Methodology

Notes on figures 79-82: In these graphs, assault includes threats; burglary does not include attempts; and auto theft excludes thefts of motorcycles.

Chart data - in spreadsheets
	Figure 79			Figure 80
	Robbery			Assault
Year	United States	England		United States	England
1988	1.9	0.7		5.4	1.9
1989
1990
1991	1.5	1.1		4.7	3.8
1992
1993
1994
1995	1.3	1.4		5.7	5.9
	Figure 81			Figure 82
	Burglary			Auto theft
Year	United States	England		United States	England
1988	3.8	2.1		2.1	1.9
1989
1990
1991	3.1	3.0		2.6	3.7
1992
1993
1994
1995	2.6	3.0		1.9	2.5

England (including Wales) has higher crime rates than the United States according to 1995 national crime victimization surveys. Are English crime rates higher than U.S. crime rates according to 1995 international crime victimization surveys?

Earlier in the report, U.S. and English crime rates were compared based on data compiled from national crime victimization surveys in the two countries. English crime rates for 1995 were found to be higher than U.S. crime rates for all four crime categories compared (robbery, assault, burglary, and motor vehicle theft) (figures 1-4). However, the two national crime victimization surveys (called the "British Crime Survey" in England and the "National Crime Victimization Survey" in the United States) were not specifically designed to be comparable to one another. For example, in the British Crime Survey, sampled respondents are asked about victimizations that occurred over a 12-month period; in the U.S. National Crime Victimization Survey they are asked about crimes over a 6-month reference period. Also, the British Crime Survey does not use bounded interviews (a technique for minimizing the tendency of respondents to unintentionally inflate the number of victimizations that occurred in the reference period by mentioning crimes that occurred before the reference period) whereas the U.S. National Crime Victimization Survey does use bounded interviews.

In view of such design differences between the two surveys, it is reasonable to ask whether England's higher 1995 survey crime rates might merely reflect counting differences rather than real differences in criminal victimization. One reason for doubting that possibility is that the design differences do not all operate to make England's crime problem appear worse than America's. For example, two counting differences are England's 12-month reference period (versus America's 6) and England's unbounded interviews (versus America's bounded ones). While England's unbounded interviews do artificially make England crime rates look worse than America's, England's 12-month reference period operates in the other direction, artificially making America's crime situation look worse than England's. That is because people have an easier time remembering crimes that occurred during the last 6 months than the last 12.

The question of whether England's higher 1995 survey crime rates are possibly attributable to different data collection procedures is perhaps best answered by comparing crime rates from victimization surveys that are more nearly identical in design. Such surveys exist. They are called the international crime victimization surveys and are completely distinct from the national crime victimization surveys. Three international crime victimization surveys have been conducted in England (including Wales) and the United States: one covering crimes committed in 1988, one covering those committed in 1991, and one covering those in 1995 (Mayhew and van Dijk, 1997, page 34 and table 1 of appendix 4). Nearly the same years (1987, 1991, and 1995) were surveyed in the two countries' national crime victimization surveys.

International surveys have a major shortcoming not present in national surveys: their small samples and low response rates. However, the major strength of the international crime victimization surveys is that they were specifically designed for international comparisons. Design differences that hinder comparisons between the national surveys are not present in the international surveys. Consequently, if results from international surveys agree with results from national surveys, confidence is enhanced that findings from the national surveys are not merely a reflection of their major weaknesses for international comparisons.

Turning now to the international crime victimization surveys, results largely parallel those from the national surveys in three respects:

First, the international survey (figures 79-82) and the national survey (figures 1-4) both indicated that England had a higher 1995 victimization rate than the United States for each type of crime (robbery, assault, burglary, motor vehicle theft).

Second, the national and the international surveys also showed some similarities in trends since around 1987. According to both, England did not have generally higher victimization rates than the United States in the late 1980's, but by 1995 England's rates had all risen to levels exceeding those in America.

Third, the country that had the higher victimization rate according to the national survey was typically the country that had the higher rate according to the international survey. For example, both surveys agree that the robbery rate during the late 1980's was higher in the United States than in England; the 1991 U.S. robbery rate was higher than the English rate; and the 1995 English robbery rate was higher than the U.S. rate. More generally, of the 12 comparisons that could be made (the 3 for robbery plus 3 for assault, 3 for burglary, and 3 for auto theft), 8 agreed on which country had the higher victimization rate. There was perfect agreement for robbery and mostly agreement for burglary and auto theft, although mostly disagreement for assault.

To the top

Sources of American data

American survey offenses

The number of victim-survey offenses, comparable household and population figures (numbers of persons aged 12 or over), and the probability of reporting to the police were obtained from the National Crime Victimization Survey (NCVS). For example, the NCVS estimated that there were 5,482,720 household burglaries in 1994 and that 51% of them were reported to the police (BJS, 1997, tables 1 and 91). Since there were an estimated 100,807,650 households in 1994, the survey burglary rate was 54.4 per 1,000 households; about one in every 18 households was burglarized in 1994 (disregarding repeat victims).

The NCVS was redesigned in 1992 to encourage more complete recall of crimes, especially those committed by intimates or family members (see Perkins and others, 1996, Appendix II). More detailed screening questions were used, with more explicit specification of different kinds of acts. To assess the effect of the redesign, half of the 1992 interviews were carried out with the old questionnaire format, and half were carried out with the new questionnaire format. It was concluded that the redesign led to an increase of about 20% in the estimated number of burglaries and of about 24% in the estimated number of aggravated assaults, but had no significant effect on the estimated number of robberies or motor vehicle thefts, as revealed by victims (Rand, Lynch, and Cantor, 1997). The redesign had a much greater effect on the estimated number of rapes (up 157%); however, victim-survey rapes are not included in this report, because there is no comparable victim-survey estimate of rape for England.

