, p-values are one-tailed, and heteroskedasticity-
consistent standard errors adjusted for residual correlation
among observations belonging to the same major language
family are used.
The evidence in Panel C supports the hypothesis that
cross-society differences in Capital Stock are associated with
differences in Exchange. The ?rst column of results indicates
that the coef?cient on Exchange equals 0.56 (p < 0.00). Fur-
ther, this result is highly robust. Exchange remains signi?-
cant after controlling for factors already captured by
Division of Labor (see second column). The third column
indicates that RK_B shows no signi?cant association with
Capital Stock after controlling for Exchange and Division of
Labor. This suggests that recordkeeping’s role is one of an
enabling technology that supports expanded exchange
and division of labor.
18
The ?nal two columns demonstrate
that the relation between Exchange and Capital Stock is not
capturing purely an effect due to size and demographic com-
plexity. The economic institutions of Exchange and Division of
Labor add considerable explanatory power to a baseline mod-
el that includes only the geographical control variables which
generate an (untabulated) adjusted R
2
of 0.30.
Viewed collectively, the evidence presented in Tables 2
and 3 provides strong support for our ?nal hypothesis that
the expansionof impersonal exchange facilitated by record-
keeping is also associated with increasingly specialized
division of labor and greater overall investment in physical,
tangible, and political capital. These results accord with
prior experimental evidence that recordkeeping is associ-
ated with greater trust (Basu et al., 2009), and that trust is
associated with investment and economic growth, both
analytically andincross-country growthregressions (Knack
& Keefer, 1997; Zak & Knack, 2001). In other words, record-
keeping enables strangers to trust each other in complex
intertemporal exchange, which then facilitates division of
labor and greater productivity, which in turn enable greater
investment and even faster economic growth.
Conclusions and implications
Our evidence suggests that recordkeeping is more likely
in large groups that cannot sustain cooperative interaction
based solely on mental memory, and that recordkeeping,
like money and inheritance of land, emerges at relatively
early stages of an economy’s development. The emergence
of recordkeeping precedes the appearance of a judiciary,
administrative hierarchies and the extension of credit, sug-
gesting that recordkeeping is a foundational institution.
Our evidence also suggests that economies where record-
keeping is possible are characterized by more extensive
impersonal exchange. Consistent with hypotheses by Smith
(1776/1976) and Stigler (1951), we also ?nd that the level
of specialization in division of labor and accumulated cap-
ital are strongly in?uenced by the extent of the market.
These ?ndings suggest that the basic accounting function
of recordkeeping is a precursor to economic development
through impersonal exchange and division of labor.
More broadly, our evidence is consistent with the
hypothesis that transaction records external to the human
brain are necessary to extend the scale of human coopera-
tion from small primitive groups to large-scale modern
societies characterized by extensive market exchange and
complex division of labor (Basu & Waymire, 2006). Our
analysis also broadly accords with conjectures offered by
an earlier generation of scholars (i.e., Sombart, Weber,
Schumpeter, and von Mises) that capitalist economic
development would be impossible without accounting
institutions like double-entry bookkeeping (Carruthers &
Espeland, 1991; Most, 1972). Thus, considerably more re-
search on how recordkeeping and more advanced analysis
of accounting’s transactional data promote economic
development is warranted.
Economic development varies considerably across conti-
nents andcountries (e.g. World Bank, 2006), as well as with-
in countries among different ethnic and sociolinguistic
groups. Our evidence is consistent with De Soto’s (2000)
argument that accurate property records are a prerequisite
for the success of capitalistic societies. The crucial role of
veri?able transaction records for legal enforcement of prop-
erty rights is explicitly indicated in legal codes ranging from
the ancient Code of Hammurabi (circa 1750 BCE) through
Athenian, Roman and French legal codes to the recent Sar-
banes-Oxley Act of 2002 (Basu&Waymire, 2006). Thus, ver-
i?able transaction records are a necessary part of the
foundations that lie beneaththe exchange-supporting insti-
tutions upon which capitalist economies have been built.
Accounting is likely an ecologically rational institution
that coordinates economic interaction through market ex-
change (Waymire, 2009; Waymire & Basu, 2007). To adapt
a metaphor from Simon (1990), historical cost accounting
records and today’s exchange agreements are like two
blades of scissors that have become increasingly effective
together over time through co-evolution. Institutional
changes such as ‘fair value’ accounting that overwrites his-
torical records of consummated transactions and erodes
the quality of memory inherent in records may make the
scissors less effective unless a matching blade is pro-
duced.
19
Contracts and regulations rely on a set of common
expectations about how performance is measured, and
wholesale changes to this performance measure may entail
‘‘unintended consequences.” One ultimate consequence of
less effective recordkeeping may be the decay of economic
institutions responsible for wealth generation in modern
developed societies.
18
In contrast, the coef?cient on RK_B equals 0.51 (p = 0.00) with Exchange
and Division of Labor excluded from the model.
19
Similarly, Ball (2001) argues that accounting harmonization is unlikely
to produce higher quality ?nancial reports unless contracts and enforce-
ment mechanisms are also changed to provide incentives for better
reporting. While Ball focuses on whether developing countries can improve
their accounting quality, we worry that developed countries are in danger
of reducing their economic effectiveness.
910 S. Basu et al. / Accounting, Organizations and Society 34 (2009) 895–917
Acknowledgements
We acknowledge helpful comments received fromMark
Kohlbeck, Eric Press, Richard Sansing, Denise Schmandt-
Besserat, two anonymous reviewers, participants in the
Emory University Anthropology Department Workshop
Series on Human Behavior and Evolution, the 2007 Univer-
sity of Oklahoma Accounting Research Conference, the
Fifth Accounting History International Conference, the
2008 FARS Midyear Meeting, the 2008 Annual Meeting of
the International Society for New Institutional Economics,
the 2008 Annual Meeting of the American Accounting
Association, and accounting seminars at Arizona State, Chi-
cago, City University of Hong Kong, Florida, Minnesota,
Northwestern, Southern California, Temple, and Wisconsin.
The Goizueta Business School at Emory University pro-
vided ?nancial support for this research.
Appendix A
Construction. of the Standard Cross-Cultural Sample (SCCS)
The Standard Cross-Cultural Sample (SCCS) provides a
cross-section of ethnographic ‘‘snapshots” that we use to
investigate cross-cultural variation in recordkeeping prac-
tices throughout the world. These pictures capture multi-
ple elements of a culture or society in a location at a
speci?c point in time. Because these data have not been
previously used in the accounting literature, we describe
the construction of this sample in some depth.
Murdock and White (1969) constructed the SCCS to
standardize the data used in cross-cultural research and
facilitate statistical analysis.
20
Paying careful attention to
ethnographic distributions, Murdock and White identi?ed
186 cultural provinces in dispersed geographical locations
and time periods (including two from before the Common
Era). Murdock and White then chose one society to repre-
sent each cultural province, picking the earliest well-de-
scribed societies to the extent practicable. The SCCS
societies include contemporary hunter-gatherers, early his-
toric states, and contemporary industrial societies. This wide
coverage re?ects Murdock and White’s (1969) conscious
decision to mitigate biases that favored societies with Eng-
lish language ethnographic sources.
A major purpose in constructing the SCCS was to in-
crease the extent to which statistically valid inferences
could be drawn from ethnographic data. Speci?cally, prior
studies had often used data that lacked independence.
Cross-cultural correlation in cultural practices arises from
the diffusion of those practices among cultures with a
common heritage. Anthropologists have recognized this
problem (a.k.a. Galton’s Problem) for over a century. The
SCCS was speci?cally designed to minimize cross-society
dependence because statistical corrections such as Fama-
MacBeth or White-Huber standard errors had yet to be de-
vised. Murdock and White (1969) addressed this problem
by ‘‘pinpointing” their societies to speci?c locales and
dates. The pinpointing of societies permitted selection of
cultures with weaker cultural and historical diffusion rela-
tionships – i.e., the SCCS was constructed to maximize
independence in terms of cultural and historical origin
while preserving a large enough sample to permit suf?-
ciently powerful statistical tests.
Murdock (1967) initiated the pinpointing process when
he analyzed nearly 1300 societies chosen for the complete-
ness of their ethnographic coverage to construct his Ethno-
graphic Atlas. He classi?ed these societies into clusters
based on the similarity of the cultures and categorized
groups of clusters into 200 ‘‘sampling provinces” (Mur-
dock, 1968; Murdock & White, 1969). From the initial
200 sampling provinces, two had no culture that could be
accurately pinpointed to a particular locale and date, two
were split in half, and 14 others were dropped because
they were too similar to others in the sample.
Murdock and White (1969) then identi?ed that culture
from each of the 186 sampling provinces with the earliest
period of satisfactory ethnographic coverage unless signif-
icantly richer data were available for a later period. The
186 cultures selected in this step comprise the SCCS. The
cultures in SCCS are assigned a number from 1 to 186,
which facilitates statistical identi?cation of those cultural
practices that originate from a common cultural heritage.
This is done because societies with close geographical
proximity are likely ones where cultural diffusion may still
be a prominent force. Thus, the correlation between adja-
cent societies within the SCCS measures the extent to
which the pinpointing process did not completely elimi-
nate Galton’s Problem.
The initial study using the SCCS coded 22 variables re-
lated to subsistence economy and related practices (Mur-
dock & Morrow, 1970). The free electronic database
includes a bibliography of ethnographic sources for each
society-year, ranked on a scale from 1 ‘‘Principal Author-
ity(ies)” to 6 ‘‘Sources to be Avoided” (White, 1986). The
hard copy archives include coding sheets indicating which
individual sources were relied on most for each culture for
coding different groups of variables. Researchers who use
the database code new variables and these additional vari-
ables are added to the database as a result. Not all 186 soci-
eties are coded for each new variable as some researchers
elected to code data for only a subset of the SCCS cultures.
For example, many variables are coded for only 93 cultures
suggesting, for example, a sampling scheme such as using
every other culture in the database.
Thus, eachnewstudyincreases thedepthof thedatabase.
There are presentlymore than2000categorical variables (as
of 2007) coded nominally or ordinally by over 60 different
studies.
21
Unlike the usual market studies, the data we use
are limited to only one observation per culture; thus, SCCS
does not provide a pooled time-series cross-sectional data
set. The SCCS was designed to ensure that standard errors
are not in?ated by multiple observations from the same unit.
20
Previous researchers had tended to analyze their own selection of
speci?c societies, which often was based on small samples that were not
comparable across studies. The SCCS helped standardize researchers’ choice
of societies and has improved cross-study comparability.
21
An online journal, World Cultures, founded by Douglas White in 1985,
maintains, re?nes and expands the SCCS. The journal is available in paper
and CD-ROM as well as for free over the Internet. The journal can be
accessed athttp://eclectic.ss.uci.edu/~drwhite/worldcul/world.htm.
