Punishment and Crime: An Examination of Some Empirical Evidence Author(s): Theodore G. Chiricos and Gordon P. Waldo Source: Social Problems, Vol. 18, No. 2 (Autumn, 1970), pp. 200-217 Published by: Oxford University Press on behalf of the Society for the Study of Social Problems Stable URL: http://www.jstor.org/stable/799582 Accessed: 26-06-2016 20:14 UTC Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://about.jstor.org/terms JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org. Society for the Study of Social Problems, Oxford University Press are collaborating with JSTOR to digitize, preserve and extend access to Social Problems This content downloaded from 137.99.31.134 on Sun, 26 Jun 2016 20:14:06 UTC All use subject to http://about.jstor.org/terms PUNISHMENT AND CRIME: AN EXAMINATION OF SOME EMPIRICAL EVIDENCE* THEODORE G. CHIRICOS AND GORDON P. WALDO The Florida State University Data from National Prisoner Statistics and Uniform Crime Reports are used to re-examine the relationship between rates of crime and the certainty and severity of punishment in the states of the United States. Recent research by Gibbs and Tittle using similar data to test deterrence hypotheses are extended in two ways: (1) by examining the relationships for three points in time which were previously examined for a single time period, and (2) by relating changes in rates of crime to prior changes in the certainty and severity of punishment. Little consistent support is found for the hypotheses that rates of crime and the certainty and severity of punishment are inversely related. The data for certainty suggest that the relationship with crime rates is somewhat variable over time and among offenses. These findings contrast with the strong and consistent negative associations reported by Tittle. An attempt is made to account for the strength and consistency of the previous findings by exploring the problem of correlated bias that exists in Tittle's measures of certainty and criminality. The inconsistency of findings, both within this study and among recent empirical tests of deterrence theory, indicate that existing sources of data may be inappropriate for use in the testing of deterrence hypotheses. The question of whether penal sanctions deter crime is still being asked (Tittle, 1969; Jensen, 1969; Zimring and Hawkins, 1968; Gibbs, 1968, 1966; Chambliss, 1966; Bittner and Platt, 1966) after more than a century of debate among sociologists, crimino- logists, and legal philosophers. Given the importance of this issue to our understanding of social control and the maintenance of social order, it is surprising that its resolution seems little nearer today than it was in the 19th century. Most of the input for this debate has been theoretical or moralistic in nature (Mead, 1918; Wood, 1938; Ball, 1955; Polier, 1956; Barnes and Teeters, 1959; Toby, 1964; Jeffery, 1965), and as noted by Chambliss (1966:70), "... the question of deter- rence has frequently turned into a de* This research was partially supported by Ford Foundation Grant #67-586. The authors would like to express their appreciation to Charles Tittle, Jack Gibbs, Roland Chilton, and Harry Allen for their com- ments and suggestions on an earlier draft of this paper. bate over the morality of capital punishment." For a time, sociologists and "enlightened" jurists showed a tendency to reject Bentham's (1843) "classical" hypothesis that man avoids criminal behavior if that behavior elicits swift, severe, and certain punishment. This common rejection of deterrence theory is summarized by Horton and Leslie (1965:154) who conclude: "there is no evidence that states with more severe penalties have lower crime rates or recidivism rates than states with lighter penalties." Indeed, until recently, empirical evidence on the question of deterrence has been limited to studies showing the apparent ineffectiveness of capital punishment in deterring crim- inal homicide (Sellin, 1967; Savitz, 1958; Schuessler, 1952). Contradic- tory empirical evidence-in support of the deterrence hypothesis-has been slow to accumulate. Several recent studies, however, have suggested the possibility that punishment may in fact deter criminality. Chambliss (1966) reported that the This content downloaded from 137.99.31.134 on Sun, 26 Jun 2016 20:14:06 UTC All use subject to http://about.jstor.org/terms Punishment and Crime 201 incidence of parking violations on a states that were above and below the college campus dropped sharply when median level of certainty, Gibbs (1968: 523-527) found an inverse relationship between certainty of punishment and the severity and certainty of punishment for this offense was increased. Having analyzed the frequency of parking vio- lations before and after a change in regulations, Chambliss (1966:74) con- cluded that "the changes had the effect of substantially reducing the number of violations."' Gibbs (1968:519-521) introduced a new dimension to the study of deterrence by presenting operational definitions of certainty and severity of punishment that were based upon existing, though admittedly limited, sources of aggregate data. Using the Federal Bureau of Investigation's Uniform crime Reports and the National Prisoner Statistics published by the Federal Bureau of Prisons, an index of the certainty of punishment was given by a ratio of prison admissions for homicide to the number of homicides known to the police. The median number of criminal homicide rate (Phi = -.48). A similar analysis for severity and rates of homicide also showed a negative, though weaker association (Phi = -.25). Noting that crime rates may be as much a product of uncontrolled etiological (extra-legal) factors as they are a response to penal sanctions, Gibbs (1968:529-530) concluded: ... all that can be said of the findings is that they question the common assertion that no evidence exists of a relationship between legal reactions to crime and the crime rate. Tittle's (1969) operational definitions of certain and severe punishment were similar to those introduced by Gibbs2 and were applied to a study of all seven F.B.I. "Crime Index" offenses. Tittle (1969:415) found that for the months served by persons for homicide seven crimes considered together, certainty of punishment and rates of "de- provided Gibbs with an index of the severity of punishment. Comparing (Tau-c = -.45). When the relation- viance,"3 were negatively associated 1 Cohen (1965:10) provides some additional insights into other factors that might have helped reduce the number of parking violations. "Several years ago the Indiana University faculty had a high rate of violation of campus parking regulations, in part because of the disjunction between the demand for parking spaces and the supply. . . . One response to this situation was to create new parking lots and to expand old ones." 2 Gibbs' (1968) index of certainty of imprisonment for homicide, and Tittle's (1969) certainty of punishment for individual felonies were given by the following ratios: # of State Prison Admissions for Homicide in 1960 Gibbs: Mean # of Homicides Known to Police for 1959-1960 Tittle: # of State Prison Admissions for "X" Offense in 1960 and 1963 # of "X" Crimes Known to the Police in 1959 and 1962 The severity of punishment was indexed by Gibbs (1968) as "the median number of months served on a homicide sentence by all persons in prison on December 31, 1960." Tittle (1969) used "mean length of time served for felony prisoners released from state prisons in 1960" as his principal index of severity. 