The redesign affected not only estimates of the frequency of aggravated assault and burglary, but also estimates of the "percent reported to police" for these two crimes. Under the old questionnaire format, 62% of aggravated assaults were reported to police in 1992, versus 55% under the new format. For burglary, 54% were reported under the old; 51% under the new. In both cases, percent reported to police declined, possibly because the biggest effect of the redesign was to stimulate recall of less serious forms of these two crimes. Being less serious forms, they are less likely to be reported to police.

As a result of the redesign, NCVS estimates (both frequency and percent reported to police) of aggravated assault and burglary for 1992 and beyond are not comparable to pre-1992 estimates. Consequently, pre-1992 NCVS estimates of frequency and "percent reported to police" for aggravated assault and burglary (but not for robberies or motor vehicle thefts) were back-adjusted to make the pre-1992 figures comparable to post-1991 figures. For example, every pre-1992 burglary estimate was increased by 20%. Similar adjustments were made to the numbers of offenses reported to the police.

For comparability with the BCS, only completed motor vehicle thefts in the NCVS were counted. All burglaries, robberies, and aggravated assaults were counted.

The average number of offenders per offense provides the link between offenses and offenders. This was obtained from three sources. For robbery, rape, and aggravated assault, the figures were estimated from the table routinely published in the NCVS (for example, BJS, 1992a, table 70). For burglary and vehicle theft, the figures were estimated from unpublished NCVS data. The homicide figures were estimated from the FBI's Supplemental Homicide Reports. Since the average number of offenders per offense was often based on small numbers but could greatly affect probability estimates, the average figure over all years (1981-94) was used in all calculations. This was 1.4 for burglary, 1.5 for motor vehicle theft, 1.8 for robbery, 1.5 for aggravated assault, 1.2 for rape, and 1.15 for homicide.

To the top

American police-recorded offenses

The numbers of police-recorded crimes were obtained from the 1995 UCR (FBI, 1996, table 1) and the 1996 UCR (FBI, 1997, table 1), which show the latest estimates of the numbers of index crimes committed in each year from 1981 to 1996. The UCR figures for forcible rape are restricted to female victims. The latest estimates of the resident population in each year were obtained from the Bureau of the Census (1996, table 14). For example, there were 2,712,800 burglaries of all kinds in 1994. Since the resident population was estimated to be 260,341,000 in 1994, the police-recorded burglary rate was 10.4 burglaries per 1,000 persons.

The numbers of police-recorded offenses that were comparable to survey offenses were estimated in various ways. For burglary, the estimated proportion of residential as opposed to non-residential burglaries was used (for example, FBI, 1995, table 23). Since this was close to two-thirds in every year, its average value in 1981-94 of .67 was used for each year's calculations. Thus, in 1994, the number of comparable police-recorded burglaries was estimated to be 1,817,576 (2,712,800 x .67) For aggravated assault, the police-recorded figure was reduced by 5% to take account of the estimated proportion of assaults on victims under age 12 (estimated from 1993 NIBRS data by Jarvis, 1994) and hence excluded from NCVS.

For robbery, the estimated proportion of offenses with personal victims was used (for example, FBI, 1995, table 23). On average in 1981-94, 78% of robberies were against persons (street, residential, miscellaneous) rather than businesses (banks, gas stations, convenience stores, commercial); this fraction was used for each year's calculations. The number of robberies was further reduced by 2% to take account of the estimated proportion of robbery victims under age 12 (Jarvis, 1994). It was then increased by 11% to convert the number of incidents to the number of victimizations, because the UCR counts robbery incidents while the NCVS counts robbery victimizations (that is, one victim = one offense). The regular table in the NCVS (for example, BJS, 1992a, table 56) shows an average ratio of victims to incidents of 1.11 in 1981-94.

For motor vehicle theft, it was assumed that the proportion of completed offenses in the UCR figures was the same as in offenses reported to the police by victims. For example, according to the NCVS in 1990 (BJS, 1992a, table 101), 94.8% of completed vehicle thefts were reported to the police; 42.5% of attempted vehicle thefts were also reported to the police. It was determined that 78.7% of reported vehicle thefts were completed as opposed to attempted. Over all years, the average was 80%, and there was no sign of a trend in these percentages, so this figure was used in all cases. Applying 80% to the number of police-recorded vehicle thefts (1,635,900) yielded an estimate of 1,308,720 police-recorded completed thefts in 1990. This figure was further reduced by 17% (to 1,086,238) to take account of the proportion of vehicle thefts involving business as opposed to personal victims (Biderman and Lynch, 1991).

The number of police-recorded offenses that were comparable to survey offenses makes it possible to estimate the probability of a survey offense being recorded by the police. For example, for burglary in 1994, comparing 5,482,720 survey offenses to 1,817,576 comparable recorded offenses yields the estimate that the police were recording 33% of all burglaries.

It is similarly possible to estimate the probability that a survey offense reported to the police would be recorded by the police. For burglary in 1994, comparing 2,796,187 offenses (5,482,720 x .51) reported to the police with 1,817,576 comparable recorded offenses yields the estimate that the police were recording 65% of offenses reported to them. This probability was slightly greater than 1 for vehicle theft in 1986, presumably reflecting the fact that the estimated number of vehicle thefts varies within a confidence interval and that almost all that are reported to the police are recorded by the police. All probabilities estimated as greater than 1.0 were set to 1.

To the top

American convictions

In the United States, the total number of convictions is the sum of State convictions plus Federal convictions plus juvenile convictions. For example, for burglary in 1994, there were an estimated 161,975 convictions (98,109 State; 146 Federal; 63,720 juvenile). The conviction rate per 1,000 population was calculated by dividing this by the resident population of persons aged 10 or over (since few persons under 10 are convicted); this was 221,890,000 in 1994 (Bureau of the Census, 1996, table 14). Hence the conviction rate for burglary in 1994 was 0.73 per 1,000 population; disregarding repeat offenders, about one in every 1,400 persons was convicted of burglary.

