1 This is the first of three articles analyzing collateral attacks on convictions in Illinois, California, Colorado, and Texas. The present paper will present the fundamental measures of filings as a necessary basis for further understanding of the subject. Comparison of the available figures across the four states provides decision makers with the opportunity to compare different systems quantitatively. From that comparison a more informative dialogue can follow and some solutions to the perceived problems of postconviction remedies can at least be considered. The first-order statistics derived in this paper also constitute the basis for the analysis and the ultimate understanding of the filing phenomenon to be discussed in the following articles. The second article will answer the question “who files?” by concentrating on the relationships between collateral attacks and other aspects of incarceration and adjudication, in particular the sentence and the direct appeal process as well as the periods of delay inherent in the system. The third article will analyze the nature of the allegations raised and their eventual disposition.Google Scholar

2 See Uniform Post-Conviction Procedure Act, National Conference of Commissioners on Uniform State Laws. For a list of states adopting postconviction acts see ABA Standards Relating to Post-Conviction Remedies, app. C at 112 (Approved Draft, 1968).Google Scholar

3 1972 Annual Report of the Director of the Administrative Office of the United States Courts, table on page A-15, “Prisoners Petitions” column.Google Scholar

4 Although there is a possibility of filing a motion while on parole or probation or after a misdemeanor conviction, it rarely occurs. Such instances were omitted from the study.Google Scholar

5 The process of selecting states for the study began with a summary of available information in each of the 50 states about the major differences and similarities in the types of postconviction remedies available as well as a review of broad indicators of prison populations and court activities. Fifteen states were found to have the full range of combinations of major factors of interest: California, Colorado, Connecticut, Florida, Illinois, Louisiana, Maine, Massachusetts, Missouri, Nebraska, New York, Pennsylvania, Texas, Virginia, and Wisconsin.Google Scholar

After a more detailed analysis of the law and an evaluation of the feasibility of data collection and its projected costs, four states were chosen for the study. Illinois was one of the earliest states to provide, independent of the traditional habeas corpus procedure, a special procedure for collateral attacks. It also provides assistance of counsel in postconviction cases and allows them as a matter of right. California represented those states in which collateral attacks were developed through additions and extensions–almost continuously–of habeas corpus statutes. The state's long established data-collection and record-keeping systems were of great help. Texas provided an opportunity to study a system where all postconviction motions were channeled to and decided in a single appellate court. It was hoped also to study Missouri, where a rule change was under study that would have consolidated the regular appellate review process and all collateral matters into a single procedure. But when it became clear that the rule would not be implemented in time for the study, Colorado was chosen to represent those states with a much smaller absolute number of prisoners. While this number is between an eighth and a quarter of the number of prisoners in the other three states, the state's number of prisoners per 100,000 population is similar to Texas and California and almost double that of Illinois.Google Scholar

6 Ill. Ann. Stat., ch. 38, secs. 122-1 to 122-7 (Smith-Hurd 1973).CrossRefGoogle Scholar

7 Cal. Penal Code secs. 1473-1508 (West 1970).Google Scholar

8 Frequent transfers of prisoners from one institution to another prevented limitation of the search to the present county of incarceration.Google Scholar

9 Tex. Code Crim. Proc. Ann. arts. 11.01 to 11.59 (Vernon 1966).Google Scholar

10 Colo. Rev. Stat. vol. 7 (Rules of Criminal Procedure) (1973).Google Scholar

11 This is the new Colo. Rev. Stat. sec. 40-1-510, effective July 1, 1972, which permits appellate review of the sentence itself, among other things.Google Scholar

12 Supra note 2.Google Scholar

13 The latter difference was not always clearly discernible from the prisoner's petition, either.Google Scholar

14 Cal. Penal Code sec. 3040 (West 1970).Google Scholar

15 Such operational definitions are not necessarily fault free; they represent the best compromise possible under the circumstances. Thus, no courts of lesser jurisdiction than the felony courts were searched for possible filings, although such filings would not be surprising, given the erratic form of filings and the sporadic behavior of filers that were encountered. In deciding when a certain communication by or on behalf of a prisoner is a “filing,” there was no better choice but to take it into account only when a court recognized it as such. There were discrepancies in the way different courts treated prisoners' communications. In California, despite the existence of a special form for a habeas corpus petition, petitions were always accepted even when the form was not utilized. A letter to a judge would not be filed in some courts but would be given full formal treatment in others, provided it met the criteria of a collateral attack, which varied from place to place. Perhaps because it was less burdened, Colorado gave full attention, as possible collateral attacks, to many of the petitions deemed unacceptable in other places; some of them even resulted in relief.Google Scholar

