Criminal sentencing in 19th-century Pennsylvania

Explorations in Economic History 46 (2009) 287–298

Contents lists available at ScienceDirect

Explorations in Economic History journal homepage: www.elsevier.com/locate/eeh

Criminal sentencing in 19th-century Pennsylvania Howard Bodenhorn John E. Walker Department of Economics, Clemson University and NBER, 222 Sirrine Hall, Clemson, SC 29634, USA

a r t i c l e

i n f o

Article history: Received 18 May 2008 Available online 2 April 2009

Keywords: Economics of crime Criminal sentencing Racial disparities Ethnic disparities

a b s t r a c t How law is interpreted and enforced at a particular historical moment reflects contemporary social concerns, attitudes and prejudices. This paper investigates the nature of criminal sentencing in 19th-century Pennsylvania. It finds that juries systematically departed from presumptive sentences based on extralegal factors, such age, sex, nativity and occupation. Older criminals and convicts with higher status preconviction occupations received longer sentences; women and ethnic minorities received shorter sentences. Unlike 20thcentury courts, 19th-century courts did not impose longer sentences on black criminals. Ó 2009 Elsevier Inc. All rights reserved.

1. Introduction How law is interpreted and enforced at a particular historical moment reflects contemporary social concerns. The republican ideal holds that all lawbreakers regardless of status or condition stand equal before the law. Such a notion is, of course, naive. As Mosher and Hagan (1994) note, much legal research consists of efforts to illuminate the gap between ideal and practice. An important component of that research agenda focuses on the extent to which extralegal factors, such as a defendant’s race, sex, ethnicity or other personal characteristics influence the dispensation of justice. Studies of 20th-century sentencing reveal persistent and large race and gender effects (Bushway and Piehl, 2001; Mustard, 2001; Sarnikar et al., 2006).1 African Americans receive longer and women shorter sentences, holding other observable factors constant. Sentencing disparities attributable to extralegal factors existed in the last third of the 20th century despite the adoption of strict sentencing guidelines by federal and several state courts. Guidelines were designed to constrain judicial discretion in sentencing in that crime severity and criminal history alone are determinative. Significant departures from the guidelines in either direction are condoned only when judges can point to compelling aggravating or mitigating circumstances (Shapiro, 1991; Seymour, 1992). That disparities emerge despite sentencing guidelines has reinforced some observers’ opinions that the criminal justice has historically served the interests of the rich and powerful (Friedman, 1973, p. 25, 1993; Walker, 1980, p. 104). Moreover, prisons are disproportionately populated with minorities, immigrants and the poor, which suggests that these groups suffer the worst injustices even after the adoption of sentencing guidelines. 19th-century courts faced fewer sentencing constraints than modern courts. Under modern Pennsylvania law, for example, a first-time offender convicted of rape faces a sentence between 46 and 66 months (Pennsylvania Commission on Sentencing, 2005). In the mid-19th century, Pennsylvania statutes provided for prison terms between 120 and 252 months, and juries were not strictly bound by even these broad limits. The narrowing of prescribed sentences reflects the diminishing latitude in sentencing afforded judges and juries over the past two centuries. The evidence presented below will show, however, that 19th-century justice was, while harsh by 20th-century standards, relatively even handed.

E-mail address: [email protected] The literature investigating the connection between personal characteristics and criminal sentencing is massive. Tractable reviews of racial disparities in criminal sentencing are provided by Wilbanks (1987), Weitzer (1996), Lauritsen and Sampson (1998) and Engen et al. (2002). 1

0014-4983/$ - see front matter Ó 2009 Elsevier Inc. All rights reserved. doi:10.1016/j.eeh.2009.03.001

288

H. Bodenhorn / Explorations in Economic History 46 (2009) 287–298

Information on more than 10,000 criminals sentenced to Pennsylvania’s two state prisons between the late 1820s and the 1870s shed light on three features of early 19th-century criminal procedure. First, blacks and immigrants were arrested and incarcerated at disproportionate rates. Blacks represented less than 2.8% of all Pennsylvanians in the antebellum era, yet they made up 16.3% of incoming prisoners. Moreover, blacks were incarcerated at significantly younger ages – 25.9 years – on average than whites (29.4 years). The foreign-born represented 14.8% of Pennsylvania’s aggregate population in 1860 and 26.8% of incoming prisoners in that year. Unlike blacks, however, the foreign-born tended to be older at incarceration – 31.4 years for the Irish and 33.2 years for Germans – than native-born whites (28.4 years). Second, discriminatory sentencing against blacks and ethnic minorities is not evident in the data. Juries were no more likely to depart from the presumptive sentence for African Americans than for whites and generally imposed shorter sentences for Irish and German immigrants relative to native-born whites. Finally, convicted felons from higher status occupations were more likely to receive sentences with upward departures from presumptive sentences than common laborers. This last clearly contradicts Friedman and Walker thesis that law is designed and used by upper classes to control lower classes, often to the detriment of the latter. 2. Nineteenth-century criminal procedure Historians describe 19th-century criminal courtrooms that little resemble the popular image. Court attendance was a popular form of lower class entertainment as people crowded into cramped spaces to jeer and cheer the procession of vagrants, drunks, prostitutes and felons brought before the court. Amidst the clamor of the crowd, judges conducted business at a pace that would astonish the modern observer. Defendants were continuously shuttled in and out. Pleading, testifying and cross examination were minimal. Rice (1996) reported that the average criminal jury trial lasted about 45 min.2 About half the time the jury remained in the box to deliberate and might hear several cases before deliberating each in turn. In 19th-century Pennsylvania most criminal prosecutions were private or citizen prosecutions (Steinberg, 1986, 1989; Rice, 1996). That is, citizens charged a defendant with a crime, such as assault or larceny, and then either represented themselves or hired a private prosecuting attorney. There were state’s attorneys, but Ireland (1995) contends that many victims preferred private to public prosecutors because the latter were typically underpaid, overworked and either young and inexperienced or old and incompetent. Bodenhamer (1985) and Rice (1996) are less critical of district attorneys. State prosecutors’ time was spent mostly drafting bills of indictment for grand jury hearings and they rarely acted as advocates. Victim preferences for private counsel arose because private prosecutors gave defendants ‘‘no quarter,” while it was common for a state’s attorney to inform a jury of flaws in the prosecution’s case (Rice, 1996, p. 467). As late as 1866 the Pennsylvania legislature amended its law allowing a victim to dismiss the state’s attorney in favor of a private prosecutor who, once recognized by the court, assumed most of the prosecutorial powers granted district attorneys (Pennsylvania General Assembly 1866, Chapter 72). Other features that distinguished 19th- from 20th-century criminal procedure were the central role of the jury and the absence of plea bargaining (Haller 1979). In 19th-century Philadelphia, accused larcenists were the defendants most likely to plead guilty, but only 10% of larcenists did so because there was little advantage to a guilty plea (Steinberg, 1989).3 With few pleas, jury trial was ubiquitous. 19th-century Americans considered the jury trial an essential bulwark against governmental power and oppression and reflected the still powerful faith in the common man (Note, 1964). Reliance on the jury also reflected a general distrust of judges appointed by faraway political elites who may or may not have had the same values as the local citizenry (Bouton, 2007). Not only was jury trial ubiquitous, it was accepted that 19th-century juries were finders of both fact and law. Acting as defense counsel for Reverend Jacob Gruber, who was charged with inciting a slave insurrection in Maryland in 1819, future Chief Justice of the United States Roger B. Taney, as Gruber’s attorney reminded the jury that they were ‘‘the judges of the law, as well as the fact” (quoted in Rice, 1996, p. 466). Judicial instruction was viewed as so much meddling with jury independence and was little tolerated in some jurisdictions. Gross (1998, p. 117) writes that in the antebellum era, ‘‘the realms of ‘law’ and ‘fact’ were far from distinct; although the 19th-century trend was towards greater power for the judge, vigorous popular efforts to limit judicial power, and strong customary traditions of juries deciding ‘law’ questions meant that the battle [for judicial determination of the law] was far from won.” Juries sometimes received instruction on the law, but instruction rarely settled the question. Jurors retained that right for themselves. As Rice (1996) noted, historians have interpreted the jury’s centrality and its power to determine law as a ‘‘license to impose locally-defined rough justice.” Because juries were not as ideologically committed to legal technicalities as attorneys and judges, it is possible that defendants were subject to more capricious applications of the law under a jury-dominated than a lawyer-dominated system. But the system placed meaningful limits on the jury, and jurors typically based their decisions on three considerations. First, was the evidence sufficient to support a conviction? Second, could the defendant contradict or explain away the evidence against him or her? And, third, what was the defendant’s character? It was only at this third stage that the criminal trial became a contest of character and statutory and common law ‘‘gave way to local, community-defined determinants of guilt or innocence” (Rice, 1996, p. 474). Although local rough justice conjures fears of unfair 2 Langbein (1978) reports similarly speedy trials at London’s Old Bailey. The typical trial lasted several minutes, not hours and certainly not days. In a two- to four-day sitting of the quarter session, the court would adjudicate 50–100 criminal cases. 3 Vogel (1999) finds that plea bargaining became increasingly common between 1830 and 1860 in Massachusetts. It remains unclear how widespread plea bargaining was elsewhere in the first half of the 19th century, though legal historians still consider it rare prior to the emergence of a powerful state attorney.