Federal convictions 1981-1994

Data on Federal convictions were obtained from the BJS Federal Justice Statistics Program (FJSP) (for 1981-1988: BJS, 1992b, 1993, table 9) (for 1990: Langan, Perkins and Chaiken, 1994) (for 1992: Langan, 1996) (for 1994: Langan and Brown, 1997b).

State convictions 1986-1994

State court convictions were obtained from the National Judicial Reporting Program (NJRP) (for 1986: Langan, 1989) (for 1988: Langan and Dawson, 1990) (for 1990: Langan, Perkins and Chaiken, 1994) (for 1992: Langan, 1996) (for 1994: Langan and Brown, 1997b). Motor vehicle theft was not shown separately in 1986 and 1988, so it was estimated as 18% of larceny convictions, since 18% of larceny convictions in 1990-94 were for motor vehicle theft. The number of rape convictions in 1986 was estimated using the 1988 figure because the figure from the 1986 NJRP survey was thought to be unreliable. The rape convictions include female offenders, but consistently 99% of convicted offenders were male (Langan, 1989: Langan and Dawson, 1990; Langan and Dawson, 1993; Langan and Graziadei, 1995; Langan and Brown, 1997a).

State convictions 1981 and 1983

Since the NJRP only began in 1986, it was necessary to back-estimate the number of convictions in State courts in 1981 and 1983. Previously (Farrington, Langan, and Wikström, 1994), this was achieved by assuming that the change in the number of adults convicted for a particular crime was the same as the change in the number of adults arrested for that crime. However, an estimation method based on prison admissions proved to be more defensible empirically.

In choosing between different estimation methods, the main criterion was how well they reproduced known changes in the number of convictions in State courts in 1988-94. The first NJRP survey, in 1986 (Langan, 1989), was based on a random sample of 100 counties, whereas all subsequent NJRP surveys have been based on a random sample of 300 counties. Hence, it was thought that changes in convictions in 1988-94 would be more reliable than changes in convictions in 1986-94. In 1988, a total of 218,096 persons were convicted in state courts of murder/nonnegligent manslaughter, rape, robbery, aggravated assault, burglary or motor vehicle theft (Langan and Dawson, 1990, table 1); the corresponding figure in 1994 was 262,435, an increase of 20% (Langan and Brown, 1997a, table 1).

The number of adults (age 18 or over) arrested for these six offenses, according to the UCR, was 965,462 in 1988 and 993,121 in 1994, an increase of 3% (FBI, 1989, 1995). However, the number of new admissions to State prisons (new court commitments plus violators returned with new sentences) for these six offenses was 123,706 in 1988 (obtained by applying the 1988 offense distribution of new court commitments to the 1988 total number of new admissions) and 144,940 in 1994 (obtained by applying the 1994 offense distribution of new court commitments to the 1994 total number of new admissions), an increase of 17% (Perkins and Gilliard, 1992, table 1-5; BJS, 1991, tables 5.10a and 5.16; Brown and others, 1996, tables 5.10a and 5.16; BJS, 1994, table 1-5). Hence, the increase in the number of new admissions to State prisons corresponded closely to the increase in the number of convictions in State courts; the number of new admissions was 57% of the number of convictions in 1988 and 55% in 1994. Also, the number of new admissions to State prisons according to State prison records (as compiled through the National Corrections Reporting Program and National Prisoner Statistics) was similar to the number of persons sentenced to prison according to State court records (as compiled in the NJRP), which was 125,471 in 1988 and 152,378 in 1994, an increase of 21% (Langan and Dawson, 1990; Langan and Brown, 1997a).

The numbers of new admissions to State prisons (new court commitments plus violators returned with new sentences) for 1981 and 1983 were obtained by applying the offense distribution of new court commitments (Greenfeld and Minor-Harper, 1984, table 3; Beck and Hester, 1986, table 3) to the total number of new admissions to State prisons (BJS, 1983; BJS, 1986). For example, there were 164,857 new admissions in 1981, 27.2% of which were for burglary, leading to an estimate of 44,841 burglars admitted. Since the corresponding figure for 1988 was 49,539, about 10.5% more burglars were admitted in 1988 than in 1981. Since the number of convictions for burglary in State courts in 1988 was 101,050, it was estimated that the corresponding number of convictions in 1981 was 91,467. All estimates of convictions in State courts in 1981 and 1983 were similarly based on changes in numbers of sentenced persons admitted to State prisons.

Estimates of juvenile convictions 1986-94

The number of juvenile convictions was defined as the number of petitioned juvenile court cases resulting in an adjudication for which the juvenile received a disposition other than transfer/waiver (to the adult court) or dismissal/release. (Transferred cases were excluded because they would appear in State conviction figures.) Two types of data were used to estimate the number of juvenile convictions by offense: the nationwide number of referrals to juvenile court, by offense; and the percentage of referrals that resulted in conviction. The source for the first was the annual publication Juvenile Court Statistics (JCS). The source for the second was a database maintained by the National Center for Juvenile Justice (Snyder, 1995; Butts, 1996). Multiplying the number of juvenile court referrals by the percentage that resulted in conviction gives the number of juvenile convictions. For example, there were an estimated 141,600 burglary cases in 1994 (Butts et al., 1996, table 1), of which 45% (Butts, 1996) were adjudicated and received placement, probation or another disposition (not including transfer or dismissal), leading to an estimated 63,720 juvenile burglary convictions.