3 Tittle (1969) referred to his measure of criminality as a "Deviance Index" and his operational definition of that index for individual felonies was given by: Mean # of "X" Offenses Known to Police for 1959 and 1962. 1960 Population This content downloaded from 137.99.31.134 on Sun, 26 Jun 2016 20:14:06 UTC All use subject to http://about.jstor.org/terms 202 SOCIAL PROBLEMS ships between certainty of punishment and rates of deviance were examined for the seven individual offenses, they proved to be consistently negative and ranged in strength from -.57 for sex offenses to -.08 for auto theft. Whereas these data for certainty offered some support for the deterrence hypothesis, severity of punishment was positively related to rates of deviance for all crimes excepting homicide (-.45). While Gibbs studied only homicide, and Tittle examined seven offenses in- cluding homicide, their findings for this offense are somewhat contradictory. Both found certainty and severity of punishment to be inversely related to rates of homicide. However, in Gibbs' analysis, the relationship is stronger for certainty than for severity, while in Tittle's research, this represents an extension of Gibbs' analysis to include additional offenses. However, an effort is made to approach the testing of the deterrence hypothesis in terms that are less "static" than those used by Gibbs and Tittle. The operational definitions of certainty and criminality allow for a test of the expected deterrence relationship at three points in time. For severity, the deterrence hypothesis will be tested at two points in time. By affording a check on the reliability of empirical relationships, this approach extends the efforts of Gibbs and Tittle. A second departure in this approach to the deterrence question is provided by a focus upon changes in rates of crime that follow changes in levels of certainty and severity. This approach is Tittle's data the situation is reversed. prompted by several considerations. In fact, the relationship between certainty of punishment and homicide was First, there is an assumption-implicit in deterrence theory-that a causal relationship exists between punishment and criminality. In the absence of experimental controls one might logically consider whether changes in "X" (certainty or severity) precede or accompany changes in "Y" (crime rates). The relevance of change is further underscored by the link between the deterrence issue and aspects of learning not statistically significant (-.17) in Tittle's study. Since the findings of Gibbs and Tittle are inconsistent, even though they used similar operational definitions for almost identical time periods, further examination of the relationship between crime rates and legal sanctions appears warranted. The research reported here is a further test of the hypothesis that rates of crime are inversely related to the certainty and severity of punishment.4 Like It will be noted that a major portion of the "deviance index"-1959 Crimes Known to Police--is composed of data which reflect theory. Homans (1961:55) has sug- gested that the more often a man re- ceives "reinforcement" in a given period, the less likely will additional (comparable) reinforcement be effect- ive. Applying this interpretation to both positive (reward) and negative criminal activity preceding the measures of certainty and severity used by Tittle. This would seem to partially violate the logic of deterrence theory, wherein reduced criminality, or deviance is expected to follow the imposition of certain or severe punishment. ment is important in the theoretical formu- 1966; Gibbs, 1968; Tittle, 1969), the celerity of punishment has not been ex- research. 4 As in previous research (Chambliss, (punishment) reinforcement, Jefferey (1965:298) concluded that "continous amined. While this dimension of punishlation of deterrence theory, avaialable data restrict its incorporation for this kind of This content downloaded from 137.99.31.134 on Sun, 26 Jun 2016 20:14:06 UTC All use subject to http://about.jstor.org/terms Punishment and Crime 203 punishment will not control behavior through the criminal justice system either, for satiation takes place .. " Even with high levels of certain and from the point at which an offense be- severe punishment, a constancy in these negative sanctions could diminish their deterrent effectiveness. Thus, in addi- judicial disposition. Regrettably, this approach is impossible under existing conditions for any meaningful number of cases and one is forced to construct tion to absolute levels of certainty and severity, important variables may be the degree of change in certainty and severity. In fact, one might argue that potential criminal offenders are more likely aware of broad trends and changes in certainty and severity of punishment, than of their absolute levels. For example, an awareness that the government is attempting to increase the certainty and severity of pun- ishment for narcotics use may serve as a greater deterrent to middle-class Americans than any understanding of the exact probability (certainty) and magnitude (severity) of punishment. This "change" orientation is also re- flected in Chambliss' (1966:73) approach to the question of deterrence through punishment: ... change in regulations (and the cor- responding increase in the certainty and severity of punishment) greatly reduced the number of transgressions and the num- ber of transgressors, and served as an effective deterrent. [Italics added.] RESEARCH METHOD As expressed by Gibbs and Tittle, and as operationalized here, certainty of punishment actually refers to the certainty of imprisonment as a final disposition for crimes known to the comes known to the police, to final estimates of certainty from available sources of information.5 Using data from Uniform Crime Reports (Federal Bureau of Investiga- tion, 1950:82-85; 1951:79-82; 1960: 38-51; 1961:38-52; 1963:38-52; 1964: 52-66) and National Prisoner Statistics (Federal Bureau of Prisons, 1954:72- 74; 1960:41; 1964a:26-27), an index of the certainty of imprisonment for each of six offenses6 was computed for 1950, 1960, and 1963. The index gives the ratio of prison admissions for one year to the average number of crimes known to the police in that year and the previous year, for a specific crime category. The three certainty indices were computed for each state as follows: 5 Unless one had the means to conduct an immensely expensive research project in all 50 states, these data are unavailable. Indeed, such a project would necessitate covering all police jurisdictions--or a meaningful sample of jurisdictions-within each state for a number of years before the desired data were available. 