289

H. Bodenhorn / Explorations in Economic History 46 (2009) 287–298

severity, there is no reason that it could not work the other way. Though far removed from 19th-century Pennsylvania, late medieval and early modern English juries were notoriously lenient (Hanawalt, 1979). Whether 19th-century juries based punishments on extralegal factors and in which direction is then an empirical question. 3. Data The data come from ledgers kept by the wardens at the Pennsylvania’s Eastern State Penitentiary in Philadelphia and Western State Penitentiary in Pittsburgh. Extant records from Eastern State include the ‘‘Descriptive Registers” maintained between 1829 and 1857 and those from Western State include the ‘‘Convict Docket and Descriptive Registers,” from 1826 through 1876. These records include basic information about convicts, including their names, ages, nativities, occupations, the crimes for which they were incarcerated, sentence lengths, prior convictions, and counties of conviction. The two ‘‘Descriptive Registers” also include identifying information, such as complexion and race. Whites are recorded with complexion descriptions such as dark, ruddy or pale. African Americans were identified as black or mulatto (hereafter mixedrace). The registers also included eye and hair color, and brief descriptions of any unique marks, scars or deformities. Clerks at Eastern State also recorded each prisoner’s height and foot length, probably to facilitate the provision of clothing and shoes. ‘‘Convict Reception Registers,” kept between 1842 and 1869 represent the second set of prisoner records from the Eastern State Penitentiary. The Reception Registers include the same information on prisoners as the ‘‘Descriptive Registers.” Table 1 provides summary statistics of the principal variables used in the analysis of sentencing. Given the nature of the data, most variables are dichotomous (0,1) variables. The exceptions are age, sentence length, crime, and prior convictions. Age is recorded in years. The average age at intake was 28.8 years. The youngest person sent to the prison was 11 years old at intake; the oldest was 89 years. The largest percentage of prisoners, regardless of race or ethnicity came to the prison in their 20s. Nearly 40% of blacks were between 20 and 24 years of age, compared to 16% of Irish, 15% of Germans and 30% of nativeborn whites. The median ages at intake for members of these groups were 24, 28, 31 and 25 years, respectively. Sentence length, recorded in years, is transformed into its equivalent in months, which follows standard practice in the sentencing literature. In the 50 years of prison records used here, convicted felons were incarcerated for 67 different crimes, ranging from abortion to vagrancy. Table 2 provides information on the most common crimes, the statutorily recommended sentences under the major revisions to the criminal code adopted in 1794, 1829 and 1860. The fourth column reports the number of commitments for each crime. More than half of all prisoners (54.4%) were larcenists. Convicted burglars made up another 9.3% of all convicts, counterfeiters 4.3% and arsonists 2.3%. Violent criminals, including those convicted of rape (2.3%), manslaughter (2.3%), attempted murder (2.3%) and second-degree murder (2.8%), represented a sizable minority of prisoners. These eight crimes account for four–fifths of all incarcerations. The fifth column of Table 2 reports mean sentences by crime for these common crimes. Jury-imposed sentences varied in proportion to crime severity; that is, larceny was less severely punished than arson, which was less severely punished than murder. There was also notable variation in sentence length within crime classifications. Although the mean and median sentences for horse theft were 36.2 and 30 months, the 75th percentile was 48 months and the 25th percentile was 24 months. For larceny, the mean and median were 24.3 and 18 months and the interquartile range was 30 and 12 months. Unfortunately, the Pennsylvania prison data do not provide any information about the dollar value of goods stolen or other Table 1 Summary statistics for convicts sentenced to Pennsylvania penitentiaries, 1819–1876. Variable

Description

Mean

SD

Sentence length Prior convictions Age Female Black Mixed race Irish German Other immigrant Professional Clerical Sales Labor Operative Service Craftsman Proprietor Duncan Index

Number of months convict sentenced to serve Number of prior convictions Defendants age in years at intake ‘=1 if defendant had female given name ‘=1 if recorded as ‘‘black” ‘=1 if recorded as ‘‘mulatto” ‘=1 if born in Ireland ‘=1 if born in a German state ‘=1 if born outside US, Ireland and Germany ‘=1 if employed in professional occupation ‘=1 if employed in clerical position ‘=1 if employed as salesman ‘=1 if employed as common laborer ‘=1 if employed in semi-skilled manufacturing ‘=1 if employed in services ‘=1 if employed as skilled craftsman ‘=1 if self-employed proprietor Occupational score

33.19 0.25 28.82 0.04 0.09 0.07 0.10 0.07 0.07 0.01 0.02 0.01 0.35 0.18 0.07 0.26 0.20 16.28

28.53 0.68 10.34

13.23

Notes: Professionals are mostly doctors and lawyers; clerical are clerks and bookkeepers; services are housekeepers, stewards, washerwomen, etc.; craftsmen include blacksmiths, coopers, saddlers, etc.; proprietors are boarding house keepers, shopkeepers and merchants. Duncan Index scores involved classifying occupations according to categories reported in Reiss (1965). The categorization is consistent with IPUMS. Sources: Author’s calculations from information found in Eastern State Penitentiary (1829–1857, 1842–1869), Western State Penitentiary (1826–1876).

290

H. Bodenhorn / Explorations in Economic History 46 (2009) 287–298

Table 2 Representative crimes, presumptive sentences, commitments and mean actual sentences. Crime

Guidelines (1794) (months)

Guidelines (1829) (months)

Guidelines (1860) (months)

Commitments

Mean sentences (months)

Arson Assault Bigamy Burglary Counterfeiting Forgery Horse theft Larceny Manslaughter Murder (2d) Rape Receiving stolen goods Robbery

60–144 None 0–24 0–120 48–180 None 0–84 0–36 24–120 60–216 120–252 0–24

12–120 None 0–24 24–120 12–84 12–84 12–48 0–36 24–72 48–144 12–144 0–24

0–240 0–60a 0–24 0–120 12–84 0–60 12–48 0–36 0–144 0–144 0–180 0–24

272 216 97 1020 472 322 550 5961 247 315 258 207

52.1 31.2 16.1 50.3 35.7 35.1 36.2 24.3 41.1 93.5 66.7 25.4

0–120

12–84

0–120

322

46.6

Notes: 1794 presumptive sentences for bigamy, burglary, larceny, receiving stolen goods and robbery were enacted in 1790. There was no presumptive sentence for assault. Tyson (1827, p. 61) reports that the ‘‘punishment of [sic] assault and battery, being at common law, may be for any period.” Sources: Pennsylvania General Assembly, vol. II (1810, Chapter MDV); Pennsylvania General Assembly, vol. III (1810, Chapter MDCCLXVI); Pennsylvania General Assembly (1829, chapter 204), Tyson (1827), Wedgewood (1844), Kuntz (1988, p. 26), Eastern State Penitentiary (1829–1857, 1842–1869), Western State Penitentiary (1826–1876). a Maximum sentence of 60 months for assault with intent to rob; simple assault carried maximum 12-month sentence. The 1860 revision also included maximum fines, generally $500 or $1000, for several crimes.