Estimates of juvenile convictions for 1981 and 1983

Data on the fractions of cases adjudicated and receiving various dispositions were not available before 1985. Previously (Farrington, Langan and Wikström, 1994), the numbers of juvenile convictions in 1981 were estimated from the numbers of delinquency cases in 1981 (Snyder, Finnegan and Hutzler, 1983, table 2), assuming that the fractions were the same as in 1986. However, the present estimates were based on the correspondence between juvenile arrests (under 18) as a percentage of all arrests, and juvenile convictions as a percentage of all convictions. For example, for burglary in 1986-94, the fraction of arrests involving juveniles (34%) was consistently about 9 percentage points lower than the fraction of convictions that were in the juvenile court (43%). Since the fraction of burglary arrests involving juveniles was 43% in 1981 (FBI, 1982, table 31), it was assumed that the fraction of burglary convictions that were in the juvenile court in 1981 was 52%. Since the total number of State and Federal convictions for burglary in 1981 was 91,606, the estimated number of juvenile convictions was 99,240 (91,606 x 52/48).

Linking the number of convictions to the total number of alleged offenders

The average number of offenders per offense provides the link between offenses and offenders. Since there were an estimated 5,482,720 residential burglaries in 1994, and an average of 1.4 offenders per offense (the average from 1981 through 1994), there were about 7,675,808 offenders (not necessarily different persons) at risk of conviction for residential burglary. Since residential burglaries comprise 67% of all burglaries, it was estimated that there were about 11,456,430 offenders at risk of conviction for burglaries of all types. Dividing this by the number of persons convicted (161,975) yields the estimate that there were 71 offenders per conviction, or that the average burglar could commit about 71 burglaries per one conviction for burglary. The corresponding probability of conviction for each burglary offender was .014, or 14 convictions per 1,000 alleged offenders.

The calculations for rape and homicide were based only on police-recorded offenses. For example, for rape in 1994, there were 102,220 recorded offenses and 1.2 offenders per offense (1.2 was the 1981-1994 NCVS aver-age), yielding an estimated 122,664 offenders at risk of conviction for rape. Dividing this by the number of persons convicted of rape (23,047) yields the estimate that there were about 5 offenders per conviction, that the probability of an offender being convicted was .19, or that there were 188 convictions per 1,000 alleged offenders.

To the top

American probability of incarceration

In the United States, the total number of incarceration sentences is the sum of State plus Federal plus juvenile sentences. For example, for burglary in 1994, 96,369 offenders were sentenced to incarceration (73,582 State; 131 Federal; 22,656 juvenile). The population incarceration rate per 1,000 population was calculated by dividing this by the resident population of persons aged 10 or over (221,890,000 in 1994). Hence, the incarceration rate for burglary in 1994 was 0.43 persons incarcerated per 1,000 population; about 1 in every 2,300 persons was sentenced to incarceration for burglary.

The probability of incarceration following a conviction was calculated by dividing the number of persons sentenced to incarceration by the number of persons convicted. For burglary in 1994, this was .59 (96,369/161,975). The probability of an offender receiving a custodial sentence was calculated by multiplying the probability of an offender being convicted (0.014) by the probability of a conviction being followed by incarceration (0.59). For burglary in 1994, this was .008 (.014 x 0.59) corresponding to 8 in every 1,000 alleged burglars receiving a custodial sentence.

Federal incarceration sentences 1981-94

Data on Federal sentences were obtained from the BJS Federal Justice Statistics Program (for 1981 and 1983: 1982 figures are used from BJS, 1993) (for 1986 and 1988: BJS, 1996) (for 1990: Langan, Perkins and Chaiken, 1994) (for 1992: Langan, 1996) (for 1994: Langan and Brown, 1997b).

State incarceration sentences 1986-94

Data on State incarceration sentences were obtained from the BJS National Judicial Reporting Program (for 1986: Langan, 1989) (for 1988: Langan and Dawson, 1990) (for 1990: Langan, Perkins and Chaiken, 1994) (for 1992: Langan, 1996) (for 1994: Langan and Brown, 1997b). Note, however, that no 1988 NJRP data exist for motor vehicle theft. For 1988 motor vehicle theft, percentages receiving prison and jail used 1988 larceny figures.

State incarceration sentences 1981-83

State courts in the United States impose two types of incarceration sentences: prison and jail. For the years in which NJRP data were not available (1981 and 1983), it was necessary to estimate the percentage of convicted persons receiving prison and the percentage receiving jail. The year with NJRP data closest to 1981 and 1983 is 1986, but 1986 NJRP data are less reliable than 1988 NJRP data because the former was based on a sample of 100 counties whereas the latter was based on 300 counties. Therefore, 1981 and 1983 percentages used 1988 percentages. Prison and jail percentages have been stable over the years. Therefore, using 1988 figures for 1981 and 1983 seemed acceptable. Applying 1988 prison and jail percentages to estimated 1981 and 1983 convictions gave the estimated number of persons sentenced to incarceration those years. Note, however, that no 1988 NJRP data exist for motor vehicle theft. For motor vehicle theft in 1981 and 1983, the percentages receiving prison and jail used 1988 larceny figures. Such percentages do not differ much between larceny and motor vehicle theft, according to NJRP data for years in which the two could be compared.

Juvenile incarceration sentences 1986-94

As the term was used here, convicted juveniles were juveniles who were petitioned to the juvenile court where they were then adjudicated and finally disposed of through placement, probation, or another disposition (not including transfer or dismissal). The number of juvenile convictions multiplied by the percentage that received placement gives the number of juveniles sentenced to incarceration. Percentages receiving different dispositions were obtained from a database maintained by the National Center for Juvenile Justice (Snyder, 1995; Butts, 1996). For example, for burglary in 1994, 36% of convicted juveniles were given a placement.

Juvenile incarceration sentences 1981 and 1983

Data on the percentage of convicted juveniles sentenced to incarceration (defined as placement) do not go back before 1985. Estimates, by offense, were taken from the average percentage from 1986 to 1994.

To the top

American sentence length, time served, and days at risk of serving

Average sentence length was estimated by combining State, Federal, and juvenile figures. For example, for burglary in 1994, the average prison sentence length was 69 months in State courts (51,998 offenders) and 60 months in Federal courts (116 offenders). The average jail sentence length was 7 months in State courts (21,584 offenders) and 9 months in Federal courts (15 offenders). The average juvenile confinement sentence was 18 months (22,656 offenders). The weighted average of all these figures was 43 months.