S. Basu et al. / Accounting, Organizations and Society 34 (2009) 895–917 911
Appendix B
De?nitions of SCCS variables used in empirical tests
SCCS variable Coding
Panel A: SCCS variables ?rst introduced in ‘‘The history of transaction records and recordkeeping in SCCS societies”
V149: Writing and Records 1 = None
Variable #149 coded by Murdock and Provost (1973, pp. 379–
380)
2 = Mnemonic devices
3 = Nonwritten records
4 = True writing; no records
5 = True writing; records
Panel B: SCCS variables ?rst introduced in ‘‘Recordkeeping emerges as group size increases”
V18: Credit source 1 = Personal loans between friends or relatives
Variable #18 ?rst coded by Murdock and Morrow (1970, p. 306) 2 = Internal money lending specialists
3 = External money lending specialists
4 = Banks or comparable institutions
V63: Community size 1 = 50,000
V89: Judiciary 1 = Absent
Variable #89 ?rst coded by Tuden and Marshall (1972, p. 441) 2 = Not local
3 = Executive
4 = Appointed by executive
5 = Priesthood
6 = Hereditary
V91: Administrative hierarchy 1 = Absent
Variable #91 ?rst coded by Tuden and Marshall (1972, pp. 441–
442)
2 = Popular Assemblies
3 = Heads of kin groups
4 = Heads of decentralized territorial divisions
5 = Heads of centralized territorial divisions
6 = Part of centralized system
V155: Money 1 = None
Variable #155 coded by Murdock and Provost (1973, p. 381) 2 = Domestically usable particles
3 = Alien currency
4 = Elementary forms
5 = True money
V234: Settlement patterns 1 = Nomadic or fully migratory
Variable #234 ?rst coded by Murdock (1967, p. 159) 2 = Seminomadic
3 = Semisedentary
4 = Compact but impermanent settlements
5 = Neighborhoods of dispersed family homesteads
6 = Separated hamlets, forming a single community
7 = Compact and relatively permanent settlements
8 = Complex settlements
V278: Inheritance of land
22
0 = Absence of individual property rights or rules
Variable #278 ?rst coded by Murdock (1967, p. 167) 1 = Inheritance based on familial ties
V1122: Population size 1 = 10–99
912 S. Basu et al. / Accounting, Organizations and Society 34 (2009) 895–917
Appendix B (continued)
SCCS variable Coding
Variable #1122 coded by Douglas White and is described in
Standard Cross-Cultural Codes, edited by White, Burton, Divale,
Gray, Korotayev, and Khalturina at:http://eclectic.ss.uci.edu/
~drwhite/courses/SCCCodes.htm
2 = 100–999
3 = 1000–9999
4 = 10,000–99,999
5 = 100,000+
6 = 1,000,000+
7 = 10,000,000+
8 = 100,000,000+
V1130: Population density 2 = less than 1 per square mile
Variable #1130 coded by Pryor (1984) 3 = 1–4.9 per square mile
4 = 5–24.9 per square mile
5 = 25–99.9 per square mile
6 = 99–499.9 per square mile
7 = 500 or more per square mile
Panel C: SCCS variables First Introduced in ‘‘The in?uence of recordkeeping on impersonal exchange and division of labor”
Variables used to measure Exchange
V1: Intercommunity trade as food source 1 = No trade
Variable #1 coded by Murdock and Morrow (1970) 2 = No food imports
3 = Salt and minerals only
4 =
Adam Smith hypothesized that impersonal exchange was necessary for a society to develop
specialized division of labor and create wealth. Douglass North and Vernon Smith argue
that successful developed economies are the result of institutions. We hypothesize and
provide evidence from ethnographic data that the basic accounting technology of recording
transactions is associated with more extensive impersonal exchange and increased specialization
in the division of labor. Our intuition is that extensive impersonal exchange
requires reliable memory of trading partners’ past behavior to sustain trust and encourage
reciprocity when a group expands beyond the size of traditional hunter-gatherer groups.
Our findings are consistent with the hypothesis that transaction records are necessary
for the emergence of complex economies as suggested by the archaeological evidence of
recordkeeping in Mesopotamian societies 10,000 years ago.
Memory, transaction records, and The Wealth of Nations
Sudipta Basu
a
, Marcus Kirk
b
, Greg Waymire
c,
*
a
Fox School of Business, Temple University, Philadelphia, PA 19122, USA
b
Fisher School of Accounting, University of Florida, Gainesville, FL 32611, USA
c
Goizueta Business School, Emory University, Atlanta, GA 30322, USA
a b s t r a c t
Adam Smith hypothesized that impersonal exchange was necessary for a society to develop
specialized division of labor and create wealth. Douglass North and Vernon Smith argue
that successful developed economies are the result of institutions. We hypothesize and
provide evidence from ethnographic data that the basic accounting technology of recording
transactions is associated with more extensive impersonal exchange and increased special-
ization in the division of labor. Our intuition is that extensive impersonal exchange
requires reliable memory of trading partners’ past behavior to sustain trust and encourage
reciprocity when a group expands beyond the size of traditional hunter-gatherer groups.
Our ?ndings are consistent with the hypothesis that transaction records are necessary
for the emergence of complex economies as suggested by the archaeological evidence of
recordkeeping in Mesopotamian societies 10,000 years ago.
Ó 2009 Elsevier Ltd. All rights reserved.
Introduction
The palest ink is better than the best memory.
Chinese Proverb
(K)nowledge of the past, the record of truths revealed
by experience, is eminently practical, as an instrument
of action and a power that goes to making the future.
Lord Acton
1
The past had not merely been altered, it had actually
been destroyed. For how could one establish even the
most obvious fact when there existed no record outside
your own memory?
George Orwell
2
Human economies vary considerably in scale, com-
plexity, and performance; some generate great wealth
while others remain mired in poverty. In The Wealth of
Nations, Smith (1776/1976, Book I, Chapter II, p. 17) ar-
gued that the growth of economies derived from exten-
sive impersonal exchange, which grew out of a human
‘‘propensity to truck, barter, and exchange one thing for
another.” Humans sustain cooperation better than other
primate species in part because we can remember and
communicate information about the cooperative acts of
others, which is a prerequisite for reciprocity and reputa-
tion formation (Axelrod & Hamilton, 1981; Nowak & Sig-
mund, 2005). Nevertheless, the evolved proclivities and
abilities of the brain that favor exchange and cooperation
can account for human groups only up to a size of about
200 persons (Dunbar, 1992). What role did the institution
of recordkeeping play in allowing some economies to cir-
cumvent the biological constraints of memory and there-
by expand impersonal exchange and produce great
wealth?
The institutions that societies use to govern economic
and social interaction have been suggested as necessary
0361-3682/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.aos.2009.07.002
* Corresponding author. Tel.: +1 404 727 6589; fax: +1 404 727 6313.
E-mail address: [email protected] (G. Waymire).
1
Inaugural lecture on the study of history, Cambridge University, June
1895, as cited in Collini, Winch, and Burrow (1983).
2
Orwell (1949/1961, p. 36).
Accounting, Organizations and Society 34 (2009) 895–917
Contents lists available at ScienceDirect
Accounting, Organizations and Society
j our nal homepage: www. el sevi er. com/ l ocat e/ aos
for economic development (North, 2005; Smith, 2008).
3
These institutions include legal codes that support property
rights and money that relaxes constraints inherent to barter
exchange, both of which date back thousands of years (Sag-
gs, 1989, pp. 156–175; Redish, 2003). Recordkeeping for ex-
change transactions is an even older institution found in the
?rst human settlements of Mesopotamia circa 8000 BCE
(Schmandt-Besserat, 1992). Humans ?rst invented writing
to keep records (Nissen, Damerow, & Englund, 1993), which
coincided in time (circa 3000 BCE) with the emergence of
the ?rst cities, underscoring the central importance of trans-
action records that objectively document the history of ex-
change. Accounting scholars have long recognized the
presence of recordkeeping in ancient societies (Carmona &
Ezzamel, 2007, provide an overview; see also Macve,
2002).
4
Despite this evidence, we have no parsimonious sci-
enti?c explanation and little broad evidence for the emer-
gence of transaction records and their role in economic
and social development.
Basu and Waymire (2006) hypothesize that transaction
records emerge to symbolically represent the history of ex-
change in a more permanent manner external to the hu-
man brain. External records lessen the chances of
individual memory failure and are valuable in tracking a
trading partner’s past behavior as a basis for current deci-
sions. Records can also establish reliable social memory
and common knowledge useful to two or more parties in
structuring an exchange. For example, ‘‘hard” records that
are veri?ed by witnesses can make it ‘‘dif?cult for people
to disagree” later about whether past promises have been
ful?lled (Ijiri, 1975, p. 36). As recordkeeping evolves to en-
code more information, it enables drafting and enforcing
contracts that govern complex exchange transactions
across time and geographical boundaries – e.g., reliable re-
cords are needed to secure the property rights that facili-
tate modern capitalism (De Soto, 2000).
Basu and Waymire’s hypothesis parallels the hypothesis
that double-entry bookkeeping enabled modern capitalist
organization, which was advanced by Sombart (1919), We-
ber (1927), Joseph Schumpeter (1942), and Ludwig von
Mises (1949).
5
Both hypotheses assert a foundational role
for basic accounting institutions in enabling the emergence
of more complex forms of economic and social interaction.
The main distinguishing feature of Basu and Waymire’s
hypothesis is that basic accounting institutions like record-
keeping play this role early in the development of complex
societies rather than in emergent business organizations
within already developed societies.
Basu and Waymire (2006) predict that (1) recordkeep-
ing emerges because increasing exchange complexity
taxes the brain’s memory resources, and (2) accounting
records work in tandem with other fundamental institu-
tions to promote economic development. These predic-
tions can be investigated in several complementary
ways. One is to conduct experiments using neuroscienti?c
methods to investigate whether the human brain’s evalu-
ation of exchange parallels culturally evolved accounting
practices (Dickhaut, Basu, McCabe, & Waymire, 2009a,
2009b). A second avenue is to investigate whether the
causal links inherent in the Basu and Waymire (2006)
story are observed in a controlled experimental setting
(Basu, Dickhaut, Hecht, Towry, & Waymire, 2009). An-
other possibility, which we explore in this paper, is to
use naturally occurring data to test whether institutional
and economic development is greater in those societies
that have developed technologies for recording
transactions.
We present evidence on whether the association be-
tween recordkeeping technology and societal size and
complexity in ancient Mesopotamia generalizes to other
human societies, and how strongly the prevalence of
recordkeeping is associated with the presence of other
fundamental economic institutions. We explore these is-
sues using ?eld data collected by ethnographers and
archaeologists from a broad cross-section of human socie-
ties, and subsequently coded into machine-readable data
by Murdock and White (1969). Murdock and White’s
Standard Cross-Cultural Sample (SCCS) provides extensive
coded data for a variety of cultural variables – presently
over 2000 – for 186 societies selected to maximize the
cross-society independence of observations. SCCS societies
are ‘‘pinpointed” to speci?c dates that vary across socie-
ties. The SCCS data include a variable that measures a
society’s recordkeeping technology as well as numerous
measures of economic, social, and institutional
development.
We use the SCCS data to empirically evaluate whether
recordkeeping is a necessary institution for unleashing
the economic forces of impersonal exchange and division
of labor hypothesized by Smith to be the ultimate source
of economic wealth. We document that recordkeeping use
and sophistication is greater in societies that have sur-
passed the modest levels suggested by Dunbar (1992),
and also that recordkeeping is present as early as or ear-
lier in economic development than other basic institu-
tions such as money, property rights, hierarchical
organizations, a judiciary, and the use of credit. This evi-
dence suggests that recordkeeping is a precursor to rather
than the result of economic complexity. Our analyses also
demonstrate that the extent of impersonal exchange is
positively associated with the use of recordkeeping, and
that specialized division of labor and the level of capital
accumulation are more extensive in societies with greater
opportunities for market exchange. Collectively, our ?nd-
ings are consistent with the hypothesis that basic
accounting institutions are necessary but not suf?cient
to foster the extensive impersonal exchange and complex
3
We use ‘‘institution” in the broad sense of Douglass North (1991, p. 97):
‘‘Institutions are the humanly devised constraints that structure political,
economic and social interaction. They consist of both informal constraints
(sanctions, taboos, customs, traditions, and codes of conduct) and formal
rules (constitutions, laws, property rights). Throughout history, institutions
have been devised by human beings to create order and reduce uncertainty
in exchange. They evolve incrementally, connecting the past with the
present and the future; history in consequence is largely a story of
institutional evolution in which the historical performance of economies
can only be understood as part of a sequential story.”
4
Earlier examples include Littleton (1927), Robert (1952), Keister (1963),
Keister (1964), Mattessich (1987) and Baxter (1989).
5
This hypothesis and its empirical basis are discussed and debated by
Yamey (1949), Yamey (1964, Yamey (2005), Most (1972), Most (1973), and
Carruthers and Espeland (1991).
896 S. Basu et al. / Accounting, Organizations and Society 34 (2009) 895–917
economic interactions that characterize modern devel-
oped economies.