indicators of crime severity. We might presume that a person convicted to 48 months for the theft of a horse stole a more valuable horse than one sentenced to 24 months, but this remains a matter of speculation. 4. Conceptual issues in criminal sentencing research Rationalist Anglo commentators since at least the time of John Locke have argued that the objective of the criminal justice system is to provide deterrence, incapacitation and proportional retribution.4 The alternative utilitarian approach to criminal sentencing emphasizes the extent of the harm, including all relevant externalities, and achieving efficient deterrence in a costeffective fashion (Waldfogel, 1993). Criminal sentencing in early America was an amalgam approach that acknowledged the philosophical justifications for proportionality in addition to public desires for rehabilitation and Christian charity. This amalgam approach, especially the Christian charity element, strongly influenced the Quaker Pennsylvanian approach, so that capital and corporal punishments gave way to fines and incarceration. By end of the first third of the 19th century, for example, the presumptive sentence for convicted counterfeiters was no longer the pillory followed by hanging as prescribed in a 1767 statute. After 1829, from one to seven years in the state penitentiary was believed to offer a more efficient mix of deterrence, incapacitation and rehabilitation.5 Under either rationalist or utilitarian approaches, the efficient price for an infraction was unrelated to extralegal factors, or personal characteristics such as sex, age, race or ethnicity. Becker’s (1968) canonical model of crime concludes that the optimal sentence is negatively related to both the marginal cost of imprisonment and the probability of imprisonment because the supply of crime is inversely related to the probability of detection, conviction and incarceration, but does not easily accommodate extralegal factors. Albonetti’s (1991, p. 249) statistical discrimination approach, on the other hand, incorporates the ‘‘past experience, stereotypes, prejudices, and highly particularized views” of judges, which may influence sentencing outcomes. The judge’s particularized views shape his or her beliefs about a criminal’s propensity to recidivate so that, in the face of uncertainty about the connection between sentence severity and recidivism or dangerousness, observable characteristics believed to be correlated with recidivism, such as race, ethnicity and sex, will influence sentence length. Although Albonetti’s statistical discrimination model is judge-based, it can be adapted to the 19th-century’s jury-based system. It is not unreasonable to assume that juries balanced their charges to apply the law equitably and to protect the public. If popular opinion accepted that a characteristic such as race or sex served as a predictor of recidivism or dangerousness, it is equally reasonable to believe that juries used existing stereotypes in determining sentences. Imperfect information and bounded rationality do not excuse discriminatory practices, of course, but understanding jury sentencing will shed light on the power and salience of racial and ethnic stereotypes to 19th-century Americans. Alternatively, if 19th-century juries accepted the primacy of the rule of law, statutory penalties will have had presumptive force and sentencing variance will be statistical noise that reflects the jury’s desire to tailor the sanction to the specific crime and not the characteristics of the criminal. 4 Locke wrote, ‘‘each transgression may be punished to that degree, and with so much Severity as will suffice to make it an ill bargain to the Offender, give him cause to repent, and terrife [sic] others from doing the like” (quoted in Hirsch, 1992, p. 81). 5 Pennsylvania General Assembly, Laws of the Commonwealth (1810) vol. I; Laws of the General Assembly (1829), Chapter 204.

H. Bodenhorn / Explorations in Economic History 46 (2009) 287–298

291

5. Empirical issues in criminal sentencing research In investigating the extent to which extralegal factors influenced jury sentencing in the 19th century, I follow the modern sentencing guidelines departure literature and estimate an empirical specification of the following general form:6

Sentence lengthi ¼ a þ b Presumptive sentencei þ cX i þ i One advantage of this formulation is that it explicitly incorporates the effects of recommended sentences. That is, the specification assumes that the observed sentence is driven by the statutory recommendation in addition to a set of personal characteristics or other factors captured in the vector X. By removing the variation due to the presumptive sentence, it is possible to isolate the jury discretion induced by extralegal factors and determine whether any group or groups were the victims of consistent departures from contemporary sentencing norms.7 A second empirical issue is that the data are positively skewed in levels. Linear regression techniques might then generate curvilinear errors terms, which will lead to inefficient standard errors and potentially inappropriate inference. The positive skew is easily dealt with by taking the natural log of the sentences. Estimating the equation with the left-hand side variable expressed in log terms has the added advantage of facilitating the interpretation of the coefficients. Estimated c’s are simply the percentage change in the dependent variable caused by a change in X. The advantage of logarithms rather than levels becomes clear if we think about comparing a six-month departure from an 18-month presumptive sentence and a six-month departure from a 48-month presumptive sentence. Clearly, the former represents a more significant proportionate departure than the latter, but estimating the equation in levels forces these two departures to have identical meanings, making interpretation of the estimated c’s less than obvious (Bushway and Piehl, 2001, p. 747). A third empirical issue involves developing an appropriate measure of the presumptive sentence. Table 2 reports summary statistics for 13 representative crimes. The first three columns report the statutory guidelines adopted in the 1794, 1829 and 1860 revisions to the criminal code; column 5 reports average sentences. A priori there is no correct choice of recommended sentence other than it should reflect an understanding of the guidelines and the institutional structure in which they are established and enforced. Pennsylvania statutes provided wide limits within which juries were given substantial discretion, yet mean sentences approximate the midpoint of the sentencing ranges. Regressions are estimated using the midpoint as the presumptive sentence. A fourth empirical issue involves how best to model the sentencing decision, because how we conceptualize jury processes will determine the preferred estimation strategy. Like Albonetti (1997) and Bushway and Piehl (2001), I assume that the sentencing decision occurs at a single stage. That is, upon receiving any instruction about the law from the judge, including the legally prescribed sentence, the jury simultaneously determined guilt and sentence length, a choice that naturally included a sentence of length zero. Sentencing constraints mean that sentences are censored between lower and upper threshold values. Two censoring points emerge in the 19th-century data. The data are left-censored because criminals convicted and sentenced to serve more than 12 months were to be sent to one of the state prisons. Criminals sentenced to less than 12 months were to serve their sentences in a city or county jail. For as yet unknown reasons, 332 criminals whose records appear in the prison records served less than 12 months in the prison and the shortest term observed in the final sample is three months. These observations, though anomalous, are retained because they provide information on contemporary sentencing practices. The data are also right censored because the longest observed term is 29 years (348 months) and there are few sentences of more than 20 years. A useful way to think about censoring observed in criminal sentences is to assume there is a latent variable y* (sentence length) that is only observed when it falls between the lower and upper bounds. The Tobit model was designed to solve the censoring problem by including all available information into a likelihood function under the assumption that y* would be normally distributed if it was fully observed. Thus, we can write the Tobit likelihood function for presumptive 19th-century sentences as

8  ! F N ð0jli ; rÞ > < 0 if yi < 12  yi ¼ yi if 12 6 yi < 348 ! fN ðyi jli ; rÞ > : ! 1  F N ð348jli ; rÞ 348 if yi P 348 which is, in effect, the product of two probits and a linear regression model, a feature exploited here. Because sentences less than 1-year are not fully observed and sentences in excess of 348 months (death penalties) are not observed at all, the effects of extralegal factors are estimated with OLS regressions on the uncensored values. Least squares estimates are, of course, inefficient and potentially biased toward zero compared to maximum likelihood estimates (Green, 1990, pp. 730–731), 6 For an introduction to the guidelines departure literature see Griswold (1987), Albonetti (1997), Reitz (1998), Engen and Gainey (2000, 2001), Mustard (2001) and Bushway and Piehl (2001). 7 Bushway and Piehl (2001) contend that studies of modern sentencing guidelines should impose the restriction that b = 1 so as to isolate judicial decisions based on defendant characteristics from judicial responses to the guidelines themselves. Imposing the restriction is probably more sensible in a judge-based 20th-century context where judges are given narrow sentencing latitude than in a jury-based system in which juries were afforded broad sentencing powers. Estimating b also affords an opportunity to understand how juries responded to presumptive sentences. The estimated c’s are not qualitatively different in specifications where b is estimated or restricted.