Similarly, average time served was estimated by combining the State, Federal, and juvenile figures. For example, burglars sentenced to State prison in 1994 would be expected to serve 39% of their sentence, assuming they would serve the same percentage of their sentence that State prisoners released in 1994 served, a figure that includes time served in jail. Burglars sentenced to jail in State courts in 1994 would be expected to serve 50% of their sentence (the figure that was used for jail sentences for all offenses and for all years, since no national data exist on percentage of jail sentence served). Incarcerated Federal burglars would be expected to serve 85% of their prison sentence and 100% of their jail sentence (since Federal law requires that 100% of Federal jail sentences and a minimum of 85% of Federal prison sentences be served). Juvenile burglars sentenced to incarceration would be expected to serve 50% of their sentence (the figure that was used for juvenile incarceration sentences for all offenses and for all years, since no national data exist on percentage of juvenile incarceration sentence served). Using the preceding figures on incarceration sentences, the overall percentage of time served for burglary was 42%. Consequently, the average time served for burglary in 1994 was estimated to be 18 months (since 42% of 43 months -- the overall average incarceration sentence for burglary -- is 18 months).

"Days of incarceration an offender risks serving" was obtained by multiplying the probability of conviction given an offense by the probability of incarceration given conviction and by the average number of days served per incarceration sentence. Thus, for burglary in 1994, this was 4.6 days (.014 x .59 x 548 days = 4.6 days when unrounded figures are used).

Federal estimates of incarceration sentence length and time served

This report calls Federal sentences of 1 year or less "jail," and Federal sentences over a year "prison." The distinction between Federal prison and jail sentences is important for estimating time served for persons sentenced in 1988 and beyond because a Federal law, which took effect in 1987, set different release policies for prison and jail sentences. Data on Federal sentence lengths were from the FJSP (for 1986-1988: BJS, 1993) (for 1990: Langan, Perkins and Chaiken, 1994) (for 1992: Langan, 1996) (for 1994: Langan and Brown, 1997b). For 1981 and 1983, sentence lengths given for 1982 were used (BJS, 1993).

The average time served for prison sentences given in Federal courts before 1987 was estimated from McDonald and Carlson's (1992) report showing that violent offenders served 54% of their sentences and property offenders served 66%.

For 1988-94 the fraction of time served for prison sentences was assumed to be 85%, since this fraction, set by a Federal law that took effect in 1987, is the minimum time that must be served. The average time served for jail sentences given in Federal court was assumed to be 100% for 1988-94, since the same Federal law required all jail sentences to be served in full. For Federal defendants sentenced from 1990 to 1994, the percentage of their incarceration sentence that they were expected to serve was estimated by taking a weighted average of the number expected to serve 85% (that is, the number sentenced to Federal prison) and the number expected to serve 100% (that is, the number sentenced to jail). For example, Federal courts in 1994 sentenced 116 burglars to prison and 15 to jail. The 116 were expected to serve 85% of their prison sentence, and the 15 were expected to serve 100% of their jail sentence. The weighted average for the 131 incarcerated burglars is 87%, the percentage of Federal incarceration sentences that burglars were expected to serve. Because data were not available, the preceding method could not be used to estimate the percentage of incarceration sentences that Federal defendants sentenced in 1988 were expected to serve. For 1988, the percentage was estimated as the average from 1990 to 1994.

State estimates of incarceration sentence length and time served 1981-1983

For 1981 State prison sentences, estimates used 1982 averages from the National Corrections Reporting Program (Minor-Harper and Greenfeld, 1985). The 1982 report was used because the 1981 report (Greenfeld and Minor-Harper, 1984) only provided median sentence lengths, not means. For 1983 prison sentences, estimates were from the 1983 National Corrections Reporting Program (Beck and Hester, 1986). The murder/nonnegligent manslaughter figures for 1981 and 1983 were weighted averages of murder (.67) and manslaughter (.33), reflecting the relative frequencies of these cases among State prison admissions. Average lengths of jail sentences given in State courts in 1981 and 1983 were assumed to be the same as NJRP published figures for 1988 (Langan and Dawson, 1990). Average jail sentence lengths for motor vehicle theft in 1981 and 1983 used 1988 sentences for larceny.

Data from the National Corrections Reporting Program (NCRP) were the source for 1981 and 1983 estimates of percentage of sentence served before release. For 1981, percentages were from 1981 NCRP data. Published 1981 figures on percentage of sentence served (Greenfeld and Minor-Harper, 1984) were adjusted upward to correct for the fact that they included time in prison before release but not credited jail time. For the murder adjustment, 5 percentage points were added; for rape, 6; robbery, 6; assault, 9; burglary, 8; motor vehicle theft, 11. The basis for the adjustment was results of an analysis of 1992-1994 NCRP data. For 1983, percentages used 1984 NCRP data on time served (Minor-Harper and Innes, 1987, table 4). This report was used because it provides time served in prison and jail for persons released from prison.

State estimates of incarceration sentence length and time served 1986-1994

With few exceptions, estimates of incarceration (prison and jail) sentence length from 1986 to 1994 are from the BJS National Judicial Reporting Program (for 1986: Langan, 1989) (for 1988: Langan and Dawson, 1990) (for 1990: Langan, Perkins and Chaiken, 1994) (for 1992: Langan, 1996) (for 1994: Langan and Brown, 1997b). For 1986 and 1988, jail sentences for motor vehicle theft use published larceny figures. For 1986 and 1988, the average lengths of prison sentences for motor vehicle theft are from the National Corrections Reporting Program (for 1986: Perkins, 1992) (for 1988: Perkins and Gilliard, 1992).