6 The category of "sex offenses" was in- corporated into Tittle's (1969) study, but eliminated here because of definitional in- consistencies between Uniform Crime Reports and National Prisoner Statistics. Uniform Crime Reports provides data for "forci- police. An estimate of this certainty for ble rape," while National Prisoner Statistics combines this crime with a large group of each offense category is given by the was also eliminated from the present analysis proportion of "crimes known to the police" which result in imprisonment. Ideally, this estimate would be achieved by tracing large numbers of cases rather diverse "sex offenses." "Total crimes" even though Tittle (1969) included this category. "Total crimes" is essentially a meaningless category since it encompasses entirely different elements in the two sources of data. This content downloaded from 137.99.31.134 on Sun, 26 Jun 2016 20:14:06 UTC All use subject to http://about.jstor.org/terms 204 SOCIAL PROBLEMS 1950 Admissions to Prison for "X" Offense 1950 Certainty = Mean of "X" Crimes Known to Police in 1949 & 1950 1960 Admissions to Prison for "X" Offense 1960 Certainty = Mean of "X" Crimes Known to Police in 1959 & 1960 1963 Admissions to Prison for "X" Offense 1963 Certainty = Mean of "X" Crimes Known to Police in 1962 & 1963 In addition to facilitating three tests of the relationship between certainty of punishment and rates of crime, the several measures of certainty may be used reflected in several sources of available to give three indices of the change in certainty over time. Change in the certainty of punishment refers to the in- points in time and change in severity crease or decrease in the probability that criminal behavior will result in im- prisonment. Operationally, change in certainty is given by the percentage change (increase or decrease) in the certainty of punishment between each information.' However, only "median length of sentence served by released prisoners" allows one to look at two over time. Thus, "median sentence served" was recorded for 1960 (Federal Bureau of Prisons, 1960:69) and 1964 (Federal Bureau of Prisons, 1964b:52); and the two were used to compute the percentage change (increase or decrease) in severity of pun- ishment over time. The operational pair of years: 1950-1960; 1950-1963; definitions of severity and change in and 1960-1963. Thus, for a particular severity are briefly summarized: 1960 Severity = Median Length of Sentence Served by State Prisoners Released in 1960 1964 Severity = Median Length of Sentence Served by State Prisoners Released in 1964 Percentage Change in Severity = (1964 Severity) - (1960 Severity) (1960 Severity) state, the changes in certainty were com- The presumed "dependent variable" puted as follows: in deterrence research is the criminal (1960 Certainty) - (1950 Certainty) 1950-1960 % Change in Certainty = (1950 Certainty) (1963 Certainty) - (1950 Certainty) 1950-1963 % Change in Certainty = (1950 Certainty) (1963 Certainty) - (1960 Certainty) 1960-1963 % Change in Certainty = (1960 Certainty) The severity of punishment administered by states for specific offenses, is 7 As noted earlier, Tittle (1969) used "mean length of time served for felony This content downloaded from 137.99.31.134 on Sun, 26 Jun 2016 20:14:06 UTC All use subject to http://about.jstor.org/terms Punishment and Crime 205 behavior which is thought to be reduced by certain and severe punishment. Rates of crime and percentage of crime rate change, provide ready, although admittedly crude, measures of the de- pendent variable in this analysis. Mean crime rates were computed for three- year periods following each point in time represented by the several indices of certainty and severity of punishment. This procedure closely approximates Gibbs' (1968) earlier methodology. The mean rates of crime for (1950, 1951, 1952), (1960, 1961, 1962), and (1963, 1964, 1965) were compared respectively with certainty in 1950 (Time 1), 1960 (Time 2), and 1963 (Time 3). The mean crime rate for 1960, 1961, and 1962 was compared with severity of punishment in 1960, while the mean for 1964, 1965, and 1966 was compared with 1964 severity. In essence, it was expected that there would be an inverse relationship between the certainty and severity of pun- ishment as measured in a single period, and the average rate of criminal offense in the subsequent period. For each measure of percentage change in severity or certainty of punishment, several indices of the percent- age change in crime rate were computed. For example, percentage change three periods of time:8 1962-1965; 1963-1966; 1964-1967. The several measures of change in crime rate cover time periods that begin either near the end of the period of change in certainty, or at its conclusion. This allows a change in certainty or severity to substantially precede a change in rates of crime. Similar computations of crime rate change were made with regard to the additional measures of change in certainty and change in severity of punishment. Classical criminology and the recent findings of Chambliss, Gibbs, and Tittle,9 suggest the following hypo- theses: 1. There is an inverse relationship between the certainty of punishment and crime rates for each time period under consideration. 2. There is an inverse relationship between the severity of punishment and crime rates for each time period under consideration. 3. There is an inverse relationship between change in certainty of punishment and change in crime rates. 4. There is an inverse relationship between change in severity of punishment and change in crime rates. in certainty of punishment between 1960 and 1963 was compared with percentage change in crime rate for The data are analyzed in 2 X 2 contingency tables using the states in the prisoners released from state prisons in age change in crime rate is given by the 1960" as a measure of the severity of pun- ishment for specific offense. For "total felonies" Tittle used "the median sentence for state felony offenders imprisoned in 1960," as well as "the number of crimes punishable by death" in a state. These data, as well as data for "median length of sentence served by prisoners" in 1960 and 1964, were considered prior to our selection of the latter. 8 The operational definitions for percent- following: Crime Rate (Time 2) - Crime Rate (Time 1) Crime Rate (Time 1) 9 It should be emphasized that Tittle (1969:415) found little support for the hypothesis that rates of crime and the severity of punishment are inversely related. The expected inverse relationship was found only for homicide. This content downloaded from 137.99.31.134 on Sun, 26 Jun 2016 20:14:06 UTC All use subject to http://about.jstor.org/terms 206 SOCIAL PROBLEMS TABLE 1 PHI-COEFFICIENTS FOR CERTAINTY OF PUNISHMENT AND CRIME RATES IN THREE TIME PERIODS FOR SIXMAJOR CRIMES Time 1 Time 2 Time 3 1950 Certainty With 1960 Certainty With 1963 Certainty With Mean of 1950, Mean of 1960, Mean of 1963, 1951, and 1952 1961, and 1962 1964, and 1965 Crime Rates Crime Rates Crime Rates Homicide .02 -.46a Robbery -.26 -.33a Burglary -.09 Assault -.35a Larceny Auto -.38b -.04 Theft -.33a -.17 -.16 -.17 -.25 -.42b --.50 -.25 -.15 -.16 a = p <.05. b = p <.01. e =p <.001. U.S. as the unit of analysis. Each variable is divided at the median into high and low scores.10 Phi-coefficients and chi-squares are computed to show the strength, direction, and significance of the associations." Because each rela- tionship is examined for six offense categories, the number of contingency tables is too large for the tables to be presented in the paper. Instead, the relationships are summarized by offense category with phi-coefficients and levels of significance reported. 