As with all studies, ours is subject to important cave-
ats. Because we examine naturally occurring data that
are subject to measurement error and are not designed
to measure intertemporal change within economies, our
analysis does not speak directly to causality in the rela-
tions between recordkeeping, exchange, and economic
development. Rather, our paper is an attempt to investi-
gate cross-cultural statistical associations between
recordkeeping and patterns of economic interaction ob-
served in societies at early stages of their development.
The evidence we present here complements the causal
evidence on the origins and consequences of basic
accounting institutions presented in Basu, Dickhaut,
Hecht, Towry, and Waymire (2009) and Dickhaut, Basu,
McCabe, and Waymire (2009a), Dickhaut, Basu, McCabe,
and Waymire (2009b). In combination, all these studies
suggest that veri?able historical transaction records play
a foundational role in the development of market econo-
mies. This study viewed in isolation is a necessary, but
not suf?cient, condition to support such a claim insofar
as it only provides evidence on the external validity of
the hypothesis.
We ?rst discuss the history of recordkeeping and de-
scribe how recordkeeping systems vary across the SCCS
societies. We then present our hypotheses. We next offer
evidence on the emergence of recordkeeping in the course
of social and economic development. Following that, we
present evidence on the association between recordkeep-
ing, impersonal exchange, and the division of labor. Con-
clusions from and implications of our ?ndings are
summarized in the paper’s ?nal section.
The history of transaction records and recordkeeping in
SCCS societies
Historical background on the origins and nature of transaction
records
Kohler’s (1952, p. 356) A Dictionary for Accountants de-
?nes a ‘‘record” as ‘‘A book or document containing or evi-
dencing some or all of the activities of an enterprise or
containing or supporting a transaction, entry, or account.”
Examples include ‘‘a book of account; subsidiary ledger;
invoice; voucher; contract; correspondence; internal re-
port; minute book.” This de?nition and examples evoke
images of a ‘‘paper trail” where writing symbolically por-
trays the types and quantities of goods exchanged, the per-
sons buying and selling the goods, and their obligations
under the exchange. Implicit in this characterization is that
a written language, numbers, and a system of weights and
measures already exist that shape the speci?cs of a trans-
action record.
Yet, this assumes more than existed in the earliest
recordkeeping systems. Indeed, the earliest transaction
records precede writing by thousands of years (Sch-
mandt-Besserat, 1992). The0 sophistication of records is
directly tied to how well various properties of exchange
goods can be categorized, described and measured. A
transaction record is merely an artifact that does not
per se require written language or the use of re?ned
weights and measures, although such things are useful
(Hutchins, 1999; Schmandt-Besserat, 1999). A record is
valuable because it allows a person to remember impor-
tant attributes of a transaction; two or more persons can
also use a transaction record to state their common
knowledge about the nature of the transaction. Preserv-
ing exchange information on records outside the brain
increases the life of common knowledge – that is, shared
understandings can be carried forward in time to a
greater extent than would be the case with the spoken
word.
The oldest-known recordkeeping systems were used in
Mesopotamian villages in 8000 BCE, with baked clay to-
kens and stones symbolizing individual agricultural com-
modities transferred between two parties (see the left-
hand side of Panel A in Fig. 1). By 4000 BCE, transfers of dif-
ferent manufactured goods were recorded with newer
complex tokens with various shapes and incisions (Nissen
et al., 1993; Schmandt-Besserat, 1995; Schmandt-Besserat,
1996). Tokens were sealed within baked hollow clay balls
(‘‘bullae”) by 3500 BCE. By 3200 BCE, personal seals of
the transacting parties and witnesses, along with indenta-
tions of the enclosed tokens, were impressed on the exte-
rior of the bulla before it was baked (see the right-hand
side in Panel A of Fig. 1).
Schmandt-Besserat (1995, p. 2100) sees the Mesopota-
mian token as a ‘‘mnemonic device by which to handle
and store an unlimited quantity of data without risking
the damages of memory failure.” Tokens and bullae
stored data on several transaction attributes long before
writing was invented. First, the unique token shapes al-
lowed the type and quantity of commodities exchanged
to be clearly identi?ed. Second, personal seals enabled
identi?cation of the exchange parties and witnesses from
the af?xed ‘‘signatures” on a bulla. Third, the external to-
ken impressions let everyone know the commodities in-
volved without breaking open the bulla. This was
advantageous since the bulla need then be broken only
at the transaction settlement, thereby facilitating com-
plex intertemporal exchange. Fourth, the baking of bullae
and their storage in a secure location rendered the infor-
mation ‘‘hard” both literally and also metaphorically in
that it made it ‘‘dif?cult for people to disagree” later (Ijiri
1975, p. 36).
Another non-written recordkeeping system was the In-
can quipu, ?rst documented after the Spanish conquered
that large and wealthy civilization around 1500 AD. The In-
can quipu (shown in panel B of Fig. 1) was a ‘‘knotted
string” on which transaction attributes were symbolized
by thread colors and length as well as the number of knots
and their location on the cord (Urton, 2002). Like the early
Mesopotamian tokens, the quipus kept track of tribute paid
by citizens to the Incan government (Assadourian 2002, p.
120; Pollock 1999, pp. 78–116).
The ‘‘tally stick” (see Panel C of Fig. 1), relied upon by
the English Treasury as recently as the 19th century, was
a receipt for tax collections by Royal agents (Robert,
1952; Baxter, 1989). The tally stick differed from tokens
and quipus in that some writing was employed (Robert,
S. Basu et al. / Accounting, Organizations and Society 34 (2009) 895–917 897
1952, p. 76). Tally sticks have been used all over the world,
at least as early as 500 BC by the Chinese (Goetzmann &
Williams, 2005), and as recently as the 1970s in rural
France (Ifrah, 2001). These pre-literate recordkeeping
systems are similar to the vouchers, contracts and other
written material that are used to generate ‘‘journal entries”
Fig. 1. Examples of recordkeeping technologies present in human history.
Picture sources:
Left hand side of Panel A in this ?gure: Schøyen Collection, Oslo and London.
Right hand side of Panel A in this ?gure: Louvre Museum, Paris, France.http://www.humanitiesinteractive.
org/ancient/mideast/index.html?collectionVar=AncientCulturesStop&pageVar=2.
Panel B in Fig. 1: Larco Museum, Lima, Peru.
Panel C in Fig. 1: Science Museum, London, UK.
Panel D in Fig. 1: Schøyen Collection, Oslo and London.
898 S. Basu et al. / Accounting, Organizations and Society 34 (2009) 895–917
in double-entry bookkeeping systems developed in 13th
century Italy. An ancient direct analog to the receipt is
the Cuneiform tablet (see Panel D of Fig. 1), which
Schmandt-Besserat (1992) has hypothesized to have
evolved from the earlier token system.
Academic accountants have long recognized the ancient
origins of accounting records (Keister, 1963; Keister, 1964;
Littleton, 1927; Mattessich, 1987; Mattessich, 2000). Inter-
est in the subject of ancient accounting has accelerated in
the past few decades, both within the accounting literature
and the archaeological literature (Carmona & Ezzamel,
2007; Hudson & Wunsch, 2004). The picture that emerges
from these literatures is that transaction records and the
accountability they render have played an important role
in ordering the economies of ancient societies in Egypt
and Mesopotamia (Hudson, 2004). Our study complements
these literatures by offering broader statistical evidence on
the emergence of recordkeeping and its role in economic
development.
Recordkeeping in SCCS societies
We use the Standard Cross-Cultural Sample (SCCS) con-
structed by Murdock and White (1969) in all of our empir-
ical analyses. Anthropologists have used the SCCS
extensively in ethnographic research; recently it has been
used by economists in studies of cross-cultural differences
in the structure and performance of economies in varying
stages of development (e.g., Baker, 2008; Baker & Miceli,
2005).
The SCCS comprises 186 societies sampled from a wide
range of time periods (including two from before the
Common Era) and geographical locations. The SCCS socie-
ties include contemporary hunter-gatherers, early historic
states, and contemporary industrial societies. This wide
coverage re?ects Murdock and White’s (1969) conscious
decision to mitigate selection biases that favored societies
with English language ethnographic sources. The sam-
pling unit is typically the local community studied in
the primary ethnography used to code a given society
(Murdock & White, 1969, p. 30). Appendix A provides
an in-depth description of how the SCCS database was
constructed.
A major issue with a sample like the SCCS is whether it
is representative of the population of societies present
throughout the world. Some have argued that the SCCS
societies constitute a sample biased in favor of those stud-
ied more intensively by Murdock, his colleagues, and their
students (Otterbein, 1976). Subsequent research demon-
strates that such bias is not present in the SCCS (Gray,
1996). Rather, the predominant bias in the SCCS is towards
societies that have been more intensively studied by eth-
nographers, as previously noted by Murdock and White
(1969, p. 332).
The SCCS presently provides data on more than 2000
categorical variables coded nominally or ordinally. Mur-
dock and Morrow (1970) coded 22 variables related to sub-
sistence economy and related practices, and these
variables are the ?rst listed in SCCS. Additional variables
are added to the free online electronic database as
researchers encode new variables by reading the pre-spec-
i?ed primary and secondary ethnographic studies.
6
Our primary variable of interest is Recordkeeping (SCCS
variable #149, entitled ‘‘Writing and Records”), which is
coded on an ordinal scale from 1 to 5 (Murdock & Provost,
1973, pp. 379–380).
7
Panel A of Fig. 2 describes how this
variable is categorized and shows the number of SCCS soci-
eties classi?ed into each category (see also Panel A of Appen-
dix B). We defer the introduction of several other SCCS
variables until the later sections in which we analyze them.
A value of 1 for Recordkeeping signi?es ‘‘writing, re-
cords, and mnemonic devices in any form are lacking or
unreported.” Seventy-three (39.3%) of the SCCS societies
are coded as completely lacking in records. One such soci-
ety is that of the Mbuti, who are nomadic, gathering Pyg-
mies living in Central/East Africa. Panel B of Fig. 2 shows
that non-recordkeeping societies are less prevalent than
recordkeeping societies (those with a score of 2 or more)
in each sub-sample classi?ed by the society’s pinpointed
year. Panel C of Fig. 2 demonstrates that a greater number
of non-recordkeeping cultures are located in Sub-Saharan
Africa and South America. Conversely, the frequency of cul-
tures possessing records is highest in the Circum-Mediter-
ranean and Eurasia.
8
The next Recordkeeping value of 2 indicates that ‘‘writ-
ing and signi?cant records are lacking but the people em-
ploy mnemonic devices, e.g., simple tallies.” Forty-nine
SCCS societies (26.3%) are coded as using mnemonic de-
vices. Examples include the ancient Mesopotamian tokens,
shown in Panel A of Fig. 1, or the shells used as wampum
by American Indians (Schmandt-Besserat, 1992; Szabo,
2005). A speci?c SCCS example is that of the Kapauku Pap-
uans of New Guinea, who use shell artifacts extensively in
exchange (Pospisil, 1963, pp. 291–293 and 300–311).
A value of 3 for Recordkeeping indicates that a society
‘‘lacks true writing but possesses signi?cant non-written
records in the form of picture writing, quipus, pictorial
inscriptions, or the like.” Twenty-one SCCS societies
(11.3%) are coded as having non-written records. One of
these societies is that of the Incas, who were pinpointed
in 1530 AD shortly after the Spanish invasion of the Amer-
icas. The Incan quipu shown in Panel B of Fig. 1 has long
been recognized as an accounting device to record transac-
tions (Keister, 1964; Urton, 2002).
6
The construction of data for a single variable in SCCS is a painstaking
process. Data coders typically read one or more primary sources and several
secondary sources to determine whether a given society follows a
particular practice. The speci?c coded values are then compiled and
included in SCCS after additional checks are made on their accuracy. SCCS
often includes additional code indicating when categorization of a speci?c
data point is more ambiguous. A quality-ordered bibliography of reference
sources is included in the database (White et al., 2005), and the coding
sheets indicate which particular sources were used for a particular society’s
code, so that subsequent researchers can assess data reliability (White,
1986). Murdock and Morrow (1970, pp. 312–330) describe this process in
detail for the ?rst variables included in SCCS and the individual sources
used in creating these data for each speci?c society.