The average time served for prison sentences given in State courts was estimated using NCRP data on average sentence length and average time served (time served in prison plus credited jail time) for first releases from State prison. Time served divided by sentence length gives percentage of sentence served. In the case of burglary in 1994, 5 months jail time was added to the average 23 months prison time served by released prisoners, making an average 28 months served, or 39% of the average sentence(28/72). This fraction of the sentence was then applied to the average 69 month prison sentence given in State courts for burglary in 1994 (Langan and Brown, 1997a, table 3) to produce an expected time to be served of 27 months.

Time-served percentages for State prison sentences were from the National Corrections Reporting Program for all years (for 1994: Langan and Brown, 1997b) (for 1992: Langan, 1996). For 1990, time-served percentages were computed by summing time served in prison before release (Perkins, 1993) plus time served in jail (based on the approximate 1992-1994 average of 10 months for murder; 6 months for all other offenses), and dividing by the maximum sentence length (table 2-7). For 1988, the steps were the same as in 1990 but the data source was different (Perkins and Gilliard, 1992, table 2-7). For 1986, the steps were the same as in 1988 but the data source was different (Perkins, 1992, table 2-6).

The average time served for jail sentences given in State courts is not known, but it was assumed to be half the sentence length (as was found in California by Petersilia, Turner and Peterson, 1986, page 13).

Juvenile estimates of incarceration sentence length and time served

There is no published source of national data on sentence length and time served by juveniles for different offenses. The only nationally representative survey of youth in long-term State juvenile institutions was carried out in 1987. This reported that the average time currently served by these inmates was 7.2 months for violent offenders and 5.0 months for nonviolent offenders (Beck, Kline and Greenfeld, 1988, table 5). Assuming that youth were surveyed at a random point in their sentences, the total time served would be 14.4 months for violent offenders and 10.0 months for nonviolent offenders. These estimates are not too dissimilar from the total length of stay figures published by the National Juvenile Corrections System Reporting Program, based on information from State juvenile correctional agencies: 385 days for person offenses and 250 days for property offenses (DeComo and others, 1995, table 4-15).

Offense-specific data were obtained from the National Juvenile Corrections System Reporting Program (Cohen, 1997). These data show the average lengths of stay for juveniles released in 1991 from State juvenile institutions: homicide 839 days, rape 655 days, robbery 386 days, aggravated assault 389 days, burglary 265 days, and motor vehicle theft 229 days. How much time these juveniles spent in aftercare supervision following their release is unknown, but their overall sentence length can be thought of as their time in confinement plus their time in aftercare. Assuming that their length of time in confinement equals their length of time in aftercare, the average juvenile sentence for murder in 1991 was 1,678 days (2 x 839); for rape, 1,310 days (2 x 655); for robbery, 772 days (2 x 386); for assault, 778 days (2 x 389); for burglary, 530 days (2 x 265); for motor vehicle theft, 458 days (2 x 229). For juveniles, therefore, their time-served percentage would be 50% of their sentence.

The preceding 1991 figures on juvenile sentence length, time served, and percentage of time served, were used in all years. The justification for using the 50% figure was results of an informal survey of four States, which indicated that the time spent in aftercare was roughly comparable to the time spent in confinement.

More adequate figures, collected on an annual basis, are clearly needed both on length of time in juvenile confinement and on length of stay in aftercare.

To the top

American homicide

Special procedures were used to estimate murder sentence lengths and time served for homicide (murder and nonnegligent manslaughter) because of the large number of life sentences. Similar allowances were not made for other offenses, because of their infrequency. For example, in 1994, whereas 23.7% of homicide cases in State courts were given life sentences, this was true of 1.6% of rapes, 0.9% of robberies, 0.2% of aggravated assaults, 0.2% of burglaries, and a negligible fraction of motor vehicle thefts (Langan and Brown, 1997a, page 3).

Death sentences were treated as equivalent to life sentences because the majority of offenders given death sentences ultimately are resentenced to life imprisonment.

According to unpublished NCRP tabulations, the average time served in State prison by released homicide cases serving life sentences was 152 months in 1990, 151 months in 1992, and 146 months in 1994. Interestingly, lifers who died in prison had served almost exactly the same time as lifers who were released (137 as opposed to 146 months in 1994). To this prison-time served was added the average time served in jail (10 months) by homicide offenders. The average total time served in 1990-94 (160 months) was used as the estimated time served in 1981-88. The average effective sentence length was estimated on the basis of the fraction of time served by homicide cases with non-life sentences. For example, this was 47% in 1994 (Langan and Brown, 1997a, table 4). Hence, the effective length of a life sentence in State court for homicide in 1994 was considered to be 332 months (156/.47).

A Federal law (the Sentencing Reform Act 1984) that went into effect in November 1987 specified that all Federal life sentences must be served in full, with no parole. In 1994, 9% of sentences in Federal court for homicide were life sentences (Sabol, 1997), and this percentage was used for all previous years. The average age of homicide offenders receiving life sentences is about 30, and about half are white and half are black. At age 30, a white male can expect to live 44.7 more years, and a black male 38.2 more (Bureau of the Census, 1994, table 116). Hence, it was assumed that the average sentence length and time served in 1988-94 for homicide offenders in Federal court receiving life sentences was about 40 years (480 months). The average sentence length and time served for 1981-86 were estimated from the State court figures.

The average sentence length and average time served for homicide were estimated by combining the State, Federal, and juvenile figures. For example, in 1994, the average prison sentence length was 332 months for lifers in State courts (2,846 sentenced to life and 276 sentenced to death), 269 months for non-lifers in State courts (8,285 offenders), 480 months for lifers in Federal courts (14 offenders), and 153 months for non-lifers in Federal courts (127 offenders). The average jail sentence length was 7 months in State courts (240 offenders) and 8 months in Federal courts (5 offenders). The average juvenile confinement sentence was 56 months (720 offenders). The weighted average of all these figures was 266 months. Similarly, the average time served by homicide cases in 1994 was estimated to be 127 months, and the percentage of time served was 48%.

