outwaite-3587-14.qxd 6/12/2007 5:32 pm page 246 246...

QUANTIFICATION AND EXPERIMENT246

WHAT IS QUANTITATIVE HISTORY?

Quantitative history is the term for an arrayof skills and techniques used to apply themethods of statistical data analysis to thestudy of history. Sometimes also called clio-metrics by economic historians, the term waspopularized in the 1950s and 1960s as social,political and economic historians called forthe development of a ‘social science history’,adopted methods from the social sciences,and applied them to historical problems.These historians also called for social scien-tists to historicize their research and con-sciously examine the temporal nature of thesocial phenomena they explored. For bothtypes of questions, historians found that theyneeded to develop new technical skills anddata sources. That effort led to an array ofactivities to promote quantitative history.

Classical historical research methodologyrelies upon textual records, archival researchand the narrative as a form of historical writ-ing. The historian describes and explains par-ticular phenomena and events, be they largeepic analyses of the rise and fall of empiresand nations, or the intimate biographicaldetail of an individual life. Quantitativehistory is animated by similar goals but takesas its subject the aggregate historical patterns

of multiple events or phenomena. Such astandpoint creates a different set of issues foranalysis. A classic historical analysis, forexample, may treat a presidential election asa single event. Quantitative historians con-sider a particular presidential election as oneelement in the universe of all presidentialelections and are interested in patterns whichcharacterize the universe or several unitswithin it. The life-course patterns of onehousehold or family may be conceived asone element in the aggregate patterns of fam-ily history for a nation, region, social class orethnic group. Repeated phenomena from thepast that leave written records, which readone at a time would be insignificant, are par-ticularly useful if they can be aggregated,organized, converted to a electronic databaseand analyzed for statistical patterns. Thusrecords such as census schedules, vote tal-lies, vital (e.g., birth, death and marriage)records; or the ledgers of business sales, shipcrossings, or slave sales; or crime reportspermit the historian to retrieve the pattern ofsocial, political, and economic activity in thepast and reveal the aggregate context andstructures of history.

The standpoint of quantitative history alsorequired a new set of skills and techniquesfor historians. Most importantly, they had to

14Quantitative History

M a r g o A n d e r s o n

Outwaite-3587-14.qxd 6/12/2007 5:32 PM Page 246

incorporate the concept of the data set anddata matrix into their practice. Floud (1972:17) defined the data set as ‘a coherent selec-tion of data from the whole range of historicaldata available to the historian, and it isselected because it relates closely to the ques-tions that the historian wishes to consider.’The myriad instances of a phenomenon—for example, all United States presidentialelections—form the cases of the data set. Thepieces of information collected about thecases—for example, the candidates running,the year of the election or the vote totals—become the variable characteristics of thedata set, that is, the varying characteristics ofany particular case. The historian arrangesthe data in tabular form, that is, in a matrix ofrows and columns, ‘consisting of a numberof rows, which will normally represent cases,and a number of columns, which will nor-mally represent variables’ (Floud, 1972: 18).The creation of quantitative data sets thusrequired the historian to carefully compileconsistent information about the phenome-non to be investigated, and prepare the datain tabular form. Historians then were pre-pared to apply the techniques of statisticaldata analysis to the data set to answer theresearch question posed.

In short, to make effective use of quantita-tive evidence and statistical techniques forhistorical analysis, practitioners had to inte-grate the rapidly developing skills of thesocial sciences, including sampling, statisti-cal data analysis and data archiving into theirhistorical work. That task led to the develop-ment of new training programs in quantita-tive methods for historians, to the creation ofnew academic journals and textbooks, and tothe creation of data archives to support theresearch.

EARLY EFFORTS

Historians had made use of quantitative evi-dence prior to the 1950s, particularly in thefields of economic and social history. TheAnnales school in France pointed the way inthe pre-World War II period. The rapid

growth and expansion of the United Stateshad long required American historians toconsider quantitative issues in their study ofthe growth of the American economy, popu-lation and mass democracy. Thus, for exam-ple, Frederick Jackson Turner’s classic 1893essay on ‘The Significance of the Frontier inAmerican History’ was largely based on areading and interpretation of the results ofthe 1890 population census.

But true ‘data analysis’ in the current sensehad to await the growth of the social and sta-tistical sciences in the first half of the twenti-eth century, and the diffusion to universitiesin the 1950s of the capacity for machine tab-ulation of numerical records, and then ofmainframe computing in the 1960s. One cansee the emerging field exemplified in semi-nal studies in the late 1950s and early 1960s.In 1959, for example, Merle Curti and hiscolleagues at the University of Wisconsinpublished The Making of an AmericanCommunity: A Case Study of Democracy in aFrontier County. Curti et al. (1959) exploredTurner’s thesis with an in depth lookat the mid-nineteenth century history ofTrempeleau County, Wisconsin, including itsrecords of newspapers, diaries, privatepapers and county histories. But they alsoadded data analysis of the employment pat-terns derived from the individual-level fed-eral census manuscripts for the censusesfrom 1850 through 1880.

Similarly, the ‘new’ economic historiansof the 1950s challenged the conventionalwisdom of the day on several key issues ineconomic history. One debate centeredon the ‘necessity’ of the US Civil War.Historians at the time argued that the war hadbeen ‘unnecessary’ since the institution ofrace-based slavery would collapse under theweight of its unprofitability. In contrast, eco-nomic historians employed economic theoryand data on output of southern agriculture toargue that the southern agricultural economycould have survived profitably into the twen-tieth century using slave labor (Conrad andMeyer, 1958). Robert Fogel challenged theconventional wisdom on the centrality ofrailroads for the industrial development of

QUANTITATIVE HISTORY 247

Outwaite-3587-14.qxd 6/12/2007 5:32 PM 47

QUANTIFICATION AND EXPERIMENT

the United States. Making use of economictheory, carefully compiled data series, andthe logic of the counterfactual, Fogelargued that canals would have also suc-ceeded as a transportation system underpin-ning nineteenth-century American industrialdevelopment (1964).

‘New political historians’ such as LeeBenson, Allan Bogue, Richard P.McCormick, and political scientists with his-torical interests, such as Warren Miller andWalter Dean Burnham, translated the emerg-ing techniques of political scientists analyz-ing contemporary election results and votersurveys to historical questions, and openedup dramatic new insights into Americanpolitical history.1 The new political historiansidentified the parameters of party systems,developed the theory of the critical election,and argued that underlying structures of elec-toral politics were accessible through histor-ical analysis of voter turnout and electionresults. In 1964 in England, demographersand historians founded the Cambridge Groupfor the History of Population and SocialStructure and began a forty-year project toretrieve, assemble and reconstruct 400 yearsof the family history of Britain.2