10 This procedure is identical with that used by Gibbs (1968:523-527), but differs from that of Tittle (1969:414). The latter rank-ordered the states on each of the several indices, and then divided each distribution into "four to nine approximately equal FINDINGS The data of Table 1 show the rela- tionship between certainty of punish- ment and rates of crime for six offenses -at three points in time: 1950 certainty with the mean of 1950, 1951, and 1952 crime rates (Time 1), 1960 certainty with the mean of 1960, 1961, and 1962 crime rates (Time 2), and 1963 certainty with the mean of 1963, 1964, and 1965 crime rates (Time 3). Although all but one of the relationships are negative-as hypothesized- the strength and statistical significance of the associations vary greatly by offense and over time. Indeed, only seven of the 18 phi-coefficients were statis- tically significant (p <.05). categories of states, depending upon the nature of the distribution and apparent 'natural' groupings . . ." Thus, it was im- When relationships are compared at three points in time, only assault shows a consistently significant inverse rela- ology, because the cutting points for the tionship between certainty of punishment and rates of crime. For burglary, the relationship appears statistically significant at Times 2 and 3, but is very possible to replicate Tittle's precise methodseveral distributions were not reported. 11 This mode of statistical analysis was also used by Gibbs (1968:523-527). The reported phi-coefficients should be somewhat comparable to Tittle's (1969:414-415) Tau-c correlations, inasmuch as Tau-c = Phi in the special case of 2 X 2 classifications. weak for Time 1. The negative cor- relations are fairly consistent for robbery, but only in Time 2 is the relation- This content downloaded from 137.99.31.134 on Sun, 26 Jun 2016 20:14:06 UTC All use subject to http://about.jstor.org/terms Punishment and Crime 207 TABLE 2 PHI-COEFFICIENTS FOR SEVERITY OF PUNISHMENT AND CRIME RATES IN Two TIME PERIODS FOR SIX MAJOR CRIMES Time 1 Time 2 1960 Severity With 1964 Severity With Mean of 1960, Mean of 1964, 1961, and 1962 1965, and 1966 Crime Rates Crime Rates Homicide -.33a Robbery Assault .12 .32a -.03 .00 -.08 Burglary .25 .08 Larceny .14 -.09 Auto Theft .13 .12 a -- p <.05. ship significant. For larceny and auto theft, the negative associations are consistently weak and non-significant at each time period. Homicide presents one of the more intriguing patterns with a low positive Table 2 summarizes the relationship between severity of punishment and crime rates for two time periods: 1960 severity with the mean of 1960, 1961, nificant negative association (-.46) in Time 2, and a small negative asso- The findings provide no consistent support for the deterrence hypothesis. The for Time 2 are similar to the findings of Gibbs, and the data for Time 3 are relationships between severity and crime rates are-with one exceptioneither positive or extremely weak. Moreover, there is little consistency association (.02) in Time 1, a sig- ciation (-.17) in Time 3. The data supportive of Tittle; but in Time 1 the data fail to support either Gibbs or Tittle.12 In short, although all but one of the associations presented in Table 1 are negative,13 the examination of the data for three points in time indicates little reliability in the relationship between the certainty of punishment and crime rates. 12 Gibbs (1968:526) found the relationship between certainty and criminality to be -.48 (phi) and Tittle (1969:415) obtained a Tau-c of -.17. 13 Since there is an artifact in the data which virtually assures that the phi-coefficients will be negative, the reader should not view this consistent pattern of negative associations as providing very strong support for deterrence theory. This issue will be ex- amined in detail in the discussion section. and 1962 crime rates (Time 1), and 1964 severity with the mean of 1964, 1965, and 1966 crime rates (Time 2). over time, within offense categories. For example, a significant inverse relationship (Phi = -.33) between severity of punishment and rates of homicide for Time 1 is reduced to -.03 for Time 2. While the earlier (Time 1) relationship is similar to the findings of Gibbs and Tittle,14 the data for Time 2 represent a substantial contradiction. For the remaining offense categories, the relationship between severity of punishment in 1960 (Time 1) and rates of crime is positive, and for assault, this association (Phi = .32) 14 The relationship between the severity of punishment and homicide rates was cal- culated by Gibbs (1968:526) at Phi= -.25, and by Tittle (1969:415) at Tau-c = -.45. This content downloaded from 137.99.31.134 on Sun, 26 Jun 2016 20:14:06 UTC All use subject to http://about.jstor.org/terms 208 SOCIAL PROBLEMS is statistically significant (p < .05). However, at Time 2 severity of punish- between the measure of percentage change in certainty and crime rate practical purposes, completely unrelated. change. However, the relationships between percentage change in certainty Three of the associations are negative and three are positive, with the highest being Phi - .12 for auto theft. If there is a correlation between the severity of magnitude. Thirty-four of the 54 associations are positive, 18 are negative, ment and rates of crime are, for all and crime rate in Table 3, are both inconsistent in direction, and low in punishment and subsequent criminality, and two are zero. Only six of the 54 our analysis-covering two points in phi-coefficients are statistically signifi- Again, questions must be raised con- Thirty of the coefficients were com- cerning the reliability of the deterrence hypothesis. An alternative test of the deterrence puted at ?+L .11 or less. Only for assault time-has not been able to show it. cant, and four of these are positive. does one find more inverse than positive relationships between change in the hypothesis is given by an analysis of certainty of punishment and the nine changes in the certainty and severity of punishment and changes in crime rates. Table 3 contains the phi-coefficients for measures of crime rate change. In brief, percentage change in certainty 1950- punishment will be followed by decreased levels of crime, or that de- 1960, 1950-1963, and 1960-1963 with percentage change in crime rate for selected periods.15 The first three columns of Table 3 summarize the com- puted relationships between change in the certainty of punishment from 1950 to 1960 and change in rates of crime for three periods of time. Columns four through six give similar phi-coefficients for change in the certainty of punishment 1950-1963, and the last three columns give the correlations of change in certainty 1960-1963, with change in crime rate. If the viability of the deterrence hypothesis is to be supported, we would expect consistent inverse relationships 15 It is difficult to determine the most logical time