7
Murdock and Provost (1973) originally coded it 0–4 but the SCCS
presently reports it from 1 to 5.
8
These ?ndings are consistent with Diamond’s (1997) theory that useful
technologies were more likely to spread along the same latitudes in Eurasia
than along the same longitudes in Africa and South America.
S. Basu et al. / Accounting, Organizations and Society 34 (2009) 895–917 899
A value of 4 for Recordkeeping indicates a society that
‘‘has an indigenous system of writing but lacks any signif-
icant accumulation of written records or alternatively has
long used the script of alien people.” Twelve SCCS societies
(6.5%) fall into this category.
9
The ethnographic texts on
which this code is based indicate that written records exist
but provide few speci?cs. For example, Longrigg (1953, pp.
21–25) describes book production and newspapers in Kurd-
istan around 1900, Barth (1960, p. 32) notes that marriage
contracts among the Basseri were drafted by specialists in
marriage rites, and Gamble (1967, pp. 22–26) remarks that
Wolof was the commercial language in the Wolof society
and that school books on this language existed as early as
1823.
The highest category for Recordkeeping (coded as 5) ap-
plies to 31 SCCS societies (16.7%) where the ‘‘society has an
indigenous system of true writing and possesses written
records of at least modest signi?cance.” This group in-
cludes the society (Babylonia) with the earliest pinpointing
Fig. 2. Description of recordkeeping variable from standard cross-cultural sample (SCCS).
9
Although these societies are ranked higher than those in the Record-
keeping category 3, their order could arguably be reversed. To mitigate such
ranking ambiguity, we redid all our analyses excluding all societies in
Recordkeeping category 4 and found very similar results.
900 S. Basu et al. / Accounting, Organizations and Society 34 (2009) 895–917
date (1750 BCE) in the SCCS sample. This date is at the end
of Hammurabi’s reign as Babylonian monarch in a period
when information on transactions and contracts was regu-
larly stored on Cuneiform clay tablets (Van De Mieroop,
2002; Van De Mieroop, 2004). Panel D of Fig. 1 shows an
example of a Cuneiform tablet.
We acknowledge that Recordkeeping is only an indirect
measure of the presence and use of transaction records
within a society. Thus, an SCCS society coded as having a
fully developed writing system could also use that system
for creating other documents – e.g., religious texts. We be-
lieve that the archaeological and historical evidence cited
in this section suggests strongly that Recordkeeping re?ects
the underlying construct of recording transactions. At the
same time, this variable, like many collected for other pur-
poses, is measured with some degree of error. Strictly
speaking, Recordkeeping re?ects the availability of a tech-
nology that could be used for recording economic
transactions.
Hypotheses
The simplest accounting systems provide a historical
record of economic transactions in which a one-way trans-
fer or bilateral exchange has occurred. Transactions gener-
ate a ‘‘paper trail” of receipts, vouchers and contracts that
can be used to verify transaction details in case of forget-
fulness or subsequent disputes.
10
Transaction records are
common to large-scale societies, even those that are pre-lit-
erate. The oldest-known accounting records using ‘‘tokens”
appear at the same time and place (circa 8000 BCE Mesopo-
tamia) as the emergence of agriculture and permanent hu-
man settlements (Schmandt-Besserat, 1992). The
Sumerians invented writing to keep records and accounts
(circa 3200 BCE), which occurred at the same time as sub-
stantial increases in group size and population density in
the earliest cities (Nissen et al., 1993). Thus, accounting
innovations in ancient Mesopotamia coincided with in-
creased societal and economic complexity, suggesting a po-
tential causal connection.
Our focus in this paper is on whether the tight coupling
between accounting advances and economic development
in the ancient Near East was unique, or a pattern that de-
scribes human societies more generally. We are also inter-
ested in the extent to which recordkeeping expands the
scope of impersonal exchange and promotes increasingly
specialized division of labor. We investigate three hypoth-
eses that are depicted in Fig. 3. Each hypothesis is stated in
the lower portion of the ?gure.
Our ?rst hypothesis concerns the emergence of record-
keeping as a society expands in size. We investigate two
implications of the hypothesis. First, recordkeeping is more
likely to emerge once a society has reached the size im-
plied by mental memory constraints. Second, recordkeep-
ing is a necessary institution for extensive impersonal
exchange. As such, recordkeeping will emerge as early as
or earlier than other economic institutions that support ex-
change. The arrow between the box labeled ‘‘Internal
Memory” and the box labeled ‘‘External Records” in Fig. 3
depicts our ?rst hypothesis.
Relying solely on mental memory, humans sustain
greater social exchange than other primates. This is largely
because our evolved brains remember past interactions
and analyze exchange opportunities more effectively than
other species (Cosmides & Tooby, 2005; Wilson, 2000).
That is, human brains are adapted for social exchange
and cooperation that improves our prospects for resource
acquisition and survival. Within a small kin-based group,
mental memory of past interactions and third-party gossip
helps actors identify trustworthy partners for a contem-
plated cooperative venture (Barkow, 1992; Demsetz,
2002). Hence, small groups have little need for external re-
cords because members can accurately track others’ repu-
tations even if they cannot perfectly recall the particulars
of all past interactions (Silk, 2004).
Keeping physical records outside the brain allows peo-
ple to reliably store greater amounts of information on past
interactions and better evaluate the desirability of ex-
change with a speci?c partner (Basu & Waymire, 2006;
Dickhaut et al., 2009a; Dickhaut et al., 2009b). Recordkeep-
ing expands human capacity to ‘‘recognize other individu-
als and keep score” (Ridley 1996, p. 83), which is a
prerequisite for sustaining repeated cooperative social ex-
change and reciprocity (Axelrod & Hamilton, 1981; Nowak
& Sigmund, 2005).
11
This suggests that sole reliance on
mental records will constrain societal expansion beyond a
modest group size.
The evolved human brain can sustain stable cooperative
groups to an upper limit of between 125 and 200 mem-
bers.
12
Dunbar (2001, p. 181) writes:
‘‘(T)here is indeed a characteristic group size of around
125–200 that reappears with surprising frequency in a
wide range of contemporary and Neolithic horticultural
societies. These groups . . . all share one crucial charac-
teristic: they consist of a set of individuals who know
one another intimately and interact on a regular basis. . .
Thus there seems to be quite strong evidence that at
least one component of human grouping patterns is as
much determined by relative neocortex size as are
groups of other primates. We have bigger, more com-
plexly organized groups than other species simply
because we have a larger onboard computer (the neo-
cortex) to allow us to do the calculations necessary to
10
Accounting scholars have long recognized the importance of basic
recordkeeping and its role in providing memory of past exchange trans-
actions (e.g., Demski, 1993; Hat?eld, 1924; Ijiri, 1975; Littleton, 1933,
1953; Potter, 1952).
11
An organism’s ability to recall past interactions with its environment
and adjust behavior in response is of ?rst-order importance to its survival.
This ability is important even for single cell organisms like the Escherichia
coli bacterium (Allman, 2000, pp. 3–8).
12
‘‘Dunbar’s Number” of 125–200 persons was calculated by correlating
troop size and (neocortical) brain size across different primates such as
monkey, baboons and chimpanzees, and extrapolating to expected human
group size using actual human brain size (Dunbar, 1992; Dunbar, 1998).
Dunbar validated his predicted number by studying the historical maxi-
mum sizes of hunter-gatherer tribes, Neolithic villages, Hutterite settle-
ments, Roman army units, and other human groups. Edney (1981, p. 27)
independently estimated that ‘‘the upper limit for a simple, self-contained,
sustaining, well-functioning commons may be as low as 150 people.”
S. Basu et al. / Accounting, Organizations and Society 34 (2009) 895–917 901
keep track of and manipulate the ever-changing world
of social relationships within which we live.”
Dunbar’s Number is the estimated limit to human
group size in the absence of institutions that store data
on past economic and social interactions outside the hu-
man brain. However, transaction records serve as an
(expandable) external hard drive for the mental computer.
This suggests that external recordkeeping should become
increasingly prevalent for groups that exceed 200 persons.
Thus, our ?rst hypothesis is that the extent of recordkeep-
ing and group size will be associated non-linearly, with lit-
tle relation for groups of 200 or fewer persons and a
positive relation for groups exceeding 200 persons.
As a group grows in size, repeated interaction with
familiar partners occurs less often. In addition, individual
cooperation with members of other groups cannot rely
on familiarity or repeated interaction. Recordkeeping helps
people successfully consummate impersonal exchange and
it subsequently enables the emergence of other exchange-
supporting institutions that rely on hard evidence of past
transactions. For example, hard transaction records often
provide a basis for establishing and enforcing property
rights subsequent to property transfers (de Soto, 2000;
VerSteeg, 2000, pp. 46–66). Records also provide the basis
for compiling an individual’s exchange and credit histories,
either in a speci?c market (e.g., eBay) or more generally
with credit ratings like Moody’s. Thus, our ?rst hypothesis
implies also that recordkeeping is a precursor to other ex-
change-supporting institutions. As such, recordkeeping is
expected to appear as early as or earlier than other ex-
change-supporting institutions as an economy expands.
Our second hypothesis is about how essential record-
keeping is for a market to expand and encompass larger
numbers of impersonal exchange transactions. This
hypothesis is depicted by the curved arrow in Fig. 3 con-
necting ‘‘External Records” to ‘‘Impersonal Exchange”
Within human families, many resource transfers are unidi-
rectional grants that can be motivated by love (by parents
for children), fear (by low status members of alpha males)
and ignorance (not recognizing that an object is valuable)
(Boulding, Pfaff, & Horvath, 1972).
13
Most primitive human
societies are extended kin groups with a norm of generalized
reciprocity, where help is expected from and is available to
all group members (Sahlins, 1972).
These societies begin to transcend the bounds of the
immediate group through gift exchanges with neighboring
groups. Such complex interactions inevitably entail rigid
norms of behavior that reduce cheating and any resultant
misinterpretation of intentions (Malinowski, 1922; Mauss,
Fig. 3. Hypotheses. Graphical depiction of hypotheses on the emergence of recordkeeping and its relation to exchange, division of labor, and economic
development.
13
Over lifetimes, such one-way resource transfers likely balance out, but
given high mortality rates in these groups, there is less expectation of stable
partnerships. In more egalitarian societies, transfers between spouses may
have more of an implicit exchange character than in less egalitarian
cultures. Thus, we do not mean to characterize these transfers as
necessarily excluding an exchange component, but rather want to empha-
size that they are not purely exchange transactions between equals.
902 S. Basu et al. / Accounting, Organizations and Society 34 (2009) 895–917
1950). As a by-product of formal gift exchange, trading in
less elaborate economic goods often develops. The scope
of exchange expands to include items that are not ‘‘gifts”
per se but rather are given outside the elaborate rituals of
gift exchange, with informal norms of balanced or sym-
metrical reciprocity in which a fair return is expected from
individual recipients in the future (Sahlins, 1972). Thus, as
the size of an exchange network grows, economic interac-
tion with less-closely-related acquaintances occurs more
often.
14
At some point in the recent human past, a new form of
economic interaction arose in the form of bilateral imper-
sonal exchange or barter between strangers (Seabright,
2004), with an associated norm of negative reciprocity
where each person gives up valuable things and expects
to be reciprocated immediately in a quid pro quo manner
(Sahlins, 1972). Basu’s et al. (2009) experimental results
suggest that recordkeeping may be crucial to this transfor-
mation towards market exchange, in that experimental
economies with recordkeeping exhibited stronger patterns
of negative reciprocity in impersonal exchange than non-
recordkeeping economies. Thus, our second hypothesis is
that recordkeeping will be associated with more extensive
and more complex exchange transactions within a given
society.