The new possibilities of quantitativehistory fit well with other trends within thediscipline of history, particularly with thegrowth of social history and calls for whatJesse Lemisch (1967) called ‘history fromthe bottom up’—that is, for historians totreat the lives of ordinary people, to comple-ment the study of elites. By the mid-1960s,the interest in the new techniques led theAmerican Historical Association to recog-nize that ‘quantification in history’ wouldrequire new skills and institutions within thehistorical profession. The AHA created aQuantitative Data Committee to consider theissues. Summer institutes and classes inquantitative methods for historians were heldin 1965, 1967 and 1973 at the University ofMichigan, Cornell University and HarvardUniversity respectively. In 1968, the Inter-university Consortium for Political Researchat the University of Michigan began offeringa four-week course in quantitative historical

analysis as part of its summer program inquantitative methods. The course continuesto be offered each summer. At the NewberryLibrary in Chicago, from 1971 to 1982Richard Jensen spearheaded a summer pro-gram in quantitative methods for historians.By the early 1980s, about 40 percent ofhistory graduate programs offered training inquantitative history as part of the graduatecurriculum (Bogue, 1983: 220ff.).3

Additional institutional infrastructure ofquantitative history can also be dated to the1960s. New journals, textbooks, and editedcollections also promoted the growth ofquantitative history. The Historical MethodsNewsletter, for example, began publishing in1967, and was renamed Historical Methodsin 1978. The Journal of InterdisciplinaryHistory began publication in 1970. TheSocial Science History Association (SSHA)was founded in 1974 and the first issue of itsjournal, Social Science History, appeared in1976. SSHA became the professional venuefor bringing together historians who con-sciously adopted theory and methodologyfrom the social sciences and social scientistsdoing historical work. The cross-fertilizationhas continued, and, as noted below, many ofthe innovations in quantitative history havebeen developed by scholars with formaltraining in the social sciences and appoint-ments in departments of economics, demog-raphy, sociology, anthropology, geographyand political science.

Textbooks in quantitative history began toappear in the early 1970s, and many havebeen published since.4 Numerous editedvolumes introduced the new field andtechniques to professional and student audi-ences.5 Finally, researchers created dataarchives. In the United States, the Inter-university Consortium for Political Research(ICPR) was founded in 1962 primarily bypolitical scientists. Renamed the Inter-university Consortium for Political andSocial Research (ICPSR) in 1975, theConsortium has also pioneered in thecreation and preservation of historical datacollections. The United States NationalArchives and Records Administration (NARA)

248


created an electronic records preservationprogram in the early 1970s for federal gov-ernment data that was ‘born digital’(Ambacher 2003; Adams, 1995, forthcom-ing; Fishbein, 1973). Similar work began inBritain with the founding of the UK DataArchive in 1967.6

Thus by 1980, historians had take majorsteps to establish the institutional structuresnecessary to integrate quantitative historyinto larger historical practice. That infra-structure has, if you will, both matured andfaced challenges in the generation of worksince, and in many ways quantitative historyis still a work in progress. Nevertheless, it ispossible to identify the types of questionsquantitative history was intended to and hasbeen able to address; the major types of datasets that have developed and the key charac-teristics of historical data sets; and the mostcommonly used techniques within the field.That background in turn provides the frame-work for a review of a number of method-ological issues historians uniquely face, for areview of the achievements of quantitativehistory, and for a discussion of emergingissues.

QUESTIONS, DATA AND ISSUES INCREATING HISTORICAL DATA SETS

Quantitative history has been most successfulin addressing big questions about long-termhistorical patterns of change. Practitionershave achieved important results by assem-bling substantial amounts of numeric orcountable information, and organizing it intotabular data matrices for statistical analysis.The first generation of studies focused espe-cially on the history of the family and socialstructure, trends in economic growth andchange, patterns of electoral behavior andvoter participation, or the record of inter-generational social mobility and living stan-dards. More recently, the examples haveproliferated. Historians of crime and thecriminal justice system, for example, haveretrieved court and newspaper records toexamine the long-term patterns of crime and

violence in the past. Historians of the familyhave examined patterns of inheritance and theinter-generational transfer of wealth. Theemerging work of ‘anthropometric’ history—the study of living standards and well-beingin the past using measures of height, weight,stature and disease in the past—has cast aneven wider net, aiming to evaluate compara-tive living standards over centuries and ulti-mately millennia.7

Making such studies possible was anexplosive growth in the data sets informingquantitative history. Quantitative history, likeother branches of the social sciences,requires what was once called ‘machine-readable’ (and are now known as ‘elec-tronic’) data for analysis. Though there aresome examples of large-scale data analysisundertaken by manual systems of tabulationand statistical analysis, most notably thenineteenth-century tabulations of census orvital registration records, social science datain the modern sense required the develop-ment of machine tabulation devices, countersorters, and other mechanized calculators.The first system was the Hollerith system ofpunch-card tabulation used for the 1890American population census; the social andstatistical sciences grew with the newmachinery. By the 1940s, social scientistshad developed rules and procedures for col-lecting quantitative data to make best use ofmachine tabulation and analysis. These con-ventions included the fixed format datamatrix, the classification of variables intonominal, ordinal, interval and ratio variables,the organization of questionnaires and surveyforms to facilitate conversion to punch-cardsfor analysis, and coding systems such as theLikert scale. Quantitative historians inheritedthese practices and adapted this existingtechnology and set of conventions to theirhistorical project. The soon recognized thatthey had to solve major new methodologicaland logistical problems before the potentialfor quantitative history could be achieved.

The first problem derives from the largerevidentiary issue faced by all historians,namely, that historical analysis must rely onthe extant record of the past. Historians are at




the mercy of their subjects’ penchant andcapacity for preservation. And before 1890,that is, for most of the historical record ofhuman history, no preserved data were‘machine-readable’. Thus all potential histor-ical data had to be created from surviving,usually text-based, records and converted tomachine-readable or electronic format. Evenrecords collected in the twentieth century andinformed by the conventions of the emergingsocial sciences frequently no longer exist inmachine-readable format. Thus, the UnitedStates Census Bureau, for example, pre-served the original paper census question-naires from the eighteenth century forward.But census officials did not retain the punch-cards they used to tabulate the censuses from1890 to 1960. These cards were destroyedonce the results of the census appeared inpublished form. Thus historians interested inreanalyzing the microdata from past cen-suses faced creating, or recreating, themachine-readable records.

Quantitative historians faced additionalmajor methodological problems resultingfrom the recalcitrance of the existing archivalhistorical records. All historians face theproblems of missing data, and the difficultiesof interpreting illegible, damaged, incom-plete or destroyed records. For quantitativehistorians, though, aiming to translate thearchival record to a data matrix for statisticalanalysis, these questions of data quality areparticularly difficult. Cases and variables fora data matrix require precise conceptual andoperational definitions, as do the allowableentries for particular cell values within thematrix, since the goal of statistical analysis isto assess extent, central tendency and disper-sion of any particular characteristic. Whatdoes one do if the records for a year or periodof years are missing? How does one handleillegible entries in the records of a company’sfinances? How does one know if the probaterecords found in a county archive are com-plete? Historians have had to confront therequirements for case and variable definition,classification and coding in building a dataset. The solutions to these problems emergedwith the overall field. The journal, Historical

Methods, in particular, became the venue foridentifying, debating and proposing method-ological solutions to these issues.