292

H. Bodenhorn / Explorations in Economic History 46 (2009) 287–298

but the bias works against finding significant coefficients on the personal characteristics. The extent to which the estimates are inconsistent is unknowable. The regressions are first reported controlling only for criminal history and the personal characteristics of the criminals. Dummy variables are then added for 18 different crimes for which both presumptive and reliable midpoint sentences could be determined.8 Offense-level controls mitigate one source of potential bias. It may be that individuals convicted of one crime consistently receive a longer sentence than individuals convicted of a second crime even if the presumptive statutory midpoints are the same. If members of a particular group are disproportionately convicted of the crime receiving the longer sentences, and the crime category is not controlled for, it will appear that members of that group receive longer sentences (Mustard, 2001). The third specification adds controls for decades because public concern with certain crimes may have changed over time. County fixed effects included in the fourth, and preferred, specifications reveal ‘‘within-county” variation. Thus, the regressions estimate the extent to which an individual with a given criminal history, convicted in a given court of a given crime received a sentence that differed from that received by an otherwise comparable individual based on the race, ethnicity, sex or occupation of the first individual. This is the definition of a sentencing disparity, which will be identified by economically meaningful and statistically significant coefficients on observable personal characteristics. Finally, any sample of people convicted and sentenced to the penitentiary is selective (Winship and Mare, 1992). Of those who commit crimes, only a fraction are arrested. Of those arrested, only a fraction stand trial. Of those who stand trial, only a fraction are convicted. And, of those convicted, only a fraction will be sent to the penitentiary. Empirically, the sample is restricted on the dependent variable – a sentence length greater than zero – which means that it is effectively selected on the regression residual. That is, for any values of the independent variables, observations with sufficiently large absolute errors are excluded from the sample. The empirical procedure discussed previously is an imperfect solution to the selection problem, as are most selection models. In the end, Winship and Mare (1992, p. 333) note that, even when sample selection is empirically modeled, ‘‘any study of criminal punishment needs to account for the process through which conviction takes place.” So long as the process is understood, studies of criminal sentencing can inform us of contemporary practice and social concerns. 6. The determinants of criminal sentences Table 1 provides summary statistics on variables included in the regression analysis. The average sentence length was 33.2 months and ranged, in the final sample, from 3 to 348 months. The average prisoner had 0.25 prior convictions, but nearly 84% had not been previously incarcerated. Because there were a small number of convicts with multiple convictions – the maximum was nine – the regressions include a dummy variable that equals one if the prisoner had one or more prior convictions.9 Table 2 previously reported summary statistics on specific crimes. The basic regression specification includes the two basic criminal indicators – crime type (a series of dummy variables for 18 crime categories) and/or whether the criminal had previously been convicted of a serious felony. If juries respected the rule of law, crime type and criminal history should emerge as the principal observable determinants of sentence length. To account for any systematic disparities that may have arisen in sentencing, the regressions also include personal characteristics of the offenders. Age and its square is included to capture any age-related disparities. A dummy variable for females is included, as are dummy variables for black and mixed-race prisoners. Bodenhorn (2002, 2006) and Bodenhorn and Ruebeck (2007) show that 19th-century Americans treated mixed-race individuals preferentially relative to blacks so African American complexion is controlled for in all regressions. Ethnicity is captured with three dummy variables – Irish, German and all other immigrants, the last of which is comprised mostly of people from Great Britain, France and Scandinavia. If antiimmigrant, especially anti-Irish, prejudice was as salient as reported by Clark (1973) and Ignatiev (1995), such prejudice may have manifested itself in unwarranted sentencing disparities. Finally, because some modern critics of the criminal justice system interpret the disproportionate incarceration rate of lower class, low-income groups as evidence of either systemic bias against the poor or of the ability of the wealthy to purchase preferential justice, the regressions control for social class (Walker, 1980). Occupations are grouped into one of nine separate categories based on Otis Dudley Duncan’s classification scheme (Reiss, 1965). Because some occupational groups, such as professionals, are underrepresented in the data which creates some cells with few observations, the value of the Duncan Index or its natural logarithm are included in alternative specifications. Duncan created index scores for hundreds of occupations. Educated professionals, such as doctors, lawyers and college professors populate the highest ranks, followed by managers and small-business proprietors. Common laborers, menial personal service providers (house servants or washerwomen) and unskilled farm workers fall at the bottom of the index. Skilled craftsmen, clerical and sales workers form the intermediate groups. Index values take values between 1 (common laborer in a tobacco factory) and 96 (physician). The average among Pennsylvania convicts was 16, or the equivalent of a sailor or semi-skilled operative. Although people from all socioeconomic classes appeared in the records, prisoners were overwhelmingly from the lower socioeconomic orders, mostly common laborers. 8 Including only 18 of the 67 crimes reduces the sample from 10,952 to 10,021. Some observations were dropped because statutory guidelines for the crimes could not be located, because there were only a handful of incarcerations for the crime in question (abortion, incest, obscenity), or both. 9 Future research will investigate the determinants of repeat offending. Given the high rates of recidivism among a small proportion of criminals, it is a topic worthy of a study of its own.

293

H. Bodenhorn / Explorations in Economic History 46 (2009) 287–298

6.1. Determinants of sentences – full sample Table 3 reports the basic regressions results for the final sample of 10,021 prisoners committed to the Eastern or Western State penitentiaries between 1826 and 1876. Column 1 is a baseline regression that includes the natural logarithm of the presumptive sentence, whether the convict was previously incarcerated in the prison in question, the prisoner’s age and its square, the prisoner’s race or ethnicity, and his or her occupation. Columns 2 and 3 sequentially include crime-category and time controls, which do not notably change the coefficients of interest. Columns 4 through 6 report preferred specifications with the full set of crime, time and place controls. Once the full set of controls is included, a number of notable results become apparent. First, 19th-century juries may have dispensed local rough justice, but they were not ‘‘hanging juries.” That is, actual sentences were consistently about 7% shorter than presumptive sentences, though the estimated coefficients are barely statistically significant at conventional levels. Second, controlling for the full set of personal characteristics, convicts with one or more prior convictions received sentences 23% (= e0.211) longer than a first-time offender. Although the result is consistent with public concerns with recidivism, it is not clear how judges and juries were made aware of the accused’s prior incarcerations. I observe prior incarcerations because the prison clerks noted when a prisoner returned, but 19th-century practice did not yet include the systematic provision of pre-sentencing information concerning the prisoner’s criminal history (Epstein, 1981, p. 54). Prosecutors, witnesses or the defendant himself may have revealed something about previous criminal history that influenced sentencing decisions. However the information was revealed, punitiveness increased markedly with recidivism, a practice that carried over from late-18th and early 19th-century sentencing practices (Epstein, 1981). The coefficients on age and age squared reveal inverted U-shaped pattern of sentences. Unreported alternative specifications using dummy variables for 5-year age groups generate a comparable but less pronounced inverted U-shaped pattern of sentences. Teenagers received significantly shorter sentences than convicts in their mid-30s, as did men over 64 years. The pattern of shorter sentences for younger offenders differs from late-19th-century practice, when grave concerns were expressed over youthful offending (Moehling and Piehl, 2007). Shorter sentences for youth likely reflect society’s hopes for youthful redemption. Although the age-sentencing profile for all groups – native-born whites, African Americans, and immigrants – followed a common inverted U-shaped pattern, African Americans were notably younger at intake than native-born and immigrant

Table 3 Determinants of sentencing outcomes – full sample.

ln(presumptive sentence) Multiple priors Age Age squared Female Black Mixed race Irish German Other immigrant Professional Proprietor Sales Service Operative Craftsman Clerical Farmer Duncan Index/100 ln(Duncan Index) Unknown occupation Constant Observations Adj R2 Crime controls Decade controls County fixed effects

(1)

(2)

(3)

(4)

(5)

(6)

0.488 (0.012)*** 0.163 (0.016)*** 0.012 (0.003)*** 0.0001 (0.000)*** 0.195 (0.032)*** 0.021 (0.022) 0.010 (0.025) 0.015 (0.022) 0.091 (0.024)*** 0.034 (0.025) 0.150 (0.066)** 0.070 (0.044)* 0.111 (0.068) 0.005 (0.027)* 0.016 (0.019) 0.005 (0.016) 0.149 (0.052)*** 0.037 (0.030)

0.101 (0.027)*** 0.252 (0.016)*** 0.013 (0.003)*** 0.0002 (0.000)*** 0.126 (0.031)*** 0.043 (0.021)** 0.042 (0.023)* 0.056 (0.020)*** 0.057 (0.022)*** 0.019 (0.023) 0.208 (0.061)*** 0.130 (0.041)*** 0.159 (0.064)** 0.042 (0.025)* 0.002 (0.017) 0.034 (0.015)** 0.129 (0.048)*** 0.038 (0.028)