Our third hypothesis is that the expanded exchange en-
abled by recordkeeping is associated with increased spe-
cialization in division of labor. Smith (1776/1976, p. 17)
?rst articulated the relation between exchange and divi-
sion of labor in The Wealth of Nations:
This division of labour, from which so many advantages
are derived, is not originally the effect of any human
wisdom, which foresees and intends that general opu-
lence to which it gives occasion. It is the necessary,
though very slow and gradual, consequence of a certain
propensity in human nature which has in view no such
extensive utility; the propensity to truck, barter, and
exchange one thing for another.
Smith (1776, Book I, Chapter III, p. 21) then states his fa-
mous theorem ‘‘As it is the power of exchanging that gives
occasion to the division of labour, so the extent of this divi-
sion must always be limited by the extent of that power,
or, in other words, by the extent of the market” (see also
Stigler, 1951). We depict this hypothesis with the arrow
from ‘‘Impersonal Exchange” to ‘‘Division of Labor” in
Fig. 3. If recordkeeping promotes market expansion (our
second hypothesis), which in turn enables division of labor,
we can also expect a positive association between record-
keeping and division of labor when market extent is ex-
cluded from the regression.
Recordkeeping emerges as group size increases
Our ?rst hypothesis is that recordkeeping becomes
more prevalent after a group reaches the maximum size
achievable under biological constraints and that record-
keeping emerges as early as or earlier than other economic
institutions. A direct test of this hypothesis requires a mea-
sure of group size that re?ects the total number of people
in the group taking account of network ties with non-
group members.
15
The closest proxy within SCCS for this
construct is variable #63, Community Size, which is available
for all but one of the 186 SCCS societies. Community Size is
only a proxy for the overall size of a networked group since
it refers to the size of a typical community (i.e., city or vil-
lage) in the society being studied (Murdock & Wilson,
1972). This is an imperfect proxy to the extent that a com-
munity may have network links to individuals outside the
local community. Community Size takes on eight possible
categorical values. At the lowest are societies where ethnog-
raphers estimate the typical community size to be less than
50 persons and at the highest are communities that each
consist of more than 50,000 people. Panel B of Appendix B
shows the various categories of Community Size.
Panel A of Fig. 4 shows a plot of the frequency of each
Recordkeeping score for a given level of Community Size,
where bubble sizes are proportional to frequency. A line
connects the mean Recordkeeping score for each of the
eight Community Size categories. Consistent with our ?rst
hypothesis, the association between the Recordkeeping
and Community Size is positive and non-linear with a
monotonically increasing mean once Community Size
reaches the level of 200 persons or more. This relation is
reliably positive and signi?cantly different from zero
(Spearman q = 0.32; p < 0.01).
Visual inspection of Panel A in Fig. 4 indicates that most
of the SCCS societies without recordkeeping cluster in the
lower Community Size communities. However, this effect
is not uniform as 14 SCCS societies containing fewer than
200 persons in the typical community also have record-
keeping systems based on a written language (i.e., Record-
keeping = 4 or 5 and Community Size = 1, 2, or 3). These 14
societies include some cultures where the size of the total
population in the society is quite large. For example, these
SCCS societies include villages in both Korea and Japan,
which were subjected to ethnographic study in 1947 and
1950, respectively.
To reduce the impact of measurement error in Commu-
nity Size, we also estimated a more comprehensive mea-
sure of a society’s demographics. We performed a factor
analysis on four SCCS variables: (1) Community Size, (2) Set-
tlement Patterns (SCCS variable #234), (3) Population Size
(SCCS variable #1122), and (4) Population Density (SCCS
variable #1130). Settlement Patterns measures the extent
to which a society is nomadic versus sedentary, Population
Size is a measure of the society’s total size (based on census
data when available), and Population Density measures
the number of persons per square mile in the society.
De?nitions of these variables are provided in Panel B of
Appendix B.
14
In industrialized societies, generalized reciprocity typically character-
izes interactions between parents and children, while balanced reciprocity
characterizes transactions with cousins, neighbors and co-workers.
15
Group size is extensively used to measure the scale of sustained
cooperation within a given species; likewise it is used as a parsimonious
measure of the scale and development of human social complexity (Chick,
1997; Johnson & Earle, 2000; Wilson, 2000, pp. 131–138; Dunbar, 2001).
S. Basu et al. / Accounting, Organizations and Society 34 (2009) 895–917 903
Panel B of Fig. 4 shows results from an iterated principal
factor analysis using communalities among these four vari-
ables. A minimum Eigenvalue of one is used to choose fac-
tors for our subsequent analysis. A single factor with an
Eigenvalue of 2.15 explains 90% of the variation of the four
variables. Accordingly, we retain a single factor to specify a
variable that we refer to as Demographics. Factor loadings
for Demographics with the four variables are shown in Pa-
nel B of Fig. 4. The factor loading is the standardized coef-
?cient in a regression of the variable on the factor and
re?ects the strength of the relationship. Population Density
has the most positive factor loading while Community Size
has the least positive factor loading. The weaker associa-
tion between Demographics and Community Size suggests
that important features of a society’s demographic com-
plexity are not fully captured by Community Size.
Panel C of Fig. 4 shows a plot of the frequency of each
Recordkeeping score for each of eight levels of Demograph-
ics, where again bubble sizes are proportional to frequency.
To enable direct comparison with Panel A, we rank the
societies on the basis of Demographics, and then partition
them into eight sub-samples using the same number of
societies in each of the eight ordered categories of Commu-
nity Size used to produce Panel A. Inspection of Panel C sug-
gests that using Demographics in lieu of Community Size
modestly increases the strength of association between
Spearman ? = 0.32 (n=185, p < .01)
Factor Eigenvalue
Proportion
Explained
Cumulative
Explained Factor Loadings
1 2.15 0.90 0.90 Population Density (SCCS #1130) 0.88
2 0.19 0.08 0.98 Population Size (SCCS #1122) 0.69
3 0.04 0.02 1.00 Settlement Patterns (SCCS #234) 0.68
4 -0.00 -0.00 1.00 Community Size (SCCS #63) 0.62
True writing
with records
Community Size
None
Mnemonic
devices
Non-written
records
True writing;
no records
< 50 50 to
99
100 to
199
200 to
399
400 to
999
1,000 to
4,999
5,000 to
49,999
> 50,000
Spearman ? = 0.37 (n=185, p < .01)
True writing
with records
None
Mnemonic
devices
Non-written
records
True writing;
no records
Demographics factor groups
1 2 3 6 7 8 4 5
(A)
(B)
(C)
Fig. 4. Relation between Recordkeeping and group size
904 S. Basu et al. / Accounting, Organizations and Society 34 (2009) 895–917
Recordkeeping and the measure of group size (Spearman
q = 0.37, p < 0.01). In addition, the plotted relation between
mean Recordkeeping score and Demographics shows a gen-
erally increasing trend.
Panel D of Fig. 4 shows estimates from a regression
model where Recordkeeping is the dependent variable and
either Community Size or Demographics is the independent
variable. The model allows a kinked relation between
Recordkeeping and the independent variable, with the kink
located at group size or complexity value equal to ‘‘4.”
Intuitively, this means that we allow the relation between
Recordkeeping and Community Size to differ once groups
have reached a level of Community Size greater than or
equal to 200 persons (or a comparable level of complexity
based on Demographics). The results indicate a non-linear
positive relation between the extent of recordkeeping
and group size. In both models, the coef?cient on the inter-
action between group size and a 0–1 indicator for whether
the society is ‘‘large/complex” is positive and signi?cant
(p < 0.02 for both models). This evidence suggests that, as
a society surpasses the modest group size threshold of
200 suggested by Dunbar (1992), recordkeeping becomes
more complex.
A second implication of our ?rst hypothesis is that
recordkeeping emerges as early as or earlier than other ex-
change-supporting institutions. We identi?ed other ex-
change-supporting institutions using SCCS variables
re?ecting the use of money and credit in an economy,
the presence of a judiciary and property rights, and the
presence of administrative hierarchies. Money and prop-
erty rights are likely fundamental to expansion of ex-
change (Demsetz, 2002; Menger, 1892) and the demand
for accounting arises in part from the existence of complex
organizations and credit markets (Kimbrough, Smith, &
Wilson, 2008; Watts & Zimmerman, 1986). We use the fol-
lowing ?ve variables: Credit Source (SCCS variable #18),
Judiciary (SCCS variable #89), Administrative Hierarchy
(SCCS variable #91), Money (SCCS variable #155), and
Inheritance of Land (SCCS variable #278). These variables
are de?ned in Panel B of Appendix B.
Fig. 5 plots the cumulative percentage of societies in
which Recordkeeping or each of the other ?ve institutions
is present, where societies are ordered by Community Size
level. A speci?c institution is deemed to be present if the
society’s code for a given variable exceeds the minimum
possible value. Each point on a given line represents the to-
tal number of societies up to that size level where the insti-
tution is present divided by the total number of societies
with a code available for that variable.
The six SCCS variables’ cumulative frequency functions
cluster into three groups referenced by the capital letters
on the right-hand side of the ?gure. The ?rst cluster (la-
beled A) includes Recordkeeping along with Money and
Inheritance of Land. These institutions are present in
approximately 60% of the SCCS societies. The cumulative
frequencies for Inheritance of Land and Money are not sta-
tistically different from the cumulative frequency of
Recordkeeping (p > 0.10 based on a z-test of proportions).
This suggests that recordkeeping is a fundamental institu-
tion that, like the use of money in exchange and simple
property rights systems, emerges early as an economy
develops. This relation is consistent with the strong corre-
lation between property rights and recordkeeping in the
SCCS documented by Baker and Miceli (2005), which they
conjectured was due to the bene?cial role of recordkeeping
in allowing property transfers. That recordkeeping and
money emerge in similar fashion is likely not surprising.
Economists have noted that shells and similar artifacts
can serve as money to promote exchange and that these
monetary artifacts provide memory of past exchanges
(Kocherlakota, 1998; Townsend, 1989).
16
Recordkeeping
i
= ? + ?
1
Size
i
+ ?
2
Size
i
*Large_Comm
i
+ ?
3
Large_Comm
i
+ ?
i
Community Size Demographics
Variable Predicted sign Coefficient p-value Coefficient p-value
4 5 . 0 + e z i S 0.01 0.21 0.17
Size * Large_Comm + 0.52 0.02 1.15 0.00
Large_Comm – – 3.35 0.00 – 4.88 0.00
Pseudo R
2
0.07 0.10
N 185 185
Recordkeeping is SCCS variable #149 (Writing and Records) with categories defined on the y-axis in Panel A. In the model using Community
Size as an independent variable, Size is SCCS variable #63 (Community Size) with categories defined on the x-axis in Panel A. Large_Comm is
an indicator variable equal to 1 if Community Size is greater than 200 persons and equal to 0 if Community Size is less than 200 persons.
Demographics refers to the factor estimated in the lower portion of Panel B. For purposes of this test, the societies were ranked on
Demographics and placed into eight groups where the total number of societies assigned to each group was identical to the number of societies
in each of the eight groups corresponding to different values for Community Size. When Demographics is used as an independent variable, Size
equals a number from one to eight corresponding to which of the ranked groups that the society has been assigned (lowest = 1 and highest = 8).
Large_Comm in this model equals 0 if the society is in groups 1 – 3 and equals 1 if the society is in groups 4 – 8. The model is estimated using
Ordered Logit. p-Values are one-tailed when the si gn is directionally predicted and are based on heteroskedasticity-consistent standard errors
adjusted for residual correlation among observations belonging to the same major language family.
(D)
Fig. 4 (continued)
16
Also, the tally sticks used by the British Exchequer beginning several
centuries ago evolved into bills of exchange, which served a monetary
function in exchange (Robert, 1952, p. 80; Goetzmann & Williams, 2005).
S. Basu et al. / Accounting, Organizations and Society 34 (2009) 895–917 905
The other two clusters include Administrative Hierarchy
and Judiciary (cluster B) and Use of Credit (cluster C), which
are present in about 45% and 33% of the SCCS societies,
respectively. In each case, the frequency of these institu-
tions is statistically different from Recordkeeping at con-
ventional levels (p 6 0.01). These data indicate that more
advanced institutions like hierarchies, courts, and the
availability of credit beyond the family are less likely to ap-
pear in the earliest stages of an economy’s development
relative to recordkeeping, money and basic property rights.