An American offender's probability of arrest, and an arrested offender's probability of conviction

An offender's probability of arrest was calculated by dividing the number of persons allegedly committing a crime by the number arrested for that crime. The number of persons allegedly committing a crime was calculated by multiplying the number of crimes committed by the average number of persons committing the crime. For robbery, assault, burglary, and motor vehicle theft, the number of persons allegedly committing a crime was obtained from the NCVS; for murder, from the FBI's Supplemental Homicide Reports; for the number of rapes, from the FBI's Uniform Crime Reports; for the average number of rapists, from the NCVS.

The probability of an arrest resulting in a conviction was calculated by dividing the number convicted by the number arrested. The number arrested was obtained from the FBI's Uniform Crime Reports; the number convicted, from sources described previously.

Racial disparities in American incarceration rates

The 1991 estimates of racial disparities were calculated from national survey data on adults incarcerated in State and Federal prisons in 1991 and in local jails in 1989 (Lynch and others, 1994). To obtain estimates of the number of adults of each race that was incarcerated in 1991, the incarceration population (Lynch and others, 1994, page 3) was multiplied by the racial distribution of the incarceration population (Lynch and others, 1994, table 11). To estimate race-specific incarceration rates per 100,000 population, each race's incarcerated population was divided by its general population size.

To the top

Sources of English data

English survey offenses

In England the number of victim-survey offenses, comparable population figures (number of households and number of persons aged 16 or over), and the probability of reporting to the police were obtained from the British Crime Survey or "BCS" (Mirrlees-Black, Mayhew and Percy, 1996; Mayhew, 1996). The figures used in the report are the latest estimates for all years, taking account of the 1991 Census. For example, the BCS estimated that there were 1,754,057 household burglaries in 1995, and that 66% of them were reported to the police (Mirrlees-Black, Mayhew and Percy, 1996, table 2.3). Since there were an estimated 21,159,000 households in 1995, the survey burglary rate was 82.9 per 1,000 households; disregarding repeat victims, about one in every 12 households was burglarized in 1994.

Necessarily, BCS survey crime rates for robbery and serious assault (wounding) are per 1,000 population aged 16 or over, whereas comparable NCVS estimates are per 1,000 population aged 12 or over. Motor vehicle theft figures refer to completed thefts only. Other population estimates came from the Office for National Statistics (1996); again, figures used in this report are the latest estimates for all years.

The main change in the BCS over the years was the addition of a new screening question for domestic violence ("Has any member of your household deliberately hit you with their fists or with a weapon of any sort or kicked you or used force or violence in any other way?") and a new victim form in 1993. This caused an increase in the number of victim-survey offenses of serious assault. Mayhew (1997) and Percy (1998) provided serious assault estimates with and without the new domestic violence screening question: 691,633 (without) and 761, 205 (with) in 1993; 823,344 (without) and 860,395 (with) in 1995. For comparability with 1981-91, the "without" figures are used in these analyses.

For burglary, vehicle theft, robbery, and serious assault, the average number of offenders per offense was provided from the BCS by Mirrlees-Black (1996). For homicide, this was obtained from the annual Criminal Statistics (for example, Home Office, 1996a, page 73). As with the American calculations, the average over all years was used in estimating probabilities. This was 1.8 for burglary, 2.1 for motor vehicle theft, 2.4 for robbery, 2.0 for serious assault, and 1.1 for homicide. For rape, the only national data on the number of offenders per offense seems to be that published by Grace, Lloyd and Smith (1992) for offenses committed in 1985. They reported that, for police-recorded offenses, there were 1.1 offenders per rape offense. This figure was used in all calculations.

English police-recorded offenses

The numbers of police-recorded offenses were obtained from the annual Criminal Statistics (for example, Home Office, 1996a, tables 2.15 - 2.18). The number of police-recorded offenses refers to the number initially recorded by the police in each offense category in each year, irrespective of later court proceedings. Assault comprised section 5 wounding (mainly causing grievous bodily harm) and section 8 wounding (mainly causing actual bodily harm). Vehicle theft comprised theft or unauthorized taking of motor vehicles and aggravated vehicle taking. Only rape of a female was counted. To illustrate rate calculations, there were 1,239,484 burglaries of all kinds in 1995. Since the resident population was estimated to be 51,820,000 in 1995, the police-recorded burglary rate was 23.9 burglaries per 1000 persons.

The numbers of police-recorded offenses that were comparable to survey offenses were provided by Mayhew (1996). The estimation procedures were explained by Mayhew, Maung and Mirrlees-Black (1993, pages 163-167). For example, the 643,645 police-recorded burglaries in a dwelling (including aggravated burglaries) in 1995 were considered to be comparable to the BCS burglaries. The other main adjustments were to exclude thefts of commercial vehicles, attempt- ed vehicle thefts, and victims under 16. However, robberies of business property were counted in the BCS.

As before, dividing the number of comparable police-recorded offenses by the number of survey offenses yields the probability of a survey offense being recorded by the police. For burglary in 1995, this was .37 (643,645/1,754,057). Similarly, the probability of a reported burglary being recorded by the police in 1995 was .55 (643,645/1,162,940). The fact that the probability of a reported offense being recorded was greater than 1 for vehicle theft in 1981 and 1983 is probably a function of the confidence interval around the number of reported offenses and the fact that almost all of the reported offenses would have been recorded. The probability was set to 1.

To the top

English convictions

The number of persons convicted for each offense was obtained from the annual Supplementary Criminal Statistics (for example, Home Office, 1996b, tables S1.1A and S2.1A). It is necessary to add convictions in Crown Courts and Magistrates' Courts. In some cases, it was also necessary to add different offense categories (murder, manslaughter, infanticide, and manslaughter due to diminished responsibility for homicide; section 5 and 8 woundings for assault; theft of a motor vehicle, unauthorized taking, and indictable and summary aggravated vehicle taking for vehicle theft). As already explained, the rape convictions for 1995 included male and female victims and male and female offenders. The Supplementary Criminal Statistics showed that 556 of the 569 convicted offenders were male. Ayres (1997) further specified that 547 of the 556 male offenders had female victims.