A related issue is the set of rules forextracting the information from a text-basedevidentiary source to create a data set.Historical archives frequently contain text-based records that lend themselves to data setconstruction, but require considerable con-ceptual work before they can be manipulatedstatistically. Historians have made use ofsales invoices, wills, parish registers and casefiles of charity or social welfare agencies, forexample, and have had to create the casesand variables from the extant texts.

Historians have had to solve these method-ological questions as they select the evidenceto be analyzed and create the code-book forthe data set. Whether one is analyzing exist-ing tabular data from the past—for example,the records of imports and exports of a nationover a period of years, or the publishedresults of a census—or whether one is creat-ing a data set from text-based sources, thehistorian needs to define the case or unit ofanalysis, define the characteristics or vari-ables to be selected to characterize the caseswithin the data set, and define the codingsystem used to organize the source informa-tion for the data set. Several examples of theissues involved best illustrate the work ofquantitative historians.

COMPILATION AND ANALYSIS OFPREVIOUSLY PUBLISHED DATA

The most accessible sources for quantitativehistorians were data that were alreadypublished in tabular format. The first genera-tion of quantitative historians in particularcompiled data sets from existing, usuallyaggregated, published data sources—forexample, tabulated census results, electionresults, government reports of tax collec-tions, imports and exports, and data fromtrade publications. Assembled into timeseries, such data permitted researchers toundertake basic analyses of historical trendsand use regression models to correlate the

250


determinants of change. For example, WalterDean Burnham’s 950+-page study ofcount-level presidential election results, pub-lished in 1955, included a compilation ofresults from state archives and newspapersources, and a discussion of the methodolog-ical issues he faced in compiling the data.Combined with denominator data fromcensus results that allowed the researchers tomeasure turnout, the new data set permittedBurnham and his colleagues to begin theanalysis of historical election analysis(Burnham, 1955). In similar ways, economichistorians made particularly good use thedata compiled in statistical abstracts, such asthe Statistical Abstract of the United States,published annually since 1878.

CONVERTING TABULAR DATA INMANUSCRIPT FORM TO ELECTRONICFORMAT

A second source of quantitative data werearchived tabular records in text-based for-mat, probably best illustrated by individual-level census manuscript schedules. SeeFigure 14.1,8 a facsimile of the 1950 USCensus population schedule.

For the United States, such original censusresponses are available for all the federalcensuses except 1890, and are available forpublic use through 1930.9 The schedule isalready in a matrix format, with rows ofcases and columns of variables. The originaldifficulty with using these records is theirvolume. With one record per person for thecensuses of 1850 and later, data set creationfor a large portion of the population wasbeyond the capacities of an individualresearcher. The first generation of quantita-tive historians resolved this problem by sam-pling, and usually by organizing a researchproject of a particular locale. The historicalsocial mobility studies were designed ascommunity studies to solve the problem ofthe volume of data.

Later generations of quantitative histori-ans have by and large solved the problem of

volume through collaboration and by build-ing historical public-use microdata samples,or PUMS files. Starting with the 1900census, historians proposed to create histori-cal PUMS files that would be similar to thecontemporary PUMS files that the CensusBureau has created since 1970. In the late1980s, researchers at the University ofMinnesota, initially led by Russell Menard,Steven Ruggles and Robert McCaa, begansystematic retrieval of the historical censusdata from the United States, and morerecently from other nations. The IntegratedPublic Use Microdata Sample (IPUMS)Project and the International IPUMS projecthave created microdata samples for theUnited States from all the censuses from1850 to 2000, and are now collecting suchdata for many nations of the world. The dataare easily downloadable from the web. Theresearchers have also built the code-books,technical support materials, and researchbibliography necessary for the user to under-stand the context of the questions andresponses to the census.10

Creating Tabular Data from TextBased Records

The most time-consuming type of data setcreation is the conversion of text-basedrecords to matrix format. For existing tabulardata, whether in manuscript or publishedform, the basic framework of the matrix isgiven in the original source. For text-basedrecords with no tabular structure, it is up tothe researcher to create the code-book, andthus all the variable definitions and codingrules. Figure 14.3,11 an illustration of a recordof a slave sale in antebellum America, illus-trates the issues.12

There are thousands of such records innewspapers, private collections and archives,and, if marshaled for analysis, providedetailed, if somewhat gruesome, evidence ofthis chapter in American economic history.Robert Fogel and Stanley Engerman com-piled such records for their study, Time on theCross (1974) from the New Orleans Slave




Market. ICPSR Study 7423 contains the dataand code-book for the New Orleans SlaveSale Sample.13

For their sample, Fogel and Engerman con-verted the text-based records into cases andvariables and codes, making decisions on unit

of analysis (the slave), sampling (2.5 percentor 5 percent, depending on the year of sale),number of variables (46), and codes. Eachdecision extracted a piece of informationfrom the original text-based records, and hadimplications for ultimate analysis. The final

252

Figure 14.1 Facsimile of 1950 Census Schedule for Orange City, Iowa


data set contained 5009 records, and includedinformation on the characteristics of the slave(e.g., age, sex, occupation, color), the termsof the sale (e.g., the date, price, whether paidin cash, the number of slaves sold together),and information on the buyer and seller.

Fogel and Engerman used the data to analyzethe inter-state slave trade, and to addressquestions about the economic viability of theslave economy (Fogel and Engerman, 1974).

The work of building the corpus ofmachine-readable databases began in the


Figure 14.2 Facsimile of 1950 Census Schedule for Orange City, Iowa



1960s, and continues both with small compi-lations and large collaborative data projects.In addition to the IPUMS project mentionedabove, one can find large-scale historical datacompilations of cost-of-living studies, elec-tion results, crime data, and the records of theheights and weights of people in the past. Thecreation and retrieval of historical data hasalso led to revision and improvement of dataseries compiled in earlier years and to theanalysis of the history of data development.Most recently, for example, economic histori-ans have produced a new ‘millennial’ editionof the Historical Statistics of the UnitedStates (Carter et al., 2006), which promises toprovide opportunities for even more quantita-tive historical analysis.

ANALYZING HISTORICAL DATA SETS

Sampling and the Universe of Cases

As have data analysts in the other socialsciences, historians have made use of thetheory of probability sampling to reduce thevolume of information for a particular study

to a manageable level. Just as one does notneed to survey the entire electorate todevelop quite precise estimates of the ulti-mate election results, so historians studyingfamily structure or economic activity or con-sumer behavior have not had to record allsuch behavior for study. As noted above, theprocess of creating historical data sets is suf-ficiently time-consuming to strongly recom-mend sampling strategies designed to reducethe volume of coding and data entry to theminimum necessary for robust analysis. Thusthe original users and secondary users of thearchived historical data sets need to attend tosampling strategy and introduce appropriatesample weights and measures of error intothe analysis.

A more difficult issue is the one facing thehistorian who cannot be sure that she knowswhat the universe of cases actually is. Do theextant newspaper reports of lynchings, forexample, encompass all lynchings (Griffinet al., 1997; Tolnay and Beck, 1995)? Are therecords of wills filed with a particular countycomplete, or might some have been destroyedor lost over the centuries? These dilemmashave their analogues in non-quantitative

254

Figure 14.3 Slave Bill of Sale, Davidson County, Tennessee, 1833


research. But as with code-book creation, theresearch must provide best estimates ofanswers to such questions before analysis,and a substantial methodological literaturehas emerged to address the issues, often withspecific reference to the kind of data set beingcompiled.