0.080 (0.040)** 0.249 (0.016)*** 0.011 (0.003)*** 0.0001 (0.000)*** 0.119 (0.030)*** 0.026 (0.021) 0.036 (0.023) 0.047 (0.020)** 0.040 (0.022)* 0.017 (0.023) 0.209 (0.062)*** 0.129 (0.04)*** 0.171 (0.065)*** 0.038 (0.025) 0.002 (0.017) 0.031 (0.015)** 0.138 (0.048)*** 0.031 (0.029)

0.070 (0.039)* 0.212 (0.016)*** 0.014 (0.003)*** 0.0002 (0.000)*** 0.147 (0.030)*** 0.002 (0.021) 0.016 (0.023) 0.069 (0.020)*** 0.049 (0.022)** 0.036 (0.022) 0.169 (0.059)*** 0.079 (0.041)** 0.129 (0.068)* 0.020 (0.024) 0.015 (0.017) 0.016 (0.015) 0.087 (0.047)* 0.029 (0.029)

0.070 (0.039)* 0.214 (0.016)*** 0.013 (0.003)*** 0.0002 (0.000)*** 0.136 (0.028)*** 0.006 (0.020) 0.017 (0.022) 0.066 (0.020)*** 0.050 (0.022)** 0.038 (0.023)*

0.070 (0.039)* 0.213 (0.016)*** 0.013 (0.003)*** 0.0002 (0.000)*** 0.140 (0.028)*** 0.004 (0.020) 0.015 (0.022) 0.066 (0.020)*** 0.051 (0.022)** 0.036 (0.023)*

0.221 (0.048)*** 0.044 (0.033) 1.334 (0.065)*** 10,021 0.20 No No No

0.078 (0.034)** 4.619 (0.133)*** 10,021 0.32 Yes No No

0.054 (0.034) 4.410 (0.205)*** 10,021 0.33 Yes Yes No

0.023 (0.035) 4.097 (0.212)*** 10,018 0.36 Yes Yes Yes

0.046 (0.034) 4.065 (0.216)*** 9944 0.36 Yes Yes Yes

0.036 (0.010)*** 0.104 (0.042)** 4.007 (0.217)*** 9944 0.35 Yes Yes Yes

Notes: Dependent variable = ln(sentence). Robust standard errors reported in parentheses. ***p < 0.01; **p < 0.05; *p < 0.10. Crime categories include: First degree murder, arson, rape, counterfeiting, manslaughter, adultery, robbery, sodomy, larceny, bigamy, receiving stolen goods, forgery/fraud/embezzlement, rioting, kidnaping, perjury, and breaking and entering. Second degree murder is the excluded category. Decades include 1830s through 1870s. The 1820s are the excluded category. The data include criminal incarcerations from 86 counties. The sample sizes are smaller in columns (5) and (6) because some 19thcentury occupations have no obvious 20th-century analog as reported in the appendices of Reiss (1965). Sources: Author’s calculations from data reported in Eastern State Penitentiary (1829–1857, 1842–1869) and Western State Penitentiary (1826–1876).

294

H. Bodenhorn / Explorations in Economic History 46 (2009) 287–298

whites. The average African American inmate arrived just before his or her 26th birthday. Nearly 10% arrived between ages 15 and 19; 40% arrived between ages 20 and 24. The average age of native-born white inmates at arrival was 28.4 years and only 3% were between 15 and 19 years old at intake. Irish (31.4 years) and German (33.2 years) immigrants were substantially older than native-born whites. Less than 1% of Irish inmates and only 4.5% of Germans arrived in their late teens.10 It is difficult to square this finding with the popular image of young, Irish toughs fighting among themselves for control of 19th-century urban streets and, in the process, terrorizing law-abiding citizens. Of course, law-abiding citizens may have feared retribution from members of the accused individual’s gang and foregone private prosecution. Younger age at intake for African Americans is consistent with the black experience before and after the period under consideration here. Blacks were consistently overrepresented in the Pennsylvania prison population throughout the whole of the 19th century (Epstein, 1981; Naylor, 1979). Du Bois (1899/1967) and Lane (1986) argue that this overrepresentation followed from poor employment opportunities for blacks, which fostered a professional criminal class among young African Americans. African Americans accounted for 16% of incarcerations between 1829 and 1876, but less than 3% of Pennsylvania’s population and 8% or less of Philadelphia’s population in any census year between 1830 and 1870. Given this racial overrepresentation in the prison population, the absence of any association between race and sentence length is surprising. Stated simply, blacks served no longer for a given crime, holding other factors constant, than native-born whites. Epstein (1981, p. 112), too, failed to uncover any association between race and sentence length for the period between 1795 and 1829. Although she contended that ‘‘other sociological factors such as socioeconomic class and sex cut through racial distinctions,” contemporaries concerned themselves with black criminality. An article appearing in De Bow’s Review (Anonymous, 1853, pp. 593) noted how the liberal treatment of free blacks in the North was repaid with crime. But in an address to the Colonization Society, Henry Clay (1829) contended that excess black criminality did not follow from any ‘‘inherent depravity in their natural constitution, but from their unfortunate situation [poverty].” Fears surrounding black crime and the lack of consensus for its root cause may explain the combination of high incarceration rates and standard sentences. In taking black criminals off the streets contemporaries followed an incapacitation strategy, but were constrained by either popular notions of fair play or limited prison space.11 Higher incarceration rates and longer sentences for blacks may have overwhelmed always crowded, sometimes overcrowded, prisons. If the treatment of blacks was rooted in contemporary fears, the lenient treatment of female criminals had deep historical roots. For centuries, male jurors have not viewed women’s crime as seriously as men’s (Hanawalt, 1979). Even when it was deemed serious enough to merit conviction, women received special consideration at sentencing. This attitude, which dates to at least the 14th century, persisted through the 19th century and continues even up to today. Modern jurors acquit women more often than men. When they convict, women receive notably shorter sentences (Mustard, 2001). Where the average black prisoner received a sentence comparable to that received by the average native-born white and women received shorter prison terms than men, immigrants received somewhat shorter sentences than native-born whites. Holding all else constant, Irish immigrants received sentences about 7% shorter than native-born whites, Germans about 5% shorter and other immigrants about 3.5% shorter sentences. Lesser sentences for immigrants is inconsistent with historical accounts of the immigrants’ plight (Clark, 1973; Ignatiev, 1995). Lane (1986) remarked that no immigrant group faced as much hardship as the Irish, especially those who fled the famine in the 1840s. No city or town had enough jobs to accommodate the influx of immigrants, and charity institutions were overwhelmed. Misery and desperation drove many Irish into crime. ‘‘The problem of law and order,” wrote Lane (1986, p. 140), ‘‘was specifically an Irish problem.” They came to be known as the Wild Irish because their songs were ‘‘too loud, their liquor too strong, and their politics too Republican” (Epstein, 1981, p. 100). Despite native white concerns with both black and Irish criminality, the Irish and black experiences were completely different. This may be because Irish criminality was less threatening than black criminality. The murder indictment rate of individuals with Irish surnames, at about 4.7 per 100,000 in the 1860s and declining, was not quite twice the overall average murder indictment rate of 2.9%. In comparison, the black murder indictment rate exceeded 8.5% and was rising (Lane, 1986). Despite their reputation for brawling, Irish criminals were less fearsome than blacks, which may explain some of the differential treatment. An alternative explanation may be that Irish crime was less serious than nonimmigrant crime. Larceny, for example, includes a wide range of offenses, from stealing loaves of bread or a dollar’s worth of bacon to 100s, even thousands, of dollars. Assault spanned the range from a drunken brawl to acts just short of murder. Without details about the seriousness of the crime, we can only group similar crimes together and speculate that the shorter sentences imposed on immigrants implies that they committed less serious crimes. For the decades after the 1870s, Naylor (1979, p. 88) concludes that ‘‘vast numbers” of arrested Irishmen were latter freed without charge, which suggests less serious offenses. But preferential treatment remains a possible explanation, as does the possibility that native white jurors viewed Irish-on-Irish crime as less threatening to the social order.