The lower part of Fig. 5 shows the percentage of socie-
ties where a given institution is present for both societies
with and without recordkeeping. Each institution is pres-
ent more frequently in societies with recordkeeping than
societies without recordkeeping. Chi-square tests reject
the null of independence at p < 0.05 for each of the ?ve
comparisons. This pattern is consistent with institutional
co-evolution where multiple interdependent institutions
emerge as a society grows in size and complexity.
Overall, the evidence in Figs. 4 and 5 is consistent with
the hypothesis that recordkeeping is a foundational insti-
tution that emerges early as an economy expands. Further-
more, recordkeeping societies are more likely to develop
other exchange-supporting institutions.
The in?uence of recordkeeping on impersonal exchange
and division of labor
Recordkeeping promotes impersonal exchange
We require a proxy for the extent of impersonal ex-
change to test our second hypothesis, which is that record-
keeping promotes impersonal exchange. We develop a
multi-attribute measure based on the extent to which
impersonal exchange takes place in the society as well as
whether the society has other institutions that support
more extensive exchange. We use a broader measure of ex-
change because Smith (1776/1976, pp. 21–33) argues that
the extent of the market depends on an effective infra-
structure for storing and moving goods as well as a med-
ium of exchange.
Our measure of impersonal exchange is based on a fac-
tor analysis applied to Money and three other SCCS vari-
ables de?ned in Panel C of Appendix B. Intercommunity
Trade as a Food Source (SCCS variable #1) is a direct mea-
sure of impersonal exchange as re?ected in a society’s food
import levels. Similarly, Food Storage Via Surplus (SCCS var-
iable #21) measures the extent to which exchange is likely
to occur as a result of food production in excess of imme-
diate consumption needs. Land Transport (SCCS variable
#154) re?ects the extent to which a society has a transpor-
tation infrastructure necessary to support exchange with
more geographically distant areas. Money (SCCS variable
#155) was used in our previous analyses and measures
whether a medium of exchange that could support more
extensive impersonal exchange is available.
We perform a principal factor analysis using the com-
munalities among these four variables to extract underly-
ing dimensions, and we used a minimum Eigenvalue of
one to determine which factors to retain. Panel A of Table
1 shows that one factor with an Eigenvalue of 1.27 ac-
counts for 91% of the total variance of the four SCCS vari-
ables. Thus, we retain only one factor to specify Exchange.
The factor loadings in Panel B of Table 1 show that Money
is strongly associated with Exchange. Intercommunity Trade
Fig. 5. Cumulative % of SCCS societies where an institution is present at a given level of community size.
906 S. Basu et al. / Accounting, Organizations and Society 34 (2009) 895–917
as a Food Source, the most direct measure of exchange, and
Land Transport also exert a strong effect. Food Surplus Via
Storage shows the weakest association.
We estimate the following empirical model of the rela-
tion between Recordkeeping and Exchange:
Exchange
i
= a
0
+ b
1
RK_B
i
+ b
2
Large_Comm
i
+ b
3
RK_B
i
*
Large_
Comm
i
+ b
4
AgriculturalPotential
i
+ b
5
Climate
i
+ b
6
Region
i
+e
i
.
Exchange is the factor derived from the factor analysis in
Panels A and B of Table 1. RK_B is a transformed binary ver-
sion of SCCS #149 (Writing and Records) where 0 indicates
recordkeeping of any kind is absent and 1 indicates record-
keeping of any kind is present. We use a 0–1 indicator for
Recordkeeping since we expect a bi-directional causal rela-
tion between the quantity and complexity of recordkeep-
ing and exchange. Because Recordkeeping appears early in
development (see Fig. 5), we expect that these effects
(and any resultant endogeneity problems) will be lessened
with use of a 0–1 variable. Large_Comm is an indicator var-
iable that equals 0 when Community Size (SCCS #63) is less
than 200 persons and equals 1 when Community Size is
greater than or equal to 200 persons. Under our second
hypothesis, we expect that b
1
and b
3
will be positive.
The model includes three control variables (de?ned in
Panel C of Appendix B): (1) Agricultural Potential (SCCS
#921) of the society’s region, (2) Climate (SCCS #857) is
an ordinal variable re?ecting open access to rich ecological
resources, and (3) Region (SCCS #200) represents a series of
0–1 variables for a society’s location within one of six ma-
jor world regions. These variables are included to capture
cross-society differences in resource endowments, which
Diamond (1997) argues are prerequisites for economic
development.
The OLS models are estimated using the 182 SCCS soci-
eties where data are available for all variables. p-Values are
one-tailed when a signed prediction is present and are
based on heteroskedasticity-consistent standard errors ad-
justed for residual correlation among observations belong-
ing to the same major language family.
17
Estimation results are shown in Panel C of Table 1. The
?rst column shows results for the model when only RK_B
and the control variables are included. The estimated coef-
?cient, b
1
, equals 0.51, which is signi?cantly different from
zero at p < 0.00 (one-tailed). Recordkeeping increases the
adjusted R
2
of the OLS model including only control vari-
ables from 0.25 (untabulated) to 0.32. b
1
remains positive
(0.26) and signi?cant at p < 0.02 for the full model (shown
in the second column) and b
3
equals 0.57, which is signif-
icant at p < 0.02 (one-tailed). This latter effect suggests that
the presence of recordkeeping exerts a substantially larger
effect on the extent of impersonal exchange in larger soci-
eties. Results in the right-most column demonstrate that a
signi?cant relation between recordkeeping and exchange
persists when Demographics is included as an additional
control for social complexity. This evidence supports our
second hypothesis that recordkeeping supports an expan-
sion in impersonal exchange.
The in?uence of impersonal exchange on division of labor
Thus far we have seen that recordkeeping becomes
more prevalent once a society has reached modest size lev-
els, recordkeeping emerges early relative to other basic
Table 1
Tests of association between recordkeeping and impersonal exchange.
A: Exchange factor estimation (Iterated principal factors)
Factor Eigenvalue Proportion explained Cumulative explained
1 1.27 0.91 0.91
2 0.08 0.06 0.97
3 0.04 0.03 1.00
4 –0.00 –0.00 1.00
B: Factor loadings for exchange (One factor retained)
Money 0.84
Intercommunity trade
as a food source
0.51
Land transport 0.50
Food surplus via
storage
0.24
C: Association between recordkeeping and exchange (p-values)
Variable Predicted
sign
Model with
Demographics as a
control
RK_B + 0.51
(0.00)
0.26
(0.02)
0.13
(0.07)
Large_Comm –0.01
(0.93)
–0.41
(0.00)
RK_B
*
Large_Comm + 0.57
(0.01)
0.48
(0.01)
Demographics 0.56
(0.00)
Adj. R
2
0.32 0.38 0.63
N 182 182 182
Model: Exchange
i
= a
0
+ b
1
RK_B
i
+ b
2
Large_Comm
i
+ b
3
RK_B
*
Large_
Comm
i
+ Controls + e
i
.
The sample used in Panel C includes the 182 SCCS societies where data for
all variables were available. The models are estimated using OLS.
Exchange is the factor derived from the factor analysis in Panels A and B.
RK_B is a transformed binary version of SCCS #149 (Records and Writing)
where 0 indicates recordkeeping of any kind is absent and 1 indicates
recordkeeping of any kind is present. Large_Comm is an indicator variable
that equals 0 when Community Size (SCCS #63) is less than 200 persons
and equals 1 when Community Size is greater than 200 persons. Demo-
graphics refers to the factor estimated in Fig. 4, Panel C. Controls included
in all models are: (1) Agricultural Potential (SCCS #921) of the society’s
region, (2) Climate (SCCS #857), a 6-scale categorical variable ordered in
terms of open access to rich ecological resources, and (3) Region (SCCS
#200) dummies that represent 0–1 variables based on the society’s
location within one of six major world regions. p-Values are one-tailed
when a signed prediction is present and are based on heteroskedasticity-
consistent standard errors adjusted for residual correlation among
observations belonging to the same major language family.
17
An analysis of the SCCS data indicates that Recordkeeping is subject to
stronger patterns of cultural and historical diffusion than other SCCS
variables. Each SCCS culture is assigned a number from 1 to 186 where the
societies are ordered according to geographical proximity and cultural
similarity. Because societies with close geographical proximity are likely
ones where cultural diffusion may still be a prominent force, the correlation
between adjacent numbered societies within the SCCS measures the extent
to which the pinpointing process did not completely eliminate cross-
sectional dependence (Murdock & White, 1969). Recordkeeping displays
strong correlation (q = 0.30) when comparing adjacent neighbors within
the SCCS database. Thus, we estimate all models using heteroskedasticity-
consistent standard errors adjusted for residual correlation among obser-
vations belonging to the same major language family.
S. Basu et al. / Accounting, Organizations and Society 34 (2009) 895–917 907
institutions as societies grow larger, and that the emer-
gence of recordkeeping is associated with increased imper-
sonal exchange. Our ?nal hypothesis is, consistent with
Smith (1776/1976, p. 21), that expanded opportunities
for impersonal exchange are associated with increasingly
specialized division of labor within an economy.
Testing this hypothesis requires ?rst that we specify a
measure of a society’s division of labor. As with our tests
examining market exchange, we specify a measure of divi-
sion of labor using factor analysis applied to multiple SCCS
variables likely related to the underlying construct. Most
SCCS societies are heavily dependent on agriculture and
depend far less on industrial manufacturing and complex
services than modern Western economies. We thus con-
struct Division of Labor using measures of the extent to
which complex agriculture and other forms of occupa-
tional specialization are present in the society.
Five SCCS variables serve as inputs to a factor analysis to
specify Division of Labor, as de?ned in Panel C of Appendix B.
The two agricultural measures include: (1) Agriculture (SCCS
variable #151), which takes on ?ve possible values ranging
from 1 indicating no agriculture is present to 5 indicating
that intensive agriculture provides the primary foodsources
for the society, and (2) Intensity of Cultivation (SCCS variable
#232), which takes on six possible values ranging from 1
(‘‘no agriculture”) to 6 (‘‘intensive irrigated agriculture”).
Technological Specialization (SCCS variable #153) is a vari-
able re?ecting whether specialist potters, loom weavers,
or metalworkers are present in the society. Administrative
Hierarchy (SCCS variable #91) measures the extent to which
decision-making in the society is delegated to heads of sub-
groups and Class Strati?cation (SCCS variable #270) mea-
sures the extent to which social status arises fromresources
or power possessed by an individual or group.
Panel A of Table 2 shows that one factor with an Eigen-
value of 2.86 accounts for 81% of the total variance of the
?ve SCCS variables used; we therefore retain only one fac-
tor to specify Division of Labor. Panel B of Table 2 indicates
that Intensity of Cultivation and Agriculture exert more
in?uence on Division of Labor than do Technological Special-
ization, Administrative Hierarchy, and Class Strati?cation.
We estimate a model of the relation between Division of
Labor and Exchange as follows:
Division of Labor
i
= a
0
+ b
1
Exchange
i
+ Controls
i
+ e
i
Division of Labor is the factor obtained from the analysis
described in Panels A and B of Table 2 and Exchange is the
factor derived from the analysis in Panels A and B of Table
1. The model includes the same control variables used in
the regressions in Table 1: Agricultural Potential (SCCS
#921), Climate (SCCS #857), and Region (SCCS #200); de?-
nitions are in Appendix B. We hypothesize that b
1
will be
positive. The sample includes the 180 SCCS societies for
which data for all variables are available. Models are esti-
mated using OLS and p-values are one-tailed and are based
on heteroskedasticity-consistent standard errors adjusted
for residual correlation among observations belonging to
the same major language family.