Techniques of Analysis

Statistical analysis of historical data hasranged from elementary data analysis of thepatterns of central tendency and dispersion ofthe phenomena under study to elaborateexplanatory models of events and behavior.Much historical quantitative analysis hasbeen descriptive, simply excavating and doc-umenting patterns of change and activity inquantitative form that cannot be revealed bytraditional historical analysis. Thus muchwork—important work—is simple countingof a phenomenon, and describing trends overtime.

Somewhat more elaborate analysisinvolves determining the correlates of thephenomenon under study, or building amodel to explicate more complex patterns inthe data. Here the standard bivariate andmultivariate techniques of statistics providethe tools necessary for the analysis.Quantitative historians have borrowedheavily from sociology, political science,demography and economics, and made useof the classic linear regression model and itsvariants as the workhorse technique for morecomplex analysis. Statistical packages, suchas SPSS, SAS, STATA and the like underpinthe analysis of quantitative historical work,as they do for the social sciences.

There is some evidence that quantitativehistory has begun to have an impact on thelarger methodological practice of the socialsciences, as quantitative historians havebrought their methodological expertise to thesocial sciences. Two brief examples shouldillustrate that impact.

The first is development of the field ofecological regression, particularly for analy-sis of electoral patterns. Political scientistscan supplement analysis of election results

with surveys of individual voters. Indeed theNational Election Survey, conducted since1948, has itself become an historical sourceof changing electoral behavior. But histori-ans cannot go back and survey voters fromthe election of 1860, and thus must make useof the aggregate election results and the eco-logical characteristics of the voting units—e.g., precincts, districts or counties—thatprovided the vote. Ecological inference suf-fers from the threat of the ecological fallacy,that is, the danger of wrongly inferring indi-vidual level behavior from the patterns ofaggregates. Practitioners of quantitativehistory have taken up new methods devel-oped by political scientists and have devotedgood effort to minimizing, if not completelysolving, this dilemma. With historicallyminded political scientists, they have pro-duced a methodological literature and newtechniques that have produced rigorousresults.14

The second contribution is serious atten-tion to the development of statistical tech-niques to conceptualize and model time andtemporal explanations. The methodologicalbread and butter for all historians is ‘thinkingin time’ (Neustadt and May, 1986), and thatstandpoint has prompted historians and his-torically attuned social scientists to thinkabout how to develop techniques of statisticalanalysis suitable for the goals of historicalanalysis.

Historians think about questions of what isan event, how is it bounded and measured;what is a turning point; what is a transition;what is a conjuncture or a rupture; and howis a period of time organized and bounded.Economists and other social and biologicalscientists have developed techniques to mea-sure time series and temporal and cyclicalevents, for example, life cycles. The entry ofquantitative historians into these discourseshas been a useful clarification of the method-ological issues involved. For example, thephrase, ‘longitudinal analysis’ that social sci-entists use does not necessarily privilege timeas a central concern for analysis. Historiansand social scientists who make temporalanalysis such a central concern have thus




argued for the need to add methods that willaddress ‘thinking in time’ to the standardrepertoire of statistical techniques. Suchtechniques as sequence analysis, eventhistory analysis and the methodological dis-cussions surrounding autocorrelation in timeseries analysis have usefully been enrichedby the growth of the field of quantitativehistory.15

THE COSTS OF DOINGQUANTITATIVE HISTORY

The cost of scholarly work in quantitativehistory, like the cost of all scholarly work,can be measured in terms of both time andmoney required for the scholarship to flour-ish. The largest change in the working envi-ronment since the 1960s is that computingcosts, which were quite expensive in theearly years of the field, have dropped as thelarger information revolution has developed.To my knowledge, there is no extant schol-arly analysis of the costs of quantitativehistory versus traditional history, though Isuspect that the underlying funding situationfor quantitative historians has had an effecton the progress of the field.

In the early years of the development ofquantitative history, in the United States theSocial Science Research Council, theAmerican Historical Association, the NationalEndowment for the Humanities, and theNational Science Foundation, as well asresearch universities around the country, allprovided sponsorship of the field by fundinggrants for data development, conference spon-sorship and the institutional work required topromote the field. This early institutional sup-port was aimed at jump-starting the field, notat providing sustained long-term support.Related to this, the National Endowment forthe Humanities, the main federally sponsoredgrant agency for historians, has a much lowerfunding level than the federal funding agenciesthat support related social science research—for example, the National Science Foundationor the National Institutes of Health. The UnitedStates, unlike European nations, does not

include history in the main governmentalfoundation for funding academic research.Accordingly, quantitative history projects inthe United States have had major difficulty incompeting with both large-scale humanitiesgrant projects, such as compilations of archivalpapers, and with large-scale long-termresearch projects in the social sciences such asthe National Election Study or the Panel Studyof Income Dynamics. Allan Bogue (1983) iden-tified the chronic problems of funding faced byquantitative historians in the late 1970s. Theyremain unsolved as the concrete example whichfollows illustrates.

Robert Fogel, by any measure, representsone of the most successful and innovativequantitative history scholars in the field, yeteven he has faced major funding obstacles.Fogel was awarded the Nobel Prize inEconomics in 1993, and in his autobiograph-ical statement prepared for the award, hedescribed his career and acknowledged theproblems of funding he faced, particularly, ashe put it, for the ‘current research projects onwhich I reported in the Prize Lecture’. TheCenter for Population Economics at theUniversity of Chicago and the WalgreenChair provided funding when federal grantswould not. ‘The data on health conditions’,he wrote:

comes from a project called ‘Early Indicators ofLater Work Levels, Disease, & Death’ which is trac-ing nearly 40,000 Union Army men from the cra-dle to the grave. It takes over 15,000 variables todescribe the life-cycle history of one of these men.These life-cycle histories are created by linkingabout a score of data sets. It took more than halfa decade of work to investigate the potential ofthese data sets, work out procedures for dataretrieval and file management, and to establish thefeasibility of the enterprise in our own minds.

The site committee of the National Institutes ofHealth which reviewed the original project pro-posal in 1986 agreed that such a project could inprinciple make a significant contribution to anunderstanding of the process of aging, but theywere skeptical about the quality of some of thedata, about whether the software and program-ming procedures we had developed by that timewere adequate for the management of such alarge data set, and about whether the projectcould be completed within the proposed budget.