10 Hazard ratios estimated from a Cox proportional hazards model (detailed results available on request) support the inferences drawn from the raw data. Blacks and mixed-race individuals were at greater hazard (earlier age at arrival) for imprisonment than native-born whites. Irish, Germans and other immigrants were at significantly lower hazard of incarceration, or were received at the prison at substantially older ages than native-born whites. This differs from the pattern of youthful immigrant offending reported by Moehling and Piehl (2007) for the late-19th and early 20th centuries. 11 Lane (1997) argues that fair play was the principal factor. He writes that white jurors were ‘‘surprisingly fair” and ‘‘reasonably evenhanded” in Boston, New York and Philadelphia courts. Naylor (1979) concludes that limited prison space placed binding constraints on black incarceration.

H. Bodenhorn / Explorations in Economic History 46 (2009) 287–298

295

Coefficients on the occupation categories in Column 4 reveal that individuals in the four white-collar occupations (professionals, proprietors, sales and clerical) received longer sentences than farmers, craftsmen, factory operatives, and laborers. Professionals, for instance, received 17% longer prison terms than laborers. Proprietors and clerical workers each received sentences about 8% longer than common laborers. Columns 5 and 6 report results based on two alternative specifications of occupational status. The coefficient on the value of the Duncan Index reveal that a 1% increase in the value of the index, evaluated at the mean, increased sentence length by about 3%. The coefficient on the natural logarithm of the Duncan Index value yields a comparable estimate of 3.7%. The meaning of these departures is further explored in the next section. How are we to summarize and interpret the regression results found in Table 3? Comparisons to 20th century outcomes may help place the 19th century results in context. In both eras, a convict’s criminal history had a notable effect on sentence length. Like the 11% average upward departure for 19th-century recidivists, Bushway and Piehl (2001) find that a prior criminal history significantly increases 20th-century departures from sentencing guidelines. A defendant’s age influenced the jury’s decision in the 19th century; it has some influence in the 20th, but not consistently (Bushway and Piehl, 2001; Mustard, 2001). Six prominent modern sentencing studies all find significant race and gender effects.12 Blacks consistently receive upward, and women downward, departures from 20th-century guidelines. In the 19th century, African Americans were not consistently sentenced to longer prison terms.13 As in modern America, however, women received notable downward departures. And as with race, the ethnicity effect differs markedly across the centuries. Three of the six modern studies find that Hispanics are more likely than whites to receive upward departures (Albonetti, 1997; Engen and Gainey, 2000; Mustard, 2001) and two find upward departures for other minority ethnic groups (Engen and Gainey, 2000; Mustard, 2001). In 19th-century Pennsylvania immigrants received shorter sentences, all else equal. Mustard (2001) identifies a negative relationship between sentence departure and income. The 19th-century results reveal a positive correlation between occupational status and upward departures from presumptive sentences. The administration of criminal justice in the 19th century differed from that of the 20th century in several dimensions. 6.2. Sentencing across occupational groups and crime categories Table 4 parses the data by crime type to reveal any subtleties in sentencing not apparent when considering all crimes together. The table reports the same sets of regressions as the fourth column of Table 3, except Column (1) includes only those convicts imprisoned for violent crimes – assault, murder, manslaughter, rape and robbery. Column (2) includes only those convicted of property offenses – burglary, counterfeiting, forgery, horse theft and larceny.14 Columns (3) and (4) further divide property crimes into white collar crimes – counterfeiting and forgery – and theft-related crimes – burglary, horse theft and larceny. In cases involving violent crime, criminal history, age and gender significantly influenced the pattern of departures from presumptive sentences. No other personal characteristic had a notable effect. The small and insignificant coefficients on personal characteristics suggests that juries took their charge to protect the public from dangerous felons seriously and did not systematically treat defendants of any particular race, ethnicity or social class differently, though contemporary juries obviously believed that violent women presented lesser threats to public order than men. Steinberg (1989) argues that many female violent offenders were abused spouses who finally fought back and were, therefore, viewed as less murderous. Middle-age violent felons were considered more deserving of upward disparities than younger felons. Extant records provide few insights into the juries’ thinking, but the harsher treatment of mid-life violent felons may reflect a concern with irredeemable, professional criminals. It is in sentencing of property offenders that personal characteristics influenced how juries departed from presumptive sentences. Criminal history remained the principal determinant of sentence length for property offenders. Recidivists were sentenced to about 7.1 months more than first-time offenders. Women received shorter sentences than men. Race had no notable effect on sentence length, but Irish and German immigrants received somewhat shorter sentences. Other immigrants were also treated leniently but not to the same extent. Once property crimes are divided into white-collar crime and theft, further differences in sentencing patterns emerge. White-collar criminals received modest upward departures from sentencing norms; thieves received sizeable downward departures. Jurors imposed sentences nearly 45% shorter on thieves than an adherence to the presumptive guidelines would predict. That is, jurors imposed sentences about half as long, on average, as the midpoint legislative recommendation. Crime historians note long histories of jury leniency for theft (Hanawalt, 1979). Whereas English juries had good reason for leniency – guilty verdicts resulted in the death penalty, Quaker Pennsylvania’s abolishment of sanguinary punishment except for first-degree murder imposed no death-penalty burden on 19th-century jurors (see Table 1). Yet jurors still found legis12 The six guidelines departures papers are Albonetti (1991), which studies sentencing in the District of Columbia in 1974; Albonetti (1997), US federal drug convictions in 1991 and 1992; Kramer and Ulmer (1999), which uses Pennsylvania sentencing in the late 1980s; Engen and Gainey (2000) use sentences in Washington between 1989 and 1992; Bushway and Piehl (2001) study Maryland sentences between 1987 and 1995; and Mustard (2001), which studies US federal sentencing in the early 1990s. The modern race, gender and ethnicity results are robust across jurisdictions. 13 It is possible that race was less meaningful in Pennsylvania than in the southern states. Preliminary results from the prison records of antebellum Maryland and Tennessee, however, also failed to find significant race effects. The Maryland and Tennessee results are not reported here because the samples are much smaller (i.e., 1000 in Maryland and 1150 in Tennessee versus 10,000 in Pennsylvania) and do not offer as complete a set of individual-level controls. 14 The third category of morals offenses, such as adultery, bigamy and sodomy among others, are not included because there were fewer than 500 such incarcerations over the period.

296

H. Bodenhorn / Explorations in Economic History 46 (2009) 287–298

Table 4 Determinants of sentencing outcomes – violent and property crime.

ln(presumptive sentence) Multiple priors Age Age squared Female Black Mixed race Irish German Other immigrant Professional Proprietor Sales Service Operative Craftsman Clerical Farmer Unknown occupation Constant Observations Adj R2 Crime Controls Decade Controls County fixed effects

(1) Violent crimes

(2) All Property crimes

(3) White collar crimes

(4) Theft

0.107 (0.062)* 0.203 (0.069)*** 0.031 (0.008)*** 0.0004 (0.000)*** 0.199 (0.116)* 0.031 (0.073) 0.009 (0.068) 0.022 (0.047) 0.099 (0.089) 0.010 (0.068) 0.256 (0.264) 0.022 (0.151) 0.057 (0.276) 0.060 (0.075) 0.012 (0.050) 0.019 (0.047) 0.120 (0.148) 0.111 (0.078) 0.221 (0.127) 3.841 (0.388)*** 1375 0.40 Yes Yes Yes

0.047 (0.052) 0.208 (0.016)*** 0.011 (0.003)*** 0.0001 (0.000)*** 0.131 (0.029)*** 0.001 (0.021) 0.015 (0.024) 0.086 (0.022)*** 0.061 (0.022)*** 0.030 (0.024) 0.196 (0.062)*** 0.110 (0.043)** 0.156 (0.070)** 0.015 (0.025) 0.010 (0.018) 0.020 (0.016) 0.102 (0.050)** 0.075 (0.032)** 0.052 (0.036) 2.887 (0.273)*** 8,344 0.28 Yes Yes Yes

0.119 (0.207) 0.193 (0.071)*** 0.030 (0.012)** 0.0003 (0.000)* 0.453 (0.220)** 0.571 (0.284)** 0.129 (0.259) 0.302 (0.101)*** 0.299 (0.123)** 0.035 (0.109) 0.315 (0.134)** 0.111 (0.101) 0.120 (0.200) 0.146 (0.251) 0.040 (0.078) 0.051 (0.073) 0.158 (0.121) 0.226 (0.111)** 0.201 (0.152) 2.453 (0.962)** 791 .26 Yes Yes Yes