The results in Panel C of Table 2 support the hypothesis
that Exchange exhibits a positive association with Division
of Labor. The coef?cient on Exchange (b
1
) equals 0.44
(p < 0.00, one-tailed), and remains signi?cantly positive
(b
1
= 0.29, p < 0.00) when Community Size is used as an
additional control. The relation between Division of Labor
and Exchange becomes weakly signi?cant (p = 0.06) when
Demographics is added as a control. This is the artifact of Ex-
change being strongly correlated with variables included in
Demographics other than Community Size. In particular,
variables capturing the dispersion and density of a popula-
tion such as Settlement Patterns (SCCS variable #234) and
Population Density (SCCS variable #1130) re?ect the loca-
tion of individuals within a society who can take advantage
of exchange opportunities. The ?nal two columns indicate
that the coef?cient on RK_B is positive and statistically sig-
ni?cant (coef?cient = 0.30, p = 0.04) when included with
Exchange as an independent variable. In contrast, RK_B is
Table 2
Tests of association tests between division of labor and impersonal
exchange.
A: Division of labor factor estimation (Iterated principal factors)
Factor Eigenvalue Proportion explained Cumulative explained
1 2.86 0.81 0.81
2 0.62 0.17 0.98
3 0.06 0.02 1.00
4 0.01 0.00 1.00
5 –0.00 –0.00 1.00
B: Factor loadings for division of labor (One factor retained)
Intensity of cultivation 0.87
Agriculture 0.82
Technological
specialization
0.71
Administrative
hierarchy
0.62
Class strati?cation 0.62
C: Association of division of labor with exchange (p-values)
Variable Predicted
sign
With
Community
Size as a
control
With
Demographics
as a
control
With RK_B
Exchange + 0.44
(0.00)
0.29
(0.00)
0.10
(0.06)
0.39
(0.00)
RK_B + 0.30
(0.04)
0.50
(0.00)
Adj. R
2
0.50 0.60 0.69 0.52 0.43
N 180 180 180 180 180
Model: Division of Labor
i
= a
0
+ b
1
Exchange
i
+ Controls + e
i
.
The sample used in Panel C includes only the 174 SCCS societies where
data for all variables were available. The models are estimated using OLS.
Division of Labor is the factor derived from the factor analysis in Panels A
and B. Exchange is the factor for the extent of market exchange derived
from the factor analysis in Panels A and B of Table 1. Controls include: (a)
Agricultural Potential (SCCS #921) of the society’s region, (b) Climate (SCCS
#857) is a 6-scale categorical variable ordered in terms of open access to
rich ecological resources, and (c) Region (SCCS #200) dummies represent
0–1 variables based on the society’s location within one of six major
world regions. Community Size is used as an additional control variable in
the second column and Demographics is used as the additional control
variable in the third column in lieu of Community Size. Community Size is
SCCS #63 and Demographics is the factor estimated in Panel C of Fig. 4.
RK_B is a transformed binary version of SCCS #149 (Writing and Records)
where 0 indicates recordkeeping of any kind is absent and 1 indicates
recordkeeping of any kind is present. RK_B is added as an additional
regressor in the fourth column. The ?nal column shows the results using
RK_B as an independent variable on a stand-alone basis. p-Values are one-
tailed and are based on heteroskedasticity-consistent standard errors
adjusted for residual correlation among observations belonging to the
same major language family.
908 S. Basu et al. / Accounting, Organizations and Society 34 (2009) 895–917
positive and highly signi?cant when included on a stand-
alone basis (coef?cient = 0.50, p < 0.00), and adds explana-
tory power to a baseline regression including only the con-
trol variables – adjusted R
2
increases from 0.37
(untabulated) to 0.43. The 40% decline from 0.50 to 0.30
for the coef?cient on RK_B suggests that the effect of
Recordkeeping on Division of Labor ?ows partially through
Exchange.
The causal force hypothesized to act upon division of la-
bor when impersonal exchange is possible is that individu-
als invest in human capital that affords thema comparative
advantage in producing specialized goods and services.
More generally, this is part of a broader investment pattern
within the society where the emergence of impersonal ex-
change leads to greater investment in all forms of capital –
e.g., tangible capital and improvements to land (Smith
1776/1976, pp. 351–371). Accordingly, we also examined
the relation between impersonal exchange and the level
of capital stock accumulated by a society.
We estimate a measure of a society’s accumulated capi-
tal stock using six SCCS variables that re?ect physical, hu-
man, and social capital. Three measures pertain to
accumulated physical capital: Large or Impressive Structures
(SCCS #66), Resource Base (SCCS #859), and Cropping Index
(SCCS #1128). Large or Impressive Structures re?ects the ex-
tent to which physical structures have been erected in the
society and ranges from 1 (‘‘none”) to 6 (‘‘economic or
industrial buildings”). Resource Base measures whether
complementary techniques or assets have been developed
that improve agricultural productivity. This variable takes
on three possible values with a code of 1 representing
‘‘low resources (ex, hunting, gathering, ?shing)” and a code
of 3representing ‘‘highresources (ex, advancedhorticulture
with metal hoes, intensive agriculture with plow, pastoral-
ism).” Cropping Index is a measure of land utilization in agri-
culture ranging from 1 (‘‘no agriculture or con?ned to non-
food crops”) to 6 (‘‘100% or more of land used per year”).
Education is a measure of accumulated human capital
and represents the sum of several variables (SCCS #’s
425, 426, 427, and 428) measuring the extent to which
children of the society receive training and education. Each
of these four variables is coded on a six-point scale from 1
(‘‘informal training, with minimal guidance”) to 6 (‘‘formal
schooling typical and frequent”). Political Autonomy (SCCS
#81) and Military Success (SCCS #908) are included as mea-
sures of the extent to which the society has developed a
government that protects against external threats. The
ability to repel external threats can enhance productivity
by reducing the extent to which another group expropri-
ates the fruits of labor by a society. Political Autonomy is
coded on a six-point scale that ranges from 1 (‘‘dependent
totally”) to 6 (‘‘fully autonomous”). Military Success is
coded on a scale that ranges from 1 (‘‘no – its bound-
aries/population are shrinking”) to 4 (‘‘yes – its bound-
aries/population are expanding”).
These six variables provide the inputs to a factor analy-
sis used to identify a variable we label Capital Stock. As be-
fore, we conduct a principal factor analysis using the
communalities among these variables, along with a mini-
mum Eigenvalue of one to determine which factors to re-
tain. Panel A of Table 3 shows that one factor with an
Eigenvalue of 1.85 accounts for 71% of the total variance
of the six SCCS variables used to identify Capital Stock. Pa-
nel B of Table 3 shows that Cropping Index is the most in?u-
ential variable and Education, Resource Base, and Large
Impressive Structures are also of importance. The political
variables, Political Autonomy and Military Success, are the
least important variables in specifying Capital Stock.
We estimate a model of the relation between Capital
Stock and Exchange as follows:
Capital Stock
i
= a
0
+ b
1
Exchange
i
+ Controls
i
+ e
i
Capital Stock is the variable identi?ed in the factor anal-
ysis in Panels A and B of Table 3 and all other variables are
as previously de?ned for purposes of our prior analyses.
Each model includes the control variables used previously
(Agricultural Potential, Climate, and Region). De?nitions of
all variables are provided in Panel E of Appendix B. The
Table 3
Tests of association between Capital Stock and Exchange.
A: Capital stock factor estimation (Iterated principal factors)
Factor Eigenvalue Proportion explained Cumulative explained
1 1.85 0.71 0.71
2 0.47 0.18 0.89
3 0.18 0.07 0.96
4 0.08 0.03 0.99
5 0.02 0.01 1.00
6 –0.00 –0.00 1.00
B: Factor loadings for capital stock (One factor retained)
Cropping index 0.75
Investment in
education
0.58
Resource base 0.55
Large impressive
structures
0.53
Political autonomy 0.38
Military success 0.29
C: Association of capital stock with exchange (p-values)
Variable With
RK_B
With
Community
Size as a
control
With
Demographics
as a control
Exchange 0.56
(0.00)
0.21
(0.00)
0.20
(0.00)
0.19
(0.00)
0.17
(0.01)
Division of
labor
0.65
(0.00)
0.64
(0.00)
0.60
(0.00)
0.58
(0.00)
RK_B 0.07
(0.24)
Adj. R
2
0.53 0.78 0.78 0.79 0.79
N 143 143 143 143 143
Model: Capital Stock
i
= a
0
+ b
1
Exchange
i
+ Controls + e
i
.
The sample used in Panel C includes the 143 SCCS societies where data for
all variables were available. The models are estimated using OLS. Capital
Stock is the factor derived from the factor analysis in Panels A and B,
Exchange is the factor for the extent of impersonal exchange derived from
the factor analysis in Panels A and B of Table 1, and Division of Labor is the
factor derived from the factor analysis in Panels A and B of Table 2. RK_B is
a transformed binary version of SCCS #149 (Records and Writing) where 0
indicates recordkeeping of any kind is absent and 1 indicates record-
keeping of any kind is present. Controls included in all models are: (1)
Agricultural Potential (SCCS #921) of the society’s region, (2) Climate (SCCS
#857) is a 6-scale categorical variable ordered in terms of open access to
rich ecological resources, and (3) Region dummies represent 0–1 variables
based on the society’s location within one of six major world regions.
Community Size equals SCCS variable #63 and Demographics is the factor
estimated in Panel C of Fig. 4. p-Values are one-tailed and based on het-
eroskedasticity-consistent standard errors adjusted for residual correla-
tion among observations belonging to the same major language family.
S. Basu et al. / Accounting, Organizations and Society 34 (2009) 895–917 909
sample used to estimate Models (3), (4), and (5) includes
the 143 SCCS societies where data for all variables are
available.
The hypothesis of interest is that Capital Stock is associ-
ated with the emergence of impersonal exchange after
controlling for other factors. Thus, we expect that b
1
will
be positive. As with the previous models, OLS estimation
is used, p-values are one-tailed, and heteroskedasticity-
consistent standard errors adjusted for residual correlation
among observations belonging to the same major language
family are used.
The evidence in Panel C supports the hypothesis that
cross-society differences in Capital Stock are associated with
differences in Exchange. The ?rst column of results indicates
that the coef?cient on Exchange equals 0.56 (p < 0.00). Fur-
ther, this result is highly robust. Exchange remains signi?-
cant after controlling for factors already captured by
Division of Labor (see second column). The third column
indicates that RK_B shows no signi?cant association with
Capital Stock after controlling for Exchange and Division of
Labor. This suggests that recordkeeping’s role is one of an
enabling technology that supports expanded exchange
and division of labor.
18
The ?nal two columns demonstrate
that the relation between Exchange and Capital Stock is not
capturing purely an effect due to size and demographic com-
plexity. The economic institutions of Exchange and Division of
Labor add considerable explanatory power to a baseline mod-
el that includes only the geographical control variables which
generate an (untabulated) adjusted R
2
of 0.30.
Viewed collectively, the evidence presented in Tables 2
and 3 provides strong support for our ?nal hypothesis that
the expansionof impersonal exchange facilitated by record-
keeping is also associated with increasingly specialized
division of labor and greater overall investment in physical,
tangible, and political capital. These results accord with
prior experimental evidence that recordkeeping is associ-
ated with greater trust (Basu et al., 2009), and that trust is
associated with investment and economic growth, both
analytically andincross-country growthregressions (Knack
& Keefer, 1997; Zak & Knack, 2001). In other words, record-
keeping enables strangers to trust each other in complex
intertemporal exchange, which then facilitates division of
labor and greater productivity, which in turn enable greater
investment and even faster economic growth.
Conclusions and implications
Our evidence suggests that recordkeeping is more likely
in large groups that cannot sustain cooperative interaction
based solely on mental memory, and that recordkeeping,
like money and inheritance of land, emerges at relatively
early stages of an economy’s development. The emergence
of recordkeeping precedes the appearance of a judiciary,
administrative hierarchies and the extension of credit, sug-
gesting that recordkeeping is a foundational institution.