periods to use in correlating percentage crime rate change with percent- age certainty change 1950-1960 and 19501963. The crime rates used were those that seemed most logical to the authors in attempting to present the dependent variable as following in part, or completely, after the independent variable. Percentages of crime rate change for other year periods were also examined, however, and the pattern of findings was quite similar to that presented. these data provide no support for the hypothesis that increased certainty of creased levels of certainty will be followed by increased crime. Table 4 summarizes the computed relationships between change in the severity of punishment from 1960 to 1964 and changes in the rates of crime for several four year periods: 19611965, 1962-1966, and 1963-1967. It was expected that increased severity of punishment would be associated with decreased rates of crime, and that a decrease in the level of severity would be followed by an increase in crime. The relationships involving change in severity are as inconsistent as the re- lationships involving change in certainty. The only statistically significant relationship was a positive association (.49) for larceny. Thirteen of the associations are positive and five are nega- tive. The only consistently negative relationships are found for burglary, wherein non-significant associations of -.21 were obtained for all three time periods. Again, the data in Table 4 provide no consistent support for the This content downloaded from 137.99.31.134 on Sun, 26 Jun 2016 20:14:06 UTC All use subject to http://about.jstor.org/terms TABLE 3 PHI-COEFFICIENTS FOR PERCENTAGE CHANGE IN CERTAINTY 1950-1960, 1950-1963, 19 OF PERCENTAGE CHANGE IN CRIME RATE FOR SIX MAJOR CRI % CHANGE IN CERTAINTY % CHANGE IN CERTAINTY % 1950-1960 1950-1963 19 % Crime % Crime % Crime % Crime % Crime % Crime % Rate Rate Rate Rate Rate Rate R Change Change Change Change Change Change C 1955-1965 1959-1968 1960-1965 1955-1965 1959-1968 1963-1965 1 Homicide Robbery Assault .10 .07 -.25 .43b .19 .11 .20 .11 -.02 .10 .43b .25 -.33a .29 .38a .33a -.16 -.11 Burglary .07 .11 .25 -.02 -.15 .11 Larceny -.02 .16 .11 -.11 .16 .11 -. Auto Theft .10 .18 .24 -.26 a = p <.05. b = p <.01. This content downloaded from 137.99.31.134 on Sun, 26 Jun 2016 20:14:06 UTC All use subject to http://about.jstor.org/terms -.04 -.0 210 SOCIAL PROBLEMS TABLE 4 PHI-COEFFICIENTS FOR PERCENTAGE CHANGE IN SEVERITY 1960-1964 AND SELECTED PERIODS OF PERCENTAGE CHANGE IN CRIME RATES FOR SIX MAJOR CRIMES % Change in % Change in % Change in Severity With % Severity With % Severity With % Change in Crime Change in Crime Change in Crime Rate 1961-1965 Rate 1962-1966 Rate 1963-1967 Homicide Robbery Assault .00 .14 .28 .23 Burglary -.21 Larceny .12 Auto Theft .06 .14 .06 .19 -.03 -.21 -.21 .16 .49a .19 .12 -.09 a = p <.001. deterrence hypothesis. The data do not support a conclusion that an increase in the severity of punishment leads to a decrease in the rate of crime, or that With regard to the certainty of pun- ishment, Gibbs (1968) discovered a strong inverse relationship between certainty and rates of homicide; and Tittle reduced severity precedes an increase in (1969:409) observed "strong and con- criminality. sistent negative associations . .. between DISCUSSION Neither Gibbs' (1968) nor Tittle's (1969) findings offered strong support for the notion that severe punishment deters crime. Though both discovered a negative relationship between severity and rates of homicide, Tittle's analysis of additional offenses showed this relationship to be consistently positive. The data in this study for Time 1 generally confirm these earlier findings; however, for Time 2 some of the rela- certainty of punishment and crime rates . . ." for several offenses. Though our data for Time 2 support their findings, the data for Time 1, Time 3, and the analyses of change in certainty in- dicate that this relationship is frequently neither strong nor consistent. In fact, these data suggest that the correlation of certainty with criminality is somewhat variable over time and highly variable among offenses. Inasmuch as the present findings are somewhat inconsistent with Tittle's tionships become slightly negative and are all of low magnitude. Analysis of change in severity and change in crime cerning the certainty of punishment and ship varies both among offenses and for different time periods. ticular, one might ask why Tittle (1969:409) was able to observe such rates further indicates that this relation- Thus, the present data indicate no consistent support for the hypothesis that severe punishment deters crime. In contrast with Gibbs (1968) and Titttle (1969), this conclusion applies to homicide as well as to the other crimes examined. (1969:419-423) major findings conrates of crime, an explanation of the discrepancies seems warranted. In par- "strong and consistent negative associations . . . between certainty of punishment and crime rates . . ."16 while 16 For the six offenses in Tittle's study which correspond to the offenses studied here, Tittle (1969:415) reported the fol- lowing Tau-c coefficients between certainty This content downloaded from 137.99.31.134 on Sun, 26 Jun 2016 20:14:06 UTC All use subject to http://about.jstor.org/terms Punishment and Crime 211 this study observed weaker and more inconsistent relationships. One clue is provided by Tittle, who recognized that the denominator of his certainty index and the numerator of his deviance index were almost iden- tical. The numerator of one, and the denominator of the other, both contain 1959 and 1962 "crimes known to the ousness is that Tittle used a product- moment correlation with the random data, while he used Kendall's Tau-C with the real data. The meaning of this comparison is difficult to assess, given the diverse character of the measures and their distributions. A second problem is that Tittle used 1,000 cases for his single random correlation, whereas fifty is the maximum number of cases police" (Tittle, 1969:414-415) : 1960 + 1963 Admissions Tittle's Certainty Index - 1959 + 1962 Crimes Known 1959 ? 1962 Crimes Known/2 Tittle's Deviance Index = Population 1960 In this regard, Tittle (1969:414) was concerned "that a negative association available for computation with real data. It may be noted that with N = automatically and spuriously produced" as a consequence of their similar opera- 1,000, a product moment correlation of r = -.07 is statistically significant at p < .05. If Tittle had used the same his findings represented stronger in- number of cases in his random computations as were present in his real com- between the two indices . . . could be tional definitions. To be assured that verse relationships than might be spuri- ously produced, Tittle (1969:414) estimated the "maximum possible spurious negative association which might occur" between his two indices. This was ac- complished by assigning random digits to the terms of his certainty and devi- ance indices in such a way that the denominator of the first fraction (cer- tainty) equalled the numerator of the second fraction (deviance) for a total of 1,000 cases. A product-moment correlation coefficient was then computed at