0.443 (0.118)*** 0.205 (0.018)*** 0.006 (0.003)* 0.0001 (0.000)** 0.142 (0.029)*** 0.022 (0.022) 0.009 (0.025) 0.057 (0.024)** 0.033 (0.023) 0.020 (0.026) 0.090 (0.075) 0.067 (0.053) 0.258 (0.082)*** 0.032 (0.026) 0.019 (0.019) 0.022 (0.017) 0.100 (0.062) 0.025 (0.036) 0.066 (0.037)* 3.979 (0.422)*** 6,376 0.31 Yes Yes Yes

Notes: Dependent variable = ln(sentence). Robust standard errors reported in parentheses. ***p < 0.01; ***p < 0.05; *p < 0.10. Decades include 1830s through 1870s. The 1820s are the excluded category. The data include criminal incarcerations from 86 counties. Sources: Author’s calculations from data reported in Eastern State Penitentiary (1829–1857, 1842–1869) and Western State Penitentiary (1826–1876).

lative prescriptions excessively harsh and responded with light punishments. Contemporary court observers wrung their hands over such practices because deterrence was best achieved through the sure and consistent application of harsh penalties. Jurors, as previously noted, considered it their prerogative to mold the punishment to the crime and did so in cases involving theft. It is possible that jurors tempered the law’s severity when the evidence demanded conviction but the circumstances of the criminal or the crime cried out for mercy. The data do not afford the opportunity to observe the magnitude or the circumstances of the crime, but the downward departures for larceny and burglary suggest that many of these were relatively petty crimes committed by men and women in dire straits. Downward departures for women, African Americans and immigrants are consistent with juridical mercy shown to the poor. But this speculation, without proof, begs the question. It is too easy to speculate that discrimination or poverty drove desperate people to theft, but neither poverty nor discrimination explains a unique historical experience for any broadly defined group. Many blacks, immigrants and irregularly employed laborers suffered the insults of poverty and prejudice without turning to crime. Moreover, compared to common laborers (the excluded category) no occupational group, except salesmen, received significant upward sentencing departures for thievery. Even Du Bois (1899/1967), who despaired of the living conditions of African Americans in late-19th-century Philadelphia, would not attribute black crime to black poverty. About 10% of Philadelphia’s blacks were paupers, yet he attributed only about 10% of black incarcerations to poverty per se. Du Bois attributed excess black criminality to poor education, rural immigrants unprepared for urban life, the ‘‘temptations of the city,” and diverse other social causes. Poverty played only a small role. Where juries treated thieves leniently, they had little patience for the white collar crimes when committed by white collar workers. Contemporaries commented on the perniciousness of forgery in a commercial society. Left unchecked, forgery, fraud and embezzlement threatened to undermine the entire edifice of trust on which commercial transactions were built. Pennsylvania’s 1767 law punished counterfeiters and forgers with the pillory followed by the gallows and the preamble to a 1781 act read: ‘‘It is the duty and interest of all governments to prevent fraud and promote the interests of just and useful commerce” (Hartz, 1948, p. 204). The state had eliminated the sanguinary punishment by 1829, but imposed a greater maximum sentence for forgery than for manslaughter. Writing shortly after Pennsylvania adopted its revised criminal code, Montagu (1830, p. 155) argued that the ‘‘nature and magnitude of the crime of forgery ... are always alarming” because it was the consequence, not of ‘‘hardened obduracy, but depraved intellect.” As such, it was deserving of especially harsh punishment. In 1859, the Pennsylvania supreme court opined forgery was a crime ‘‘of the most dangerous character” because if it could be done with impunity ‘‘all confidence... must be destroyed.”15 Given these attitudes, it is not surprising to find notable

15

Biles v. Commonwealth (Pa. 1859), 537.

H. Bodenhorn / Explorations in Economic History 46 (2009) 287–298

297

upward departures for white collar workers convicted of white collar crimes, with the 31% upward departure for professionals being particular severe. 7. Concluding remarks A fundamental problem in comparing modern and historical episodes in criminal sentencing follows from the centrality of the jury in the 19th century and that of the judge in the modern era. Albonetti (1991, 1997) believes that modern judges are subject to prejudices that may lead to systematically longer sentences for some groups as a result of statistical discrimination. This article represents an investigation into the nature of mid-19th-century criminal procedure using a newly compiled data set and studies how 19th-century juries navigated the complexities of imposing punishments when they were just as racked by uncertainty and bounded rationality as modern judges and just as subject to well-known social prejudices toward certain groups. Unlike judges constrained by formal or informal sentencing guidelines, 19th-century juries were provided with greater discretion. Uncertainty and bounded rationality does not excuse the discriminatory exercise of discretion, but it sheds light on the extent to which racial and ethic stereotypes had salience for juries. When jury’s members attributed certain criminal propensities to different groups, sentencing disparities emerged. Evidence on 19th-century Pennsylvania sentencing practices yields several unexpected results. Traditional accounts of the era portray rampant and virulent racism and nativism, yet African Americans and immigrants were not subject to unwarranted upward sentencing departures. On average, blacks were treated no differently than native-born whites at the sentencing stage of the criminal process, though they were more likely to receive an excessive sentence and they were notably overrepresented in Pennsylvania’s prison population. A second intriguing result is the favorable treatment of immigrants, especially the Irish who typically received shorter sentences for a given crime than native-born whites. Social interactions are believed to positively influence the level of criminal activity (Glaeser et al., 1996; Pattachini and Zenou, 2008), and may have done so in the mid-19th century. Social interactions, as evidenced through membership in social clubs, churches, and beneficial societies may have worked to the benefit of immigrant offenders because they provided resources or other forms of support for criminal offenders. Alternatively, if the traditional interpretation of immigrant poverty and misery are accurate, contemporary juries may have taken pity on immigrant offenders or believed that they posed a lesser threat than native-born criminals. Future research may be able to sort out the relative importance of these alternative explanations. Much is yet to be learned about criminal practice in the 19th-century United States. This study represents an effort to shed some light on the economics of crime, but it raises nearly as many questions as it answers. Historical prison records exist for other states and it would be worthwhile to compare sentencing practices outside Pennsylvania. Another potentially fruitful line of inquiry would be to better control for selectivity by following criminal trials from their inception. Naylor (1979) used records on arrests and criminal trials for his study of late-19th-century Philadelphia criminality. It would be informative to link these records to the prison records to better understand how the criminal justice system resolved cases that did not necessarily end with a prison commitment. Acknowledgments I thank Hal Hochman, Carolyn Moehling, Ron Oaxaca, Anne Piehl, Martin Schonger, seminar participants at Clemson University and two anonymous referees for insightful comments on earlier versions of this paper. I thank Professor John Langbein at Yale Law School for affording me the opportunity to attend his course on the history of the common law in which I learned much about the history of criminal procedure. I also thank Supriya Sarnikar and Ron Oaxaca for permission to cite their unpublished paper. Veronica Hart provided invaluable research assistance. Financial support was provided by the Irish American Cultural Institute made possible by a gift from the Lawrence M. O’Shaughnessy family and the National Science Foundation, SES-0109165. References Albonetti, C.A., 1991. An integration of theories to explain judicial discretion. Social Problems 38 (2), 247–266. Albonetti, C.A., 1997. Sentencing under the federal sentencing guidelines: effects of defendant characteristics, guilty pleas, and departures on sentence outcomes for drug offenses, 1991–1992. Law & Society Review 31 (4), 789–822. Anonymous. 1853. Slavery and slave statistics of the South, etc. De Bow’s Review 14 (n.s.), 593–596. Becker, G., 1968. The economics of crime. Journal of Political Economy 76 (2), 169–217. Bodenhamer, D.J., 1985. Criminal justice and democratic theory in antebellum America: the grand jury debate in Indiana. Journal of the Early Republic 5 (4), 481–502. Bodenhorn, H., 2002. The mulatto advantage: the biological consequences of complexion in rural antebellum Virginia. Journal of Interdisciplinary History 33 (1), 21–46. Bodenhorn, H., 2006. The economic consequences of colorism and complexion homogamy in the black community: some historical evidence. American Economic Review 96 (2), 256–260. Bodenhorn, H., Ruebeck, C.S., 2007. Colourism and African–American wealth: evidence from the 19th-century south. Journal of Population Economics 20 (4), 599–620. Bouton, T., 2007. Taming Democracy: ‘‘The People”, the Founders, and the Troubled Ending of the American Revolution. Oxford University Press, New York. Bushway, S.D., Piehl, A., 2001. Judging judicial discretion: legal factors and racial discrimination in sentencing. Law & Society Review 35 (4), 733–764. Clark, D., 1973. The Irish in Philadelphia: Ten Generations of Urban Experience. Temple University Press, Philadelphia. Clay, H. 1829. An Address Delivered to the Colonization Society of Kentucky, at Frankfort, December 17, 1829. In Report of the Board of Managers of the Pennsylvania Colonization Society. Philadelphia: Thomas Kite.