256


To resolve these doubts it was necessary to draw asix percent subsample which linked together allof the separate sources and which demonstratedthe effectiveness of the software by analyzing theinformation in the subsample. It took an additionalfour years to complete the second phase of thejustification of the project. Thus nearly a decade ofpreliminary research, much of it funded byWalgreen and the CPE, was required before theproject was accepted by the peer reviewers of NIHand NSF.16 (Fogel, 1993)

Despite such barriers, quantitative histori-ans have been able to take advantage of thetechnological developments in computingand data management to make majoradvances in the ease of analysis, in terms ofboth time and money. For example, historiansof the 1960s through the 1980s who wished tohave access to the archived data sets atICPSR had to order tapes and paper code-books which were delivered by mail. The tapewas then mounted on a mainframe computer,to be accessed in a statistical package run in amainframe environment (with computerusage often charged by the university in thesame way that phones or paper werecharged). By the early 1990s, users couldaccess files using FTP (file transfer protocol),and micro-computers on university desktopswere providing direct access to statisticalpackages, even if those programs were some-times still lodged on a mainframe. By themid-1990s, desktop computing had replacedmainframe computing for most applications,and by the early 2000s, ICPSR initiatedICPSR Direct, the application that permittedan authorized user to download data files andPDF code-books directly to a desktop.

CRITIQUES OF QUANTITATIVEHISTORY

From the outset of the development of thefield of quantitative history, powerful criticshave challenged practitioners on their work,and even challenged the usefulness of the fielditself. In the early 1960s, Carl Bridenbaughdevoted a portion of his 1962 AmericanHistorical Association Presidential Addressto a condemnation of quantitative history

(Bridenbaugh,1962), memorably labeling it a‘bitch goddess’ (Bogue, 1983). Even duringthe period of the rapid growth of quantitativehistory in the 1960s and 1970s, ‘traditional’historians expressed doubts about the newmethods, challenging them as reductionist,brittle and not pertinent to the main goal of thehistorical narrative. Critics were extremelydubious of the ‘scientific’ claims of quantita-tive historians, and resisted the challenge ofthe quantifiers that traditional historical writ-ing was not theoretically rigorous or concep-tually consistent.

In the 1980s, some of the original propo-nents of the field also renounced their earlierenthusiasm and suggested that quantitativemethods had not fulfilled their promise. Mostnotable among these critics were Lee Bensonand Lawrence Stone, early enthusiasts whohad changed their minds (Benson, 1984;Stone, 1977, 1979). Such recantations gavesupport to the anti-quantifiers at a time whenmajor new methodological challenges werefacing historians, most notably from thepostmodernists and what came to be called‘the cultural turn’. Through this welter ofdebate, quantitative practitioners continuedtheir efforts, somewhat chastened by theirfall from the heights of fashion of earlieryears, but grounded sufficiently institution-ally and intellectually to continue to work.17

Through some twenty years of debate, nei-ther side of the traditional/quantitative divide‘won’ their arguments. Rather, by the 1990s,the debate cooled into something of anuneasy truce, with practitioners acknowledg-ing some of the points of their opponents, butagreeing to disagree on the larger validity oftheir enterprise.18 In practical terms, quantita-tive techniques did not become a routine partof history graduate student training as theydid in the social sciences, but have remaineda specialty of some historians in some grad-uate training programs, considered more akinto language requirements for reading histori-cal literature and texts of a non-English-speaking society than to a methodologicalnecessity for all practicing historians. Thiscompartmentalization of the skills of quan-tification for historians has in turn affected




the practice of quantitative historians withinthe larger history profession.

History as a field has maintained its rootsas a ‘humanities’ discipline and quantitativehistorians’ connections to the social sciencesseem to many to be a betrayal of the histori-cal project. The methodological ‘traininggap’ has meant that when quantitative histo-rians research and write for other historians,as opposed to other social scientists, theycannot expect their readers to appreciate oreven understand the technical issuesinvolved in their work. The history profes-sion has maintained its commitment toaccessible writing as well, and thus whenwriting for the broader audience of histori-ans, quantitative historians have had to avoidtechnical jargon—for example, by avoidingthe use of variable names in the explicationof a model—and be mindful to explicatetheir arguments clearly.

The critiques have also encouraged quantita-tive historians to attend to the limitations in sta-tistical methodology for analyzing historicalprocesses, as discussed above. Much of thisnew work on statistical techniques for analyz-ing temporal processes is still in developmentand has yet to provide enough empirical workto demonstrate the robust nature of the newtechniques, and hence convince non-quantita-tive historians, as well as the larger socialscience community, of the need to integrateexplicitly temporal analysis into basic methods.But the promise is there, and as noted below,there are encouraging signs on the horizon.

THE FUTURE OF THE FIELD

The intellectual achievements of quantitativehistory in conjunction with the larger infor-mation technology revolution makes theprognosis for the future of the field bettertoday than it has been for many years.19

Almost a half-century on, one can look backat steady development, though not alwaysin a satisfyingly linear pattern.20 Perhapsthe most interesting recent development isthe impact of the information technologyrevolution on the larger practice of historians.

When quantitative history as a field was in itsmost rapid initial development, most tradi-tional historians labored much as theirnineteenth-century predecessors had withpen, pencil, typewriter and note-card as tech-nological support. Bibliographic workentailed using library card catalogs or readinglarge indexed tomes of articles, books, com-pilations, and the like. ‘Data management’meant developing a file of index cards, not anelectronic spread sheet or database.Secretaries typed manuscripts for publication,and though some large research institutionshad introduced line editors for manuscriptproduction by the 1970s, these weremachines for staff, not faculty or students. Bythe 1980s, the situation changed. Desktopcomputers proliferated and for most histori-ans, word processing opened up the possibil-ity of the electronic future. By the 1990s,email replaced typed letters. After 1995, thecontent on the internet exploded, and firstbibliographical work, and then much actualarchival work, shifted to a computerized for-mat. In short, non-quantitative historians hadcome to operate in a technological environ-ment that was very similar to their quantify-ing peers. Most recently, cheap computinghas made multimedia evidence—visual andoral, video and audio—accessible to the prac-ticing historian. One can see these develop-ments in particularly acute form in thedeveloping field of historical geographicinformation systems, or historical GIS. GIS,was until quite recently, a very expensivetechnology, and thus adding historical mapsto geographic databases has only just begun.As with the digitizing projects of the 1960sand 1970s, the payoff for the large initialcosts of first translating maps to a newmedium to become ‘data’, and then the devel-opment of new theory, software programs andmethods to make the best use of these newdata, are just beginning (Knowles, 2006).

More broadly, the effect of these technolog-ical changes has been to produce a conver-gence of work of what one might call‘technologically enabled’ history. Traditionalhistorians and humanists in general—forexample, in the work of Franco Moretti

258


(2005)—also now work with electronic data-bases, learn new computer programs to ana-lyze the rapidly proliferating data, and explorenew forms of presentation of the results oftheir analysis. Quantitative historians had tolearn the skills necessary to prepare and pre-sent statistical results in print. Historians moregenerally are using visual images, audio andvideo in their presentations, not as ‘illustra-tion’ to enhance or supplement an analysis butas core evidence for analysis.21

Richard Steckel (2005) recently proposed anagenda for what he called ‘Big Social ScienceHistory’, which would extend the capacities ofquantitative history and translate some of itsmethods of work to non-quantitative projects.22

Andrew Abbott (2005) has also proposedsuch possibilities. As with the first generationof quantitative history, these large agendaswill require collaborative efforts to managethe enormously expanding data infrastructureand the myriad computer technologiesrequired to make best use of the expandingcorpus of digitized historical evidence, and todevelop appropriate theoretical approaches tosuch historical work.