r = -.07 for "random certainty" and "random deviance." Tittle (1969: 415) concluded that "associations greater than -.07 in the data . . . represent real rather than spurious asso- ciations." One problem with this test of spuriand deviance: Assault: -.46; Larceny: -.37; Robbery: -.36; Burglary: -.31; Homicide: -.17; Auto Theft: -.08. For all but homicide and auto theft, the coefficients were statistically significant. putations, (circa 50) a correlation of r = -.27 would be required for sta- tistical significance. Again, the comparison of diverse measures based upon diverse sample sizes seems difficult to interpret. A third problem with Tittle's test of spuriousness is that it was conducted only for "total felonies," and thereby failed to take into account real differ- ences in the magnitude of appropriate random numbers for the several specific offenses. For example, the range in the real data for 1959 homicides known to the police is from 2 to 524. For burglary, however, the range of real 1959 crimes known is from 543 to 111,643. Assignments of "random crimes known" for burglary should in- volve much larger digits than those assigned to homicide, or other crimes. Finally, Tittle's test of spuriousness is based upon a single generation of random data and a single correlation co- This content downloaded from 137.99.31.134 on Sun, 26 Jun 2016 20:14:06 UTC All use subject to http://about.jstor.org/terms 212 SOCIAL PROBLEMS TABLE 5 COMPARISON OF PHI-COEFFICIENTS FOR TITTLE'S CERTAINTY AND DEVIANCE MEASURES USING REAL AND RANDOM DATA FOR SIX MAJOR CRIMES Lower Upper Mean of Std. Dev. Range of Range of 100 Phi- of Phi- 100 Phi- 100 Phi- Phi- Coefficients Coefficients Coefficients Coefficients Coefficients Using Using Using Using Using Real Random Random Random Random Data Homicide Data -.17 Data -.33 Data .14 .04 Data -.60 Robbery --.33a -.35 .17 .20 -.76 Assault -.54e -.34 .16 .04 -.68 Burglary -.38b -.37 .15 -.04 -.68 Larceny Auto -.10 Theft -.34 -.20 .15 -.35 .04 .14 .04 -.68 -.68 a=p <.05. b = p <.01. e = p <.001. efficient. Thus, it is impossible to assess the sampling distribution from which the r = -.07 was drawn. is repeated a total of 100 times for each crime category. The results of computations with random data are then com- In the present study an effort is made to compute the random spurious factor in a manner that tries to avoid the pared with recomputations of Tittle's (1969:415) original relationships using aforementioned problems. The first two association.17 problems are minimized by using the same measure of association (Phi) for both random and real data and by using 50 cases in both sets of computations. The differences among the offenses are taken into account by randomly as- signing digits that fall within the real data, and the same measure of Table 5 contains the phi-coefficients obtained when Tittle's (1969) certainty and deviance indices are correlated using real data. The table also summarizes the results of correlating Tittle's (1969) indices using random data. The mean of 100 "random phi- parameters of real data for the terms of Tittle's certainty and deviance in- coefficients" for each offense is given, as well as the standard deviation and dices. That is, Tittle's (1969:412-413) the range of the 100 correlations. From this table, it is apparent that the random operational definitions for certainty and deviance are used, and digits are randomly assigned from within the range of actual data for "Admissions to Prison," "Crimes Known to Police," and "Population" for each offense and state. The indices of "random cer- 17 That is, Tittle's (1969:412-413) orig- inal operational definitions of certainty and deviance were used, and the real data were analyzed in 2 X 2 tables, with the respective distributions divided at the median. Phicoefficients were computed for each of the six offenses, and are referred to as "real" tainty" and "random deviance" for the 50 states are then correlated for each phi-coefficients, because they were computed of the six crimes, and the entire process definitions. from real data, using Tittle's operational This content downloaded from 137.99.31.134 on Sun, 26 Jun 2016 20:14:06 UTC All use subject to http://about.jstor.org/terms Punishment and Crime 213 relationships between certainty of punishment and rates of crime vary from one random assignment of digits to an- other (columns 4 and 5) but are--on the average-appreciably higher for individual offenses (column 1) than reported by Tittle (1969:414-415) using product moment correlation (r = -.07) for total felonies. Most important, there appears to be little difference between the mean of 100 "random" phi-coefficients and the phi-coefficients based on real data. The largest (-.54) and smallest (-.10) of the real relationships fall well within the range of random relationships for the respective offenses (assault and larceny). Of the total of 600 correlations between random certainty and random deviance, only none are positive. For the one offense-assault-in which the real relationship exceeds the average random relationship, it may be noted that nine of the random phi-coefficients are greater than the apparent real correlation. From this perspective, it may be sug- gested that Tittle's (1969) findings for specific offenses probably do not exceed what could "be automatically and spuriously produced" by the similarity of terms in his certainty and deviance indices. It should be noted at this point that the operational definitions for cer- tainty and deviance used in this study are somewhat affected by the methodological problems discussed above. The indices used in this study do not have identical terms in the denominator of the certainty measure and the numerator of the deviance measure; however, crimes known to the police are so highly correlated from year to year that the same problem of correlated bias is likely to be present. This probably explains why the phi-coefficients in Table 1, though not as consistent as Tittle's, are negative in all but one instance. If it were not for this correlated bias the findings would have likely been even more sporadic and inconsistent for the various crimes at the different points in time. If it is assumed that Tittle's findings for the certainty of punishment are no greater than could be spuriously and randomly produced, and that the present findings, although not consistent with Tittle's, are still affected by the same correlated bias, one of two con- clusions may be appropriate. The most obvious conclusion might be that cer- tain and severe punishment fails to deter crime, or at best, provides a deterrent effect that varies among of- fenses and over time. For reasons to be summarized below, however, this conclusion may be inappropriate. In this regard one might at least suggest that conclusive support for deterrence theory has not been found in this research. A second possible conclusion, and one that bears directly upon the first, is that the entire question of deterring crime through punishment may be inadequately addressed through the use of available aggregate data. If, in fact, punishment does deter criminal behavior-a conclusion unsupported in this study-then it may be unreasonable to expect data