As an example, 35,346 persons were convicted for burglary in 1995. Since there were an estimated 45,031,000 persons aged 10 or older in 1995 (Office for National Statistics, 1996), the conviction rate for burglary was 0.78 per 1000 population at risk. Since there were an estimated 1,754,057 household burglaries in 1995 and an estimated 1.8 offenders per offense, there were about 3,157,303 offenders (not necessarily different persons) who could in theory have been convicted for household burglary. Since household burglaries comprised about 52% of all recorded burglaries in 1995, there were about 6,080,100 offenders at risk of conviction for burglaries of all types. Dividing this by the number of persons convicted (35,346) yields the estimate that there were 172 offenders per conviction, or that the probability of conviction for each burglary offender was .0058 in 1995 (5.8 convictions per 1,000 offenders). As before, the calculations for rape and homicide were based only on police-recorded offenses.

English probability of incarceration

The number of persons sentenced to incarceration for each offense was obtained from the Supplementary Criminal Statistics (for example, Home Office, 1996b, tables S1.1A and S2.1A). As before, it was necessary to add Crown Courts and Magistrates' Courts and different offense categories. It was also necessary to add different types of custodial sentences (in 1995: imprisonment, detention in a young offender institution, secure hospital order, and detention under section 53 of the Children and Young Persons Act 1933; in 1981: imprisonment, borstal (reformatory), detention center, secure hospital order, detention under section 53 of the Children and Young Persons Act 1933).

As an example, 13,497 offenders were given custodial sentences for burglary in 1995, or 38% of all convicted burglary offenders. The incarceration rate for burglary in 1995 was 0.30 per 1000 persons aged 10 or over. As before, the probability of an offender receiving a custodial sentence was calculated by multiplying the probability of an offender being convicted by the probability of a conviction being followed by incarceration. For burglary in 1995, this was .0022, corresponding to one in every 450 burglars receiving a custodial sentence (2.2 per 1,000 offenders).

To the top

English sentence length, time served, and days at risk of serving

Offense-specific data on average sentence length and average time served in prison by released prisoners (including juveniles) were supplied by Burns (1996). Figures are for offenders released in 1982, 1984, 1988, 1991, 1993, and 1995. Initially, it was thought that the following year was the most relevant (for example, because most robbers sentenced to incarceration in 1981 were released in 1982). However, 1992 data were not available at the time of the last analysis (Farrington, Langan and Wikström 1994), so 1991 data were used, and hence 1993 and 1995 data were used subsequently.

Offense-specific data on average sentence length were published for the first time in the 1995 Supplementary Criminal Statistics (Home Office, 1996b, tables S1.4, S1.5, S2.5 and S2.6), but these figures were not used in this report (in the interests of comparability with earlier years).

The offense categories of prisoners are slightly different from those used in the Criminal Statistics, but "assault" and "wounding" in the prison data are approximately equivalent to the two "wounding" categories in the Criminal Statistics, and "taking and driving away" in the prison data covers at least 95% of those sentenced to incarceration for vehicle theft in the Criminal Statistics (Barclay, 1993). The average sentence length and average time served for assault were weighted combinations of the assault and wounding categories (weighted by the number of released offenders).

As an example, the average sentence length for burglary in 1995 was 14.9 months, and the average time served was 6.5 months, or 44% of the sentence. The average number of days served per conviction was 76 (.382 x 6.5 x 365.25/12). The average number of days served per burglar was 0.44 (76 x .0058).

There are no English national data routinely published on time served in incarceration before conviction. However, the report of the Carlisle committee (1988, page 147), based on 1987 releasees, estimated that about 10% of a prisoner's sentence was spent on remand before sentencing. The English figures on time served have not been adjusted to take this time into account.

English homicide

As in the United States, special estimation procedures were used for homicide because of the large number of life sentences. In England, a life sentence (or an equivalent indeterminate sentence for a juvenile) is mandatory for murder. In 1995, 214 life sentences (including indeterminate section 53 orders for juveniles) were given for murder, 12 life sentences were given for manslaughter, and 182 non-life prison sentences (including section 53 orders) were given for manslaughter (Home Office, 1996b, tables S2.1A, S2.5 and S2.6). There were also 21 secure hospital orders for homicide. Apart from these, 55% (226/408) of the custodial sentences for homicide in 1995 were life sentences.

In 1995, the average time served by homicide offenders first released from life sentences was 13.6 years or 163 months (Burns, 1996). In concordance with the American estimates, the effective length of life sentences was estimated on the basis of the fraction of time served by homicide cases with non-life sentences. For non-life homicide offenders released in 1995, the average sentence length was 48.4 months and the average time served was 21.0 months, or 43% of the sentence (Burns, 1996). Hence, the effective length of a life sentence for homicide in 1995 was considered to be 376 months (163/.43).

The average sentence length and average time served for homicide were weighted averages of life and non-life sentences. For example, in 1995, the average sentence length was 226 x 376 (life sentences) + 182 x 48.4 (non-life sentences), divided by 408, which came to 230 months.

Racial disparities in English incarceration rates

Racial disparities were calculated from 1991 national survey data on adults incarcerated in Prison Service facilities. To obtain estimates of the number of adults of each race who were incarcerated in 1991, the incarceration population (Lynch and others, 1994, page 3) was multiplied by the racial distribution of the incarceration population (Lynch and others, 1994, table 11). To estimate race-specific incarceration rates per 100,000 population, each race's incarcerated population was divided by its general population size (Lynch and others, 1994, table 11 note).

Prior research