ACKNOWLEDGEMENTS

* This essay has been improved considerablyby comments from colleagues, particularlyPeggy Adams, Erik Austin, Morgan Kousser,Jim Oberly, Lex Renda, Jack Reynolds,Steve Ruggles, Carole Shammas and DanScott Smith, The overall interpretation andremaining errors are all mine.

REFERENCES AND SELECTBIBLIOGRAPHY

Abbott, Andrew (2001) Time Matters: On History andMethod. Chicago: University of Chicago Press.

——— (2005) ‘Looking backward and looking for-ward, social science history at 2000’ in Harvey Graff,Leslie Page Moch and Philip McMichael with JuliaWoesthoff (eds), Looking Backward and LookingForward: Perpspectives on Social Science History.Madison: University of Wisconsin Press, pp. 69–72.

Abbott, Andrew and Tsay, A. (2000) ‘Sequence analysisand optimal matching methods in sociology’,Sociological Methods and Research, 29: 3–33.

Adams, Margaret (1995) ‘Punch card records:Precursors of electronic records’, American Archivist,58 (Spring): 182–201.

——— (forthcoming) ‘Analyzing archives and findingfacts: Use and users of digital data records’, ArchivalScience.

Alter, George (1988) Family and the Female LifeCourse. Madison: University of Wisconsin Press.

Alter, George and Gutmann, Myron (1999) ‘Castingspells: Database concepts for event-history analysis’,Historical Methods, 32 (4): 165–76.

Ambacher, Bruce I. (ed.) (2003) Thirty Years ofElectronic Records. Lanhan, MD: The ScarecrowPress.

Annual Report of the American Historical Association(1893).

Aydelotte, William, Bogue, Allan and Fogel, Robert (eds)(1972) The Dimensions of Quantitative Research inHistory. Princeton: Princeton University Press.

Benson, Lee (1957) ‘Research problems in Americanpolitical historiography’, in Mirra Komarovsky (ed.),Common Frontiers of the Social Sciences. Glencoe,IL: Free Press, pp. 113–83, 418–21.

——— (1961) The Concept of Jacksonian Democracy:New York as a Test Case Princeton: PrincetonUniversity Press.

——— (1984) ‘The mistransference fallacy in explana-tions of human behavior’, Historical Methods, 17 (3):118–31.

Bogue, Allan G. (1983) Clio and the Bitch Goddess:Quantification in American Political History. BeverlyHills: Sage Publications.

——— (1986) ‘Systematic revisionism and a genera-tion of ferment in American history’, Journal ofContemporary History, 21 (2): 135–62.

——— (1990) ‘The quest for numeracy: Data andmethods in American political history’, Journal ofInterdisciplinary History, 21 (1): 89–116.

Bourke, Paul, DeBats, Donald and Phelan, Thomas(2001) ‘Comparing individual-level voting returnswith aggregates: A historical appraisal of the Kingsolution’, Historical Methods, 34 (3): 127–34.

Bridenbaugh, Carl (1962) ‘The Great Mutation’, AHAPresidential Address. Available at: http://www.histo-rians.org/info/AHA_History/cbridenbaugh.htm.(Accessed on May 15 2006.)

Burnham, Walter Dean (1955) Presidential Ballots,1836–1892. Baltimore: Johns Hopkins UniversityPress.

——— (1970) Critical Elections and theMainsprings of American Politics. New York:W.W. Norton & Co.




Burton, Orville Vernon (2002) Computing in the SocialSciences and Humanities. Urbana: University ofIllinois Press.

Cameron, Sonja and Richardson, Sara (2005) UsingComputers in History. New York: PalgraveMacmillan.

Carter, Susan, Gartner, Scott, Haines, Michael R., Olmstead,Alan, Sutch, Richard and Wright, Gavin (2006) HistoricalStatistics of the United States: Millennial Edition. NewYork: Cambridge University Press.

Chambers,William Nisbet and Burnham,Walter Dean (eds)(1967) The American Party Systems: Stages of PoliticalDevelopment. New York: Oxford University Press.

Chopwhittle, Darcy and Taleglad, Lars Mooson (2001)‘Review of Philinda Blank’s When the Cows ComeHome: Barn Architecture and Changes in BovinePublic Space,’ Social Science History, 25 (5):609–14. (See note 19.)

Conrad, Alfred and Meyer, John (1958) ‘The economicsof slavery in the ante-bellum South’, Journal ofPolitical Economy, 66 (2): 95–130.

Curti, Merle, Curti, Margaret Wooster, Daniel, Robert,Livermore Jr., Shaw, and Van Hise, Joseph (1959) TheMaking of an American Community: A Case Study ofDemocracy in a Frontier County. Stanford: StanfordUniversity Press.

Darcy, R. and Rohrs, Richard C. (1995) A Guide toQuantitative History. Westport, CT: Praeger.

Dollar, Charles M. and Jensen, Richard J. (1971)Historian’s Guide to Statistics: Quantitative Analysisand Historical Research. New York: Holt, Rinehartand Winston, Inc.

Feinstein, Charles and Thomas, Mark (2002) MakingHistory Count: A Primer in Quantitative Methods forHistorians. New York: Cambridge University Press.

Fishbein, Meyer H. (ed.) (1973) The National Archivesand Statistical Research. Athens, OH: Ohio UniversityPress.

Fitch, Nancy (1986) ‘Statistical fantasies and historicalfacts: History in crisis and its methodological impli-cations’, Historical Methods, 18 (4): 239–54.

Floud, Roderick (1972) An Introduction to QuantitativeMethods for Historians. Princeton: PrincetonUniversity Press.

Floud, Roderick, Wachter, Kenneth and Gregory,Annabel (eds) (1990) Height, Health and History:Nutritional Status in the United Kingdom,1750–1980. New York: Cambridge University Press.

Fogel, Robert W. (1964) Railroads and AmericanEconomic Growth: Essays in Econometric History.Baltimore: Johns Hopkins University Press.

——— (1993) Nobel Prize in Economics AutobiographicalStatement. Available at: http://www.nobelprize.org/economics/laureates/1993/fogel-autobio.html.(Accessed on May 15, 2006.)

——— (2003) The Slavery Debates: A Retrospective.Baton Rouge: Louisiana State University Press.

——— (2004) The Escape from Hunger andPremature Death, 1700–2100: Europe, America,and the Third World. New York: CambridgeUniversity Press.

Fogel, Robert and Costa, Dora L. (1997) ‘A theory oftechnophysio-evolution, with some implications forforecasting population, health care costs, and pen-sion costs’, Demography, 34 (1): 49–66.

Fogel, Robert William and Elton, G.R. (1983) WhichRoad to the Past? New Haven: Yale University Press.

Fogel, Robert and Engerman, Stanley (1974) Time onthe Cross, Vols. 1 & 2. Boston: Little Brown.

Goldin, Claudia (1997) ‘Exploring the “Present Throughthe Past”: Career and family across the last century’,The American Economic Review, 87 (2): 396–9.