from Uniform Crime Reports and National Prisoner Statistics to establish that fact.18 In the first place, this "structural" approach minimizes the possibilities of controlling for non-punitive factors affecting the crime rate (dependent variable). As noted by Gibbs (1968:517): s18 Both Gibbs (1968:522-523) and Tittle (1969:411-412) have indicated a concern for the adequacy of these kinds of data, and their conclusions are apparently tempered by acknowledged deficiencies in the available sources of information. This content downloaded from 137.99.31.134 on Sun, 26 Jun 2016 20:14:06 UTC All use subject to http://about.jstor.org/terms 214 SOCIAL PROBLEMS At least two sets of factors may jointly determine the crime rate: (1) the eti- changes in the age distribution of the population, or changes in the concen- (2) the repressive, meaning aspects of areas.21 In brief, "crimes known to the ological, meaning those extra-legal conditions which are conducive to crime; and reaction to crime which operate as deterrents. If there is any validity whatever to theories on criminality, then there are extra-legal conditions that generate high rates of crime; and whether a given type of reaction to crime has a deterrent effect cannot be determined unless these etiological conditions are held constant.19 In addition to etiological variables, any number of additional factors bear directly upon the level of crime as tration of that population in urban police" are probably sensitive to so many extra-punitive factors, that their use in measures of certainty and criminality is ill-advised for deterrence research. This sensitivity could account, in large measure, for the discovered vari- ability of relationships between Time 1, Time 2, and Time 3, reported in Table 1. represented by "crimes known to the police." The level of crime is respon- The variability of the findings, both between and within offense categories, further suggests that future research crime to the police, and the recording of might fruitfully test these relationships for other types of criminality and other sive to changes in the reporting of crime by the police. In no way can a consistency be assumed in either dimension of reporting,20 and changes in the dependent variable may be more a consequence of reporting fluctuation than a consequence of changes in the certainty and severity of punishment. Further, the level of crime may reflect 19 Tittle (1969:414-415) has apparently attempted to control for several variables with presumed etiological significance for crime rates. The relationships between certainty, severity, and deviance were compared for different levels of urbanization, age, and sex composition, and educational and industrial characteristics of the states. Little appears to have been added to an understanding of the relationships between certainty, severity, and deviance with the inclusion of the additional variables. 20 For a discussion of problems in the under-reporting of crimes to the police, see the victimization survey conducted by the National Opinion Research Center (President's Commission, 1967:17-19) wherein the extent of under-reporting and reasons for non-reporting of crimes to the police are suggested. The impact of irregularities in the reporting of crime by the police is discussed in the same volume (1967:22-24), wherein precipitous jumps in rates for specific crimes and jurisdictions may be attributed to changes in police administrative policy. periods of time. It may be that deterrence theory is more applicable to those offenses and situations in which the fear of stigma, produced by arrest and adjudication, is greatest. The "respect- able middle-class" citizen who might engage in fraud or embezzlement may be deterred by this threat, while the potential armed robber or thief, having less to lose, might be willing to ignore the probabilities of apprehension.22 For white-collar criminals, the threat of de- gradation and loss of status may be the crucial dimension of deterrence, far out-weighing the certainty of incarceration or the severity of sentence. However desirable, the expansion of this analysis to include additional offenses and time periods is all but im- possible with the existing data used here and by Gibbs and Tittle. Alternative periods of time are difficult to study 21 Crime and its Impact: An Assessment (President's Commission, 1967:24-28) also summarizes the effects of these and other sociological factors upon rates of crime. 22 For a discussion of the effects of ap- prehension on "respectable" citizens, see Cameron's (1967:115-120) study of naive shoplifters. This content downloaded from 137.99.31.134 on Sun, 26 Jun 2016 20:14:06 UTC All use subject to http://about.jstor.org/terms Punishment and Crime 215 because admissions to prison by state and offense are available in National Prisoner Statistics for only a few time periods. Alternative offenses cannot be examined because "crimes known to the police" are not specifically enumerated by Uniform Crime Reports for offenses other than the seven included in the "crime index." Finally, by using these broadly based available data, we may be ignoring a third dimension of classical deterrence theory: the celerity of punishment. Even without the problems surrounding "crimes known to the police," the necessary elimination of celerity of punishment from this type of analysis could account for our inability to show the deterrent effect of certain and severe punishment. Quite possibly, if one were to control for the speed of administering justice, then the effects of certainty and severity might be realized. With available structural data, however, such control is not possible. For these reasons, it is suggested that future empirical research on the question of deterrence should not be based upon available macroscopic data. Rather, an attempt might be made to follow a number of individual cases through the criminal justice system to determine the celerity, severity and certainty of punishment, and to assess the impact of that punishment upon be- havior of the individual and of his peers and significant others. The opera- tional definition of "punishment" might be expanded to include arrest and adjudication, inasmuch as these may prove to be as effective as incar- ceration in deterring some types of offenders. To examine the broader societal im- probabilities of apprehension, adjudication, and incarceration for different kinds of criminal acts; and (2) the severity of punishment that is likely to accompany that adjudication. Without such awareness, existing penalty structures may have little deterrent effect. Finally, future research should make distinctions between general deterrence (the impact of punishment on potential offenders in the population) and specific deterrence (the impact of punishment on the offender).23 Some of these distinctions are implicit in past research but need to be made explicit and ex- amined in more detail since the two kinds of deterrence imply widely divergent mechanisms of control. REFERENCES Ball, John C. 1955 "The deterrence concept in crim- inology and law." Journal of Criminal Law, Criminology and Police Science 46 (September-Octo- ber): 347-354. Barnes, Harry E., and Negley K. Teeters 1959 New Horizons in Criminology. Englewood Cliffs, New Jersey: Prentice-Hall, Inc. Bentham, Jeremy 1843 "Principles of penal law," in John Bowring, (ed.), The Works of Jeremy Bentham. Edinburgh: W. Tait. Bittner, Egon and Anthony Platt 1966 "The meaning of punishment." Issues in Criminology 2(Spring): 77-99. Cameron, Mary 1964 The Booster and The Snitch: De- partment Store Shoplifting. New York: The Free Press. Chambliss, William J. 1966 "The deterrent influence of punishment." Crime and Delinquency 12 (January): 70-75. Cohen, Albert K. 1965 "The sociology of the deviant act: Anomie theory and beyond." pact of punishment, one might survey a representative population to determine whether it is aware of: (1) the 23 This point was suggested by Charles Tittle. This content downloaded from 137.99.31.134 on Sun, 26 Jun 2016 20:14:06 UTC All use subject to http://about.jstor.org/terms 216 SOCIAL PROBLEMS American Sociological Review 30 (February): 5-14. oners in State and Federal Institu. 