298

H. Bodenhorn / Explorations in Economic History 46 (2009) 287–298

Du Bois, W.E.B. 1899/1967. The Philadelphia Negro: A Social Study. Schocken Books, New York. Eastern State Penitentiary. 1829–1857. Population Records, Descriptive Registers. Record Group 15. Records of the Department of Justice. Harrisburg, Pennsylvania State Library. Eastern State Penitentiary. 1842–1869. Convict Reception Register. Record Group 15. Records of the Department of Justice. Harrisburg, Pennsylvania State Library. Engen, R.L., Gainey, R.R., 2000. Modeling the effects of legally relevant and extralegal factors under sentencing guidelines: the rules have changed. Criminology 38 (4), 1207–1229. Engen, R.L., Steen, S., Bridges, G.S., 2002. Racial disparities in the punishment of youth: a theoretical and empirical assessment of the literature. Social Problems 49 (2), 194–220. Epstein, B.S. 1981. patterns of sentencing and their implementation in Philadelphia city and county, 1795–1829. Unpublished Ph.D. dissertation, University of Pennsylvania. Friedman, L.M., 1973. A History of American Law. Simon and Schuster, New York. Friedman, L.M., 1993. Crime and Punishment in American History. Basic Books, New York. Hartz, L., 1948. Economic Policy and Democratic Thought: Pennsylvania, 1776–1860. Harvard University Press, Cambridge. Glaeser, E.L., Sacerdote, B., Scheinkman, J., 1996. Crime and social interactions. Quarterly Journal of Economics 111 (2), 508–548. Green, W.H., 1990. Econometric Analysis. Macmillan Publishing Company, New York. Griswold, D.B., 1987. Deviation from sentencing guidelines: the issue of unwarranted disparity. Journal of Criminal Justice 15 (4), 317–329. Gross, A., 1998. Litigating whiteness: Trials of racial determination in the nineteenth-century south. Yale Law Journal 108 (1), 109–188. Haller, M.H., 1979. Plea bargaining: the 19th century context. Law & Society Review 13 (2), 273–279. Hanawalt, Barbara A. 1979. Crime and Conflict in English Communities, 1300–1348. Harvard University Press, Cambridge and London. Hirsch, A.J., 1992. The Rise of the Penitentiary: Prisons and Punishment in Early America. Yale University Press, New Haven. Ignatiev, N., 1995. How the Irish Became White. Routledge, New York. Ireland, R.M., 1995. Privately funded prosecution of crime in the 19th-century United States. Journal of Legal History 39 (1), 43–58. Kuntz, William Francis, II. 1988. Criminal Sentencing in Three 19th-Century Cities: Social History of Punishment in New York, Boston, and Philadelphia, 1830–1880, Garland Publishing, Inc., New York and London. Lane, Roger. 1986. Roots of Violence in Black Philadelphia, 1860–1900, Harvard University Press, Cambridge and London. Lane, Roger. 1997. Murder in America: A History, Ohio State University Press, Columbus. Langbein, J.H., 1978. The criminal trial before the lawyers. University of Chicago Law Review 45 (2), 263–316. Lauritsen, J.L., Sampson, R.J. 1998. Minorities, Crime, and Criminal Justice. The Handbook of Crime and Punishment. Edited by M. Tonry. Oxford University Press, New York. Moehling, C.M., Piehl, A.M. 2007. Immigration and Crime in Early 20th Century America, Cambridge, Mass., NBER working paper. Montagu, B., 1830. Thoughts on the Punishment of Death or Forgery. William Pickering, London. Mosher, C., Hagan, J., 1994. Constituting class and crime in upper Canada: the sentencing of narcotics offenders, circa 1908–1953. Social Forces 73 (3), 613– 641. Mustard, D.B., 2001. Racial, ethnic, and gender disparities in sentencing: evidence from the US federal courts. Journal of Law and Economics 44 (1), 285–314. Naylor, Timothy J. 1979. Criminals, Crime, and Punishment in Philadelphia, 1866–1916. Unpublished Ph.D. dissertation, University of Chicago. Note. 1964. The changing role of the jury in the 19th century. Yale Law Journal 74(1), 170–192. Pattachini, E., Zenou, Y., 2008. The strength of weak ties in crime. European Economic Review 52 (2), 209–236. Pennsylvania Commission on Sentencing. 2005. Sentencing Guidelines (6th ed.) Available from: (accessed May 2008). Pennsylvania General Assembly. 1810. Laws of the Commonwealth of Pennsylvania, vols. 4, John Bioren, Philadelphia. Pennsylvania General Assembly. 1829. Laws of the General Assembly of the State of Pennsylvania Passed at the Session of 1828–29. Office of the ‘‘Reporter, Harrisburg.” Pennsylvania General Assembly. 1866. Laws of the General Assembly of Pennsylvania. Singerly & Myers, Harrisburg. Reiss Jr., A.J., 1965. Occupations and Social Status. Free Press, New York. Reitz, K.R., 1998. Modeling discretion in American sentencing systems. Law & Policy 20 (4), 389–428. Rice, J.D., 1996. The criminal trial before and after the lawyers: authority, law, and culture in Maryland jury trials, 1681–1837. American Journal of Legal History 40 (4), 455–475. Sarnikar, S., Sorensen, T., Oaxaca, R., 2006. Do You Receive a Lighter Prison Sentence Because You Are a Woman? An Economic Analysis of Federal Criminal Sentencing Outcomes. Working paper, University of Arizona. Seymour, G., 1992. Downward departures from the federal sentencing guidelines based on the defendant’s drug rehabilitation efforts. University of Chicago Law Review 59 (2), 837–864. Shapiro, D.L., 1991. Sentencing the reformed addict: departure under the federal sentencing guidelines and the problem of rehabilitation. Columbia Law Review 91 (8), 2051–2073. Steinberg, A., 1986. The spirit of litigation: Private prosecution and criminal justice in nineteenth century Philadelphia. Journal of Social History 20 (2), 231– 249. Steinberg, A., 1989. The Transformation of Criminal Justice. Philadelphia, 1800–1880. University of North Carolina Press, Chapel Hill. Tyson, J.R., 1827. Essay on the Penal Law of Pennsylvania. Mifflin & Parry, Philadelphia. Vogel, M.F., 1999. The social origins of plea bargaining: conflict and the law in the process of state formation, 1830–1860. Law & Society Review 33 (1), 166– 246. Waldfogel, J., 1993. Criminal sentences as endogenous Taxes: are they ‘‘Just” or ‘‘Efficient”? Journal of Law and Economics 36 (1), 139–151. Walker, S., 1980. Popular Justice. A History of American Criminal Justice. Oxford University Press, New York. Wedgewood, W.B. 1844. The Revised Statutes of the State of Pennsylvania, and Additional Laws to 1844, Reduced to Questions and Answers for the Use of Schools and Families. Thomas, Copperthwait & Co., Philadelphia. Weitzer, R., 1996. Racial discrimination in the criminal justice system: findings and problems in the literature. Journal of Criminal Justice 24 (4), 309–322. Western State Penitentiary. 1826–1876. Convict Docket & Descriptive Register. Record Group 15. Records of the Department of Justice. Pennsylvania State Library, Harrisburg. Winship, C., Mare, R.D., 1992. Models for sample selection bias. Annual Review of Sociology 18, 327–350.