Graff, Harvey, Moch, Leslie Page and McMichael, Philipwith Woesthoff, Julia (eds) (2005) Looking Backwardand Looking Forward: Perpspectives on Social ScienceHistory. Madison: University of Wisconsin Press.

Greif, Avner (1997) ‘Cliometrics after 40 Years’, TheAmerican Economic Review, 87 (2): 400–3.

Griffin, Larry (1993) ‘Narrative, event-structure analysis,and causal interpretation in historical sociology’,American Journal of Sociology 98: 1094–133.

Griffin, Larry and Isaac, Larry W. (1992) ‘Recursiveregression and the historical use of “time” in time-series analysis of historical process’, HistoricalMethods, 25: 166–79.

Griffin, Larry J., Clark, P. and Sandberg, J. (1997)‘Narrative and event: Historical sociology and lynch-ing’, in W. Fitzhugh Brundage (ed.), Under Sentenceof Death: Lynching in the New South. Chapel Hill:University of North Carolina Press, pp. 24–47.

Gutmann, Myron and Alter, George (1993) ‘Familyreconstitution as event history analysis’, in DavidSven Reher and Roger S. Schofield (eds), Old andNew Methods in Historical Demography. Oxford:Oxford University Press, pp. 159–77.

Hacker, David and Fitch, Catherine (eds) (2003a)‘Building Historical Data Infrastructure: New Projectsof the Minnesota Population Center’, HistoricalMethods 36 (1).

——— (eds) (2003b) ‘Building Historical DataInfrastructure: New Projects of the MinnesotaPopulation Center’, Historical Methods 36 (2).

Harvey, Charles and Press, Jon (1996) Databases inHistorical Research: Theory, Methods andApplications. London: Macmillan Press Ltd.

Haskins, Loren and Jeffrey, Kirk (1990) UnderstandingQuantitative History. New York: McGraw-Hill.

Heckman, James J. (1997) ‘The value of quantitativeevidence on the effect of the past on the present’,The American Economic Review, 87 (2): 404–8.

260


Hudson, Pat (2000) History by Numbers: AnIntroduction to Quantitative Approaches. New York:Oxford University Press.

Isaac, Larry W. and Griffin, Larry (1989) ‘Ahistoricismin time-series analyses of historical process:Critique, redirection, and illustrations from USlabor history’, American Sociological Review,54: 873–90.

Jarausch, Konrad and Hardy, Kenneth (1991)Quantitative Methods for Historians: A Guide toResearch, Data, and Statistics. Chapel Hill: Universityof North Carolina Press.

King, Gary (1997) A Solution to the Ecological InferenceProblem: Reconstructing Individual Behavior fromAggregate Data. Princeton, NJ.: Princeton UniversityPress.

Kousser, J. Morgan (1973) ‘Ecological regression andthe analysis of past politics’, The Journal ofInterdisciplinary History, 4 (2): 237–62.

———(1974) The Shaping of Southern Politics:Suffrage Restriction and the Establishment of theOne-party South, 1880–1910. New Haven: YaleUniversity Press.

——— (1984) ‘The revivalism of narrative: A responseto recent criticisms of quantitative history’, SocialScience History, 8 (2): 133–49.

——— (1986) ‘Must historians regress?’ HistoricalMethods 19 (2): 62–81.

———(1989) ‘The state of social science history in thelate 1980s’, Historical Methods, 22: 13–20.

———(2001a) ‘Evaluating ecological inference: Anintroduction’, Historical Methods, 34 (3): 100.

——— (2001b) ‘Ecological inference from Goodman toKing’, Historical Methods, 34 (3): 101–26.

Knowles,Anne Kelly (2006) ‘GIS and history’, Paper pre-sented at the Annual Meeting of the AmericanHistorical Association, Philadelphia.

Lemisch, Jesse (1967) ‘The American Revolution seenfrom the bottom up’, in Barton Bernstein (ed.),Towards a New Past: Dissenting Essays in AmericanHistory. New York: Vintage Books, pp. 3–45.

Lewis, Jeffrey B. (2001) ‘Understanding King’s ecologi-cal inference model: A method-of-momentsapproach’, Historical Methods 34 (4): 170–88.

Lorwin, Val and Price, Jacob (1972) The Dimensions ofthe Past: Materials, Problems, and Opportunities forQuantitative Work in History. New Haven: YaleUniversity Press.

McCormick, Richard P. (1966) The Second AmericanParty System; Party Formation in the Jacksonian Era.Chapel Hill: University of North Carolina Press.

McDonald, Terrence (1986) The Parameters of UrbanFiscal Policy: Socioeconomic Change and PoliticalCulture in San Francisco, 1860–1906. Berkeley:University of California Press.

Meyer, John R. (1997) ‘Notes on cliometrics’ fortieth’,The American Economic Review, 87 (2): 409–11.

Monkkonen, Eric (2001) Murder in New York City.Berkeley: University of California.

Neustadt, Richard E. and May, Ernest R. (1986)Thinking in Time: The Uses of History for DecisionMakers. New York: Free Press.

Moretti, Franco (2005) Graphs, Maps, Trees: AbstractModels for a Literary History. New York: Verso.

North, Douglass C. (1997) ‘Cliometrics – 40 years later’,The American Economic Review, 87 (2): 412–14.

Palmquist, Bradley (2001) ‘Unlocking the aggregatedata past – Which keys fit?’ Historical Methods, 34(4): 159–69.

Redding, Kent and James, David R. (2001) ‘Estimatinglevels and modeling determinants of Black andWhite voter turnout in the South, 1880 to 1912’,Historical Methods 34 (4): 141–58.

Reiff, Janice (1991) Structuring the Past: The Use ofComputers in History. Washington, DC: AmericanHistorical Association.

Reynolds, John F. (1998) ‘Do historians count anymore’,Historical Methods, 31 (4): 141–8.

Rowney, Don Karl and Graham, James Q. Jr.(eds) (1969) Quantitative History: Selected Readingsin the Quantitative Analysis of Historical Data.Homewood, Ill.: Dorsey Press.

Shreibman, Susan, Siemens, Ray and Unsworth, John(2004) A Companion to Digital Humanities. Malden,MA: Blackwell Publishing.

Shammas, Carole (1990) The Pre-industrial Consumerin England and America. Oxford: Clarendon Press.

Shammas, Carole, Salmon, Marylynn and Dahlin, Michel(1987) Inheritance in America from Colonial Timesto the Present. New Brunswick, NJ: RutgersUniversity Press.

Shorter, Edward (1971) The Historian and theComputer: A Practical Guide. Englewood Cliffs, NJ:Prentiss Hall.

Silbey, Joel, Bogue, Allan and Flanigan, William (eds)The History of American Electoral Behavior.Princeton: Princeton University Press.

Smith, Dan Scott (1984) ‘A mean and random past: Theimplications of variance for history’, HistoricalMethods, 17: 14– 48.

——— (1992) ‘Context, time, history’, in Peter Karstenand John Modell (eds), Theory, Method, and Practicein Social and Cultural History. New York: New YorkUniversity Press, pp. 13–32.