1950 Crime in the United States: Uni- 1964b National Prisoner Statistics: State Prisoners, Admissions and Re- Federal Bureau of Investigation form Crime Reports-1949. Washington, D. C.: 82-85. 1951 Crime in the United States: Uni- form Crime Reports-1950. Washington, D. C.: 79-82. 1952 Crime in the United States: Uni- form Crime Reports-1951. Washington, D. C.: 76-79. 1953 Crime in the United States: Uni- form Crime Reports-1952. Washington, D. C.: 78-81. 1956 Crime in the United States: Uni- form Crime Reports-1955. Washington, D. C.: 77-80. 1960 Crime in the United States: Uni- form Crime Reports-1959. Washington, D. C.: 38-51. 1961 Crime in the United States: Uni- form Crime Reports-1960. Wash- ington, D. C.: 38-52. 1962 Crime in the United States: Uni- form Crime Reports-1961. Washington, D. C.: 38-52. 1963 Crime in the United States: Uniform Crime Reports-1962. Washington, D. C.: 38-52. 1964 Crime in the United States: Uni- form Crime Reports-1963. Washington, D. C.: 52-66. 1965 Crime in the United States: Uni- form Crime Reports-1964. Washington, D. C.: 54-59. 1966 Crime in the United States: Uniform Crime Reports-1965. Washington, D. C.: 56-70. 1967 Crime in the United States: Uniform Crime Reports-1966. Washington, D. C.: 66-67. 1968 Crime in the United States: Uniform Crime Reports-1967. Washington, D. C.: 68-79. 1969 Crime in the United States: Uni- form Crime Reports-1968. Washington, D. C.: 66-75. Federal Bureau of Prisons 1954 National Prisoner Statistics: Prisoners in State and Federal Insti- tutions, 1950. Washington, D. C.: 72-74. 1960 National Prisoner Statistics: Characteristics of State Prisoners, 1960. Washington, D. C.: 41 and 69. 1964a National Prisoner Statistics: Pris. tions, 1963. Washington, D. C.: 22-24. leases, 1964. Washington, D. C.: 52. Gibbs, Jack P. 1966 "Sanctions." Social Problems 14 (Fall): 147-159. 1968 "Crime, punishment, and deter- rence." Southwestern Social Sci- ence Quarterly 48(March): 515- 530. Homans, George C. 1961 Social Behavior: Its Elementary Forms. New York: Harcourt. Horton, Paul B. and Gerald R. Leslie 1965 The Sociology of Social Problems. New York: Appleton-Century-Crofts. Jeffrey, C. Ray 1965 "Criminal behavior and learning theory." Journal of Criminal Law, Criminology and Police Science 56 (September) 294-300. Jenson, Gary F. 1969 "'Crime doesn't pay': Correlates of a shared misunderstanding." Social Problems 17(Fall): 189201. Mead, George H. 1918 "The psychology of punitive justice." American Journal of Sociology 23(March): 577-602. President's Commission on Law Enforcement and Administration of Justice 1967 Task Force Report: Crime and Its Impact-An Assessment. Washington: U.S. Government Printing Office. Polier, Justine W. 1956 "The woodshed is no answer." Federal Probation 20(September): 3-6. Savitz, Leonard D. 1958 "A study in capital punishment." Journal of Criminal Law, Criminology and Police Science 49(November-December) 338-341. Schuessler, Karl 1952 "The deterrent influence of the death penalty." The Annals 284 (November): 54-62. Sellin, Thorsten 1967 "Homicides in retentionist and This content downloaded from 137.99.31.134 on Sun, 26 Jun 2016 20:14:06 UTC All use subject to http://about.jstor.org/terms Punishment and Crime 217 abolitionist states," in Thorsten Sellin, (ed.), Capital Punishment. New York: Harper and Row: 135138. Tittle, Charles R. 1969 "Crime rates and legal sanctions." Social Problems 16(Spring): 409423. Toby, Jackson 1964 "Is punishment necessary?" Jour- and Police Science 55 (September): 332-337. Wood, Ledger 1938 "Responsibility and punishment." Journal of Criminal Law, Criminology and Police Science 28 (January-February): 630-640. Zimring, Frank and Gordon Hawkins 1968 "Deterrence and marginal groups." Journal of Research in Crime and Delinquency 5(July): 100-114. nal of Criminal Law, Criminology BEING DISCOVERED: A STUDY OF HOMOSEXUALS IN THE MILITARY COLIN J. WILLIAMS and MARTIN S. WEINBERG Indiana University Current conceptions of deviance have placed less emphasis on the role played by the deviant himself in being assigned a deviant label. To examine the role of the deviant, a sample of male homosexuals who had received less than honorable discharges from the military for homosexual conduct (LHD group) were compared with a sample who had received honorable discharges (HD). It was found that, compared to the HD group, the LHD group was more likely (a) to have been engaging in homosexual sex more frequently before induction, (b) to have experienced more frequent homosexual sex while in the service, and (c) to have had predominately other servicemen as homosexual partners. Further examination of the data showed how these factors specified the manner in which members of the LHD group were discovered. Current among sociological conceptions of deviance is an approach that concerns itself less with the attributes of for both labellers and labelled. How- ever, despite the emphasis laid on devi- ance as a product of interaction, in the person or persons said to have violated a social rule than with the char- practice most attention has been paid acter of the reactions of other persons to Social control agencies have been especially focused on; most of the findings have documented the discrepancies between formal statements of procedures and actual operating procedures and the immense power of control agencies in the social production of deviants (cf. these attributes and events. This approach, sometimes called the "labelling" approach to deviance, sees the deviant as a social product, the outcome of interaction sequences between labellers and labelled ) cf. Becker, 1964a). The questions that are raised by this ap- proach thus concern the behaviors that are labelled as deviant, what the processes are by which the labels are successfully applied or avoided, and what the consequences of such processes are to the labeller's role in this process. Scheff, 1964; Sudnow, 1965; Chambliss and Liell, 1966). The emphasis on labelling, rather than the attributes of deviants, has been rich in its contribution to a better understanding of deviance as a social This content downloaded from 137.99.31.134 on Sun, 26 Jun 2016 20:14:06 UTC All use subject to http://about.jstor.org/terms