Our evidence also suggests that economies where record-
keeping is possible are characterized by more extensive
impersonal exchange. Consistent with hypotheses by Smith
(1776/1976) and Stigler (1951), we also ?nd that the level
of specialization in division of labor and accumulated cap-
ital are strongly in?uenced by the extent of the market.
These ?ndings suggest that the basic accounting function
of recordkeeping is a precursor to economic development
through impersonal exchange and division of labor.
More broadly, our evidence is consistent with the
hypothesis that transaction records external to the human
brain are necessary to extend the scale of human coopera-
tion from small primitive groups to large-scale modern
societies characterized by extensive market exchange and
complex division of labor (Basu & Waymire, 2006). Our
analysis also broadly accords with conjectures offered by
an earlier generation of scholars (i.e., Sombart, Weber,
Schumpeter, and von Mises) that capitalist economic
development would be impossible without accounting
institutions like double-entry bookkeeping (Carruthers &
Espeland, 1991; Most, 1972). Thus, considerably more re-
search on how recordkeeping and more advanced analysis
of accounting’s transactional data promote economic
development is warranted.
Economic development varies considerably across conti-
nents andcountries (e.g. World Bank, 2006), as well as with-
in countries among different ethnic and sociolinguistic
groups. Our evidence is consistent with De Soto’s (2000)
argument that accurate property records are a prerequisite
for the success of capitalistic societies. The crucial role of
veri?able transaction records for legal enforcement of prop-
erty rights is explicitly indicated in legal codes ranging from
the ancient Code of Hammurabi (circa 1750 BCE) through
Athenian, Roman and French legal codes to the recent Sar-
banes-Oxley Act of 2002 (Basu&Waymire, 2006). Thus, ver-
i?able transaction records are a necessary part of the
foundations that lie beneaththe exchange-supporting insti-
tutions upon which capitalist economies have been built.
Accounting is likely an ecologically rational institution
that coordinates economic interaction through market ex-
change (Waymire, 2009; Waymire & Basu, 2007). To adapt
a metaphor from Simon (1990), historical cost accounting
records and today’s exchange agreements are like two
blades of scissors that have become increasingly effective
together over time through co-evolution. Institutional
changes such as ‘fair value’ accounting that overwrites his-
torical records of consummated transactions and erodes
the quality of memory inherent in records may make the
scissors less effective unless a matching blade is pro-
duced.
19
Contracts and regulations rely on a set of common
expectations about how performance is measured, and
wholesale changes to this performance measure may entail
‘‘unintended consequences.” One ultimate consequence of
less effective recordkeeping may be the decay of economic
institutions responsible for wealth generation in modern
developed societies.
18
In contrast, the coef?cient on RK_B equals 0.51 (p = 0.00) with Exchange
and Division of Labor excluded from the model.
19
Similarly, Ball (2001) argues that accounting harmonization is unlikely
to produce higher quality ?nancial reports unless contracts and enforce-
ment mechanisms are also changed to provide incentives for better
reporting. While Ball focuses on whether developing countries can improve
their accounting quality, we worry that developed countries are in danger
of reducing their economic effectiveness.
910 S. Basu et al. / Accounting, Organizations and Society 34 (2009) 895–917
Acknowledgements
We acknowledge helpful comments received fromMark
Kohlbeck, Eric Press, Richard Sansing, Denise Schmandt-
Besserat, two anonymous reviewers, participants in the
Emory University Anthropology Department Workshop
Series on Human Behavior and Evolution, the 2007 Univer-
sity of Oklahoma Accounting Research Conference, the
Fifth Accounting History International Conference, the
2008 FARS Midyear Meeting, the 2008 Annual Meeting of
the International Society for New Institutional Economics,
the 2008 Annual Meeting of the American Accounting
Association, and accounting seminars at Arizona State, Chi-
cago, City University of Hong Kong, Florida, Minnesota,
Northwestern, Southern California, Temple, and Wisconsin.
The Goizueta Business School at Emory University pro-
vided ?nancial support for this research.
Appendix A
Construction. of the Standard Cross-Cultural Sample (SCCS)
The Standard Cross-Cultural Sample (SCCS) provides a
cross-section of ethnographic ‘‘snapshots” that we use to
investigate cross-cultural variation in recordkeeping prac-
tices throughout the world. These pictures capture multi-
ple elements of a culture or society in a location at a
speci?c point in time. Because these data have not been
previously used in the accounting literature, we describe
the construction of this sample in some depth.
Murdock and White (1969) constructed the SCCS to
standardize the data used in cross-cultural research and
facilitate statistical analysis.
20
Paying careful attention to
ethnographic distributions, Murdock and White identi?ed
186 cultural provinces in dispersed geographical locations
and time periods (including two from before the Common
Era). Murdock and White then chose one society to repre-
sent each cultural province, picking the earliest well-de-
scribed societies to the extent practicable. The SCCS
societies include contemporary hunter-gatherers, early his-
toric states, and contemporary industrial societies. This wide
coverage re?ects Murdock and White’s (1969) conscious
decision to mitigate biases that favored societies with Eng-
lish language ethnographic sources.
A major purpose in constructing the SCCS was to in-
crease the extent to which statistically valid inferences
could be drawn from ethnographic data. Speci?cally, prior
studies had often used data that lacked independence.
Cross-cultural correlation in cultural practices arises from
the diffusion of those practices among cultures with a
common heritage. Anthropologists have recognized this
problem (a.k.a. Galton’s Problem) for over a century. The
SCCS was speci?cally designed to minimize cross-society
dependence because statistical corrections such as Fama-
MacBeth or White-Huber standard errors had yet to be de-
vised. Murdock and White (1969) addressed this problem
by ‘‘pinpointing” their societies to speci?c locales and
dates. The pinpointing of societies permitted selection of
cultures with weaker cultural and historical diffusion rela-
tionships – i.e., the SCCS was constructed to maximize
independence in terms of cultural and historical origin
while preserving a large enough sample to permit suf?-
ciently powerful statistical tests.
Murdock (1967) initiated the pinpointing process when
he analyzed nearly 1300 societies chosen for the complete-
ness of their ethnographic coverage to construct his Ethno-
graphic Atlas. He classi?ed these societies into clusters
based on the similarity of the cultures and categorized
groups of clusters into 200 ‘‘sampling provinces” (Mur-
dock, 1968; Murdock & White, 1969). From the initial
200 sampling provinces, two had no culture that could be
accurately pinpointed to a particular locale and date, two
were split in half, and 14 others were dropped because
they were too similar to others in the sample.
Murdock and White (1969) then identi?ed that culture
from each of the 186 sampling provinces with the earliest
period of satisfactory ethnographic coverage unless signif-
icantly richer data were available for a later period. The
186 cultures selected in this step comprise the SCCS. The
cultures in SCCS are assigned a number from 1 to 186,
which facilitates statistical identi?cation of those cultural
practices that originate from a common cultural heritage.
This is done because societies with close geographical
proximity are likely ones where cultural diffusion may still
be a prominent force. Thus, the correlation between adja-
cent societies within the SCCS measures the extent to
which the pinpointing process did not completely elimi-
nate Galton’s Problem.
The initial study using the SCCS coded 22 variables re-
lated to subsistence economy and related practices (Mur-
dock & Morrow, 1970). The free electronic database
includes a bibliography of ethnographic sources for each
society-year, ranked on a scale from 1 ‘‘Principal Author-
ity(ies)” to 6 ‘‘Sources to be Avoided” (White, 1986). The
hard copy archives include coding sheets indicating which
individual sources were relied on most for each culture for
coding different groups of variables. Researchers who use
the database code new variables and these additional vari-
ables are added to the database as a result. Not all 186 soci-
eties are coded for each new variable as some researchers
elected to code data for only a subset of the SCCS cultures.
For example, many variables are coded for only 93 cultures
suggesting, for example, a sampling scheme such as using
every other culture in the database.
Thus, eachnewstudyincreases thedepthof thedatabase.
There are presentlymore than2000categorical variables (as
of 2007) coded nominally or ordinally by over 60 different
studies.
21
Unlike the usual market studies, the data we use
are limited to only one observation per culture; thus, SCCS
does not provide a pooled time-series cross-sectional data
set. The SCCS was designed to ensure that standard errors
are not in?ated by multiple observations from the same unit.
20
Previous researchers had tended to analyze their own selection of
speci?c societies, which often was based on small samples that were not
comparable across studies. The SCCS helped standardize researchers’ choice
of societies and has improved cross-study comparability.
21
An online journal, World Cultures, founded by Douglas White in 1985,
maintains, re?nes and expands the SCCS. The journal is available in paper
and CD-ROM as well as for free over the Internet. The journal can be
accessed athttp://eclectic.ss.uci.edu/~drwhite/worldcul/world.htm.
S. Basu et al. / Accounting, Organizations and Society 34 (2009) 895–917 911
Appendix B
De?nitions of SCCS variables used in empirical tests
SCCS variable Coding
Panel A: SCCS variables ?rst introduced in ‘‘The history of transaction records and recordkeeping in SCCS societies”
V149: Writing and Records 1 = None
Variable #149 coded by Murdock and Provost (1973, pp. 379–
380)
2 = Mnemonic devices
3 = Nonwritten records
4 = True writing; no records
5 = True writing; records
Panel B: SCCS variables ?rst introduced in ‘‘Recordkeeping emerges as group size increases”
V18: Credit source 1 = Personal loans between friends or relatives
Variable #18 ?rst coded by Murdock and Morrow (1970, p. 306) 2 = Internal money lending specialists
3 = External money lending specialists
4 = Banks or comparable institutions
V63: Community size 1 = 50,000
V89: Judiciary 1 = Absent
Variable #89 ?rst coded by Tuden and Marshall (1972, p. 441) 2 = Not local
3 = Executive
4 = Appointed by executive
5 = Priesthood
6 = Hereditary
V91: Administrative hierarchy 1 = Absent
Variable #91 ?rst coded by Tuden and Marshall (1972, pp. 441–
442)
2 = Popular Assemblies
3 = Heads of kin groups
4 = Heads of decentralized territorial divisions
5 = Heads of centralized territorial divisions
6 = Part of centralized system
V155: Money 1 = None
Variable #155 coded by Murdock and Provost (1973, p. 381) 2 = Domestically usable particles
3 = Alien currency
4 = Elementary forms
5 = True money
V234: Settlement patterns 1 = Nomadic or fully migratory
Variable #234 ?rst coded by Murdock (1967, p. 159) 2 = Seminomadic
3 = Semisedentary
4 = Compact but impermanent settlements
5 = Neighborhoods of dispersed family homesteads
6 = Separated hamlets, forming a single community
7 = Compact and relatively permanent settlements
8 = Complex settlements
V278: Inheritance of land
22
0 = Absence of individual property rights or rules
Variable #278 ?rst coded by Murdock (1967, p. 167) 1 = Inheritance based on familial ties
V1122: Population size 1 = 10–99
912 S. Basu et al. / Accounting, Organizations and Society 34 (2009) 895–917
Appendix B (continued)
SCCS variable Coding
Variable #1122 coded by Douglas White and is described in
Standard Cross-Cultural Codes, edited by White, Burton, Divale,
Gray, Korotayev, and Khalturina at:http://eclectic.ss.uci.edu/
~drwhite/courses/SCCCodes.htm
2 = 100–999
3 = 1000–9999
4 = 10,000–99,999
5 = 100,000+
6 = 1,000,000+
7 = 10,000,000+
8 = 100,000,000+
V1130: Population density 2 = less than 1 per square mile
Variable #1130 coded by Pryor (1984) 3 = 1–4.9 per square mile
4 = 5–24.9 per square mile
5 = 25–99.9 per square mile
6 = 99–499.9 per square mile
7 = 500 or more per square mile
Panel C: SCCS variables First Introduced in ‘‘The in?uence of recordkeeping on impersonal exchange and division of labor”
Variables used to measure Exchange
V1: Intercommunity trade as food source 1 = No trade
Variable #1 coded by Murdock and Morrow (1970) 2 = No food imports
3 = Salt and minerals only
4 =