Steckel, Richard (2005) ‘Presidential Address’, AnnualMeeting of the Social Science History Association,Portland, Oregon.

Steckel, Richard H. and Floud, Roderick (eds) (1997)Health and Welfare During Industrialization.Chicago, IL: University of Chicago Press.




Steckel, Richard and Rose, Jerome (eds) (2002) TheBackbone of History: Health and Nutrition in theWestern Hemisphere. New York: CambridgeUniversity Press.

Stone, Lawrence (1977) ‘History and the social sciencesin the twentieth century’, in C. Delzell (ed.), TheFuture of History. Nashville: Vanderbilt UniversityPress. pp. 3–42.

——— (1979) ‘The revival of narrative: Reflections ona new old history’, Past and Present, 89: 3–24.

Swierenga, Robert (ed.) (1970) Quantification inAmerican History: Theory and Research. New York:Atheneum.

Tolnay, Stewart E. and Beck, E.M. (1995) A Festival ofViolence: An Analysis of the Lynching of African-Americans in the American South, 1882–1930.Urbana: University of Illinois Press.

Turner, Frederick Jackson (1893) ‘The Significance of theFrontier in American History’, Annual Report of theAmerican Historical Association: 199–227.

Whaples, Robert (1991) ‘A quantitative history of theJournal of Economic History and the cliometric revo-lution’, Journal of Economic History, 51 (2): 289–301.

NOTES

1 See, for example, Benson (1957; 1961);Burnham (1970); Chambers and Burnham (1967);Richard P. McCormick (1966).

2 For information on the Cambridge Group, seetheir website, http://www-hpss.geog.cam.ac.uk.

3 See also Kousser (1989) and Reynolds (1998).4 See for example, Darcy and Rohrs (1995); Dollar

and Jensen (1971); Feinstein and Thomas (2002);Floud (1972); Haskins and Jeffrey (1990); Hudson(2000); Jarausch and Hardy (1991); Shorter (1971).

5 See, for example, Aydelotte et al. (1972); Lorwinand Price (1972); Rowney and Graham (1969); Silbeyet al. (1978); Swierenga (1970).

6 ICPSR, founded in 1962 as ICPR, changed itsname to the Inter-university Consortium for Politicaland Social Research (ICPSR) in 1975.

7 See, for example, Floud et al. (1990); Monkkonen(2001); Shammas et al. (1987); Shammas (1990);Steckel and Floud (1997); Steckel and Rose (2002).

8 The schedule in Table 14.1 is available on theIPUMS website at http://www.ipums.umn.edu/usa/voliii/form1950.html.

9 The United States maintains census schedules asconfidential records for 72 years. The 1890 Censusmanuscript schedules were destroyed by fire in 1921.

10 See the special issues of Historical Methods(Hacker and Fitch 2003a; 2003b) on ‘BuildingHistorical Data Infrastructure: New Projects of theMinnesota Population Center’ and the website ofIPUMS at www.ipums.org for details.

11 Nattional Archives and Records Administration,‘Inside the National Archives – Southeast Region,1825-1863 Slave Sale Documents’. Available at:http://www.archives.gov/southeast/exhibit/2.php.

Transcription of Slave Sale Document in Figure 14.3Know all men by these presents, That I, Albert G.

Ewing, of the county of Davidson and state ofTennessee have this day for and in consideration offive hundred dollars, to me in hand paid by JosephWoods and John Stacker, Trustees for SamuelVanleer, his wife and chldren, under the will ofBernard Vanleer, now recorded in the office of theDavidson county court, state of Tennessee, bargainedand sold unto said Trustees, a certain negro boyname George aged about seventeen years; whichsaid slave I warrant to be sound and healthy; and Ialso will warrant the right and title of said slave, untosaid Trustees, their heirs, executors, &c. &c. and thatsaid negro boy George is a slave for life.

Witness my hand and seal, this Sixth day ofNovember 1833.

A.G. Ewing

Frederick BradfordOrville ErvingNov. 6. 1833.

12 For other examples of slave sale documents,see the Slave Documents Collection from the EnochPratt Free Library, Baltimore, Maryland, available athttp://www.pratt.lib.md.us/exhibits/slavery/

13 The data set and code-book are available at:ht tp: / /www. icpsr.umich.edu/cocoon/ ICPSR-STUDY/07423.xml.

14 See, for example, Kousser (1973, 1974). Forrecent methodological developments in the field andtheir impact in history, see King (1997), and the arti-cles in the Summer and Fall 2001 (34 (3 & 4)) issuesof Historical Methods on the time period by: Kousser(2001a, 2001b); Bourke et al. (2001); Redding andJames (2001); Palmquist (2001); and Lewis (2001).

15 See Abbott (2001); Abbott and Tsay (2000);Alter and Gutmann (1999); Alter (1988), Gutmannand Alter (1993); Griffin (1993); Griffin and Isaac(1992); Isaac and Griffin (1989); Reher and Schofield(1993). On time series, see also McDonald (1986).On quantification and historical explanation, seeSmith (1984; 1992).

16 For the results of this research, see Fogel(2004); Fogel and Costa (1997).

17 For discussion of Benson’s change in positionand critiques of the change, see Bogue (1986; 1990)and Kousser (1986). See also Fogel and Elton (1983);Kousser (1984); Fitch (1984); and Fogel (2003).

18 For a hilarious parody of the issues involved,see the Winter 2001 issue of Social Science History.Outgoing editors Paula Baker and Elizabeth Fauepublished reviews by Darcy Chopwhittle and LarsMooson Taleglad of Philinda Blank’s (2001) When theCows Come Home: Barn Architecture and Changesin Bovine Public Space (2001).

262


(The reviewed book does not exist, though per-haps it might. Many people contributed to thereview; Paula Baker and Elizabeth Faue take respon-sibility for it.)

19 On anthropometric history, for example, seethe Summer 2004 Special Issue of Social ScienceHistory, Volume 28, no. 2, guest edited by JohnKomlos and Jorg Baten. For the impact of the IPUMSproject, see the bibliography of work listed on theIPUMS website, http://www.ipums.org. For recentevaluations of ‘social science history’ as a field, seeGraff et al. (2005).

20 For retrospectives on quantitative history, seeReynolds (1998). For retrospective analysis of ‘clio-metrics’, see the special section of the ‘Papers andProceedings of the Hundred and Fourth AnnualMeeting of the American Economic Association’ inThe American Economic Review (1997), 87 (2), on

‘Cliometrics After 40 Years’. Papers in this sectioninclude Goldin (1997); Greif (1997); Heckman(1997); Meyer (1997); and North (1997). See alsoWhaples (1991).

21 See, for example, Burton (2002); Cameron andRichardson (2005); Harvey and Press (1996); Reiff(1991); Shreibman et al. (2004).

22 Steckel listed the large data projects socialscience historians have produced in the last genera-tion and then added his own wish list: including aninventory all archeological sites; an inventory all arti-facts at these sites; a database on natural disastersand human history; and an international catalogueof films and photos. He called for extending the dig-itization of all extant manuscript censuses in thepast; a digitized and annotated collection of diaries;voting records at the precinct level; and probaterecords.