16 The prison records of 18 people could not be located–a small number considering that many prisoners are in transit from one prison to another and many records were no longer in the active prison file because prisoners either were on parole or had been discharged.Google Scholar

17 The sources of the data were the various official documents from the state departments of correction, the prisons, and the courts in which the petitions were filed. There were no personal interviews. If one allows for the differential availability of the data in the four states and, more important, for the difference in definitions of such terms as “short time,”“earliest parole date,” and others (even of an “acceptable motion!”), it can be said that essentially the same type and detail of information was collected in each state.Google Scholar

18 The same situation was simultaneously encountered in California.Google Scholar

19 See note 24 infra.Google Scholar

20 An inconsequential amount of additional information could be obtained from this group about filings in the 54th to 66th months of incarceration, but it can be more effectively calculated by statistical methods. See note 24 infra.Google Scholar

a The filing period is the interval from but not including the first number in each pair through the second number; i.e., “6-12” indicates that in this category motions were filed in the period after 6 months and up to 12 months after admission.Google Scholar

b Two 6-month periods were pooled in this category because of the small number of cases.Google Scholar

23 The calculations presented for the filers or the nonfilers are not necessarily based on the full 244 or 970 cases, respectively; information for some variables or combination of variables may be missing from the records, thus reducing the sample size on certain data.Google Scholar

24 The date the petition was filed was defined to be the date on the petition, not the date the petition was filed in court. In this way possible discrepancies due to delays in the system between the manifestation of the intention to file and the legal act of filing were avoided. To enable use of all available information, the mean difference between these two dates was calculated to be.61 months. This figure was used to derive the date filed for those few cases where only the date filed in court was available.Google Scholar

25 That is, e.g., one can be 95 percent confident that the probability of filing within the 6-month or under period is between.023 –.005 and.023 +.005, or.018 and.028. An interval of the form.003 ±.005 should, of course, be interpreted to mean 0 to.008.Google Scholar

26 The exponential density function is f(t) =θe^−θt , where θ is the only parameter that determines the distribution and t represents the time of filing in months. The probability of filing within any period t (in months) is given by 1 –e^−θt . In particular the probability of filing within 54 months is given by 1 –e ^−54θ and therefore the probability of filing after 54 months is given by e ^−54θ. Thus, if one could estimate θ and prove that the exponential distribution is a good fit to the data, one would then be able to estimate the probability of filing after 54 months and consequently the probability of ever filing a motion.Google Scholar

For the exponential distribution f(t) =θe^−θt , the corresponding truncated distribution at 54 months is f(t) –θe^−θt /(1 – e^−54θ). The maximum likelihood estimator for θ was developed and evaluated through an iterative process utilizing the mean time to filing within 54 months as calculated from the sample. The solution of the equation 1/θ–±= 54 e^−54θ/(1 – e^−54θ) in Illinois was found at θ=.0435 for the mean time to filing of ±= 17.3 months.Google Scholar

Using this estimate for θ one then calculates the probability of filing after 54 months as.0929 e^−54θ/(1 – e^−54θ)=.0098 and the probability of ever filing.0929 +.0098 =.1027. The probability of filing in each category is calculated by the formula.1027 (e^−θt1 – e^−θt 2), where t ₁ and t ₂ are the two ends of the interval. Thus, for example, the probability of filing in the 6-12-month period is.1027 (e ^{−.0435 × 6}–e ^{−.0435 × 12})=.0182. The probability of filing within each period according to this fitted distribution appears in col. 2 of the table below. The expected number of filers according to the fitted distribution is calculated in col. 4 by multiplying the number of people at risk by the probability of filing in the period as given by this distribution. The last column gives the X ² figure, which evaluates the discrepancy between the two preceding columns according to the standard formula of (observed-expected)²/expected.Google Scholar

Fitting an Exponential Distribution to the Filing Data for Illinois^a Google Scholar

a For an estimated ±=.0435Google Scholar

b The two periods are combined into one in order that the expected number of cases may be greater than 5 to assure validity of the X ² test.Google Scholar

c The 5.02 figure is not significant at the.05 level when compared with the X ² with 5 degrees of freedom for this level, which equals 11.07. One can thus conclude that this exponential distribution adequately fits the data.Google Scholar

The sum of the figures in the last column is the statistic on which the goodness of fit test is performed. In this case it is found that 5.02 is smaller than the 11.07 figure that corresponds to the X ² distribution with 5 degrees of freedom at the.05 level; therefore there is no significant difference between the sample distribution and the fitted exponential one that thus adequately represents the data. One can thus further estimate from it the probability of filing after 54 months to be.0098 and the probability of ever filing in Illinois to be.1027.Google Scholar

A final correction is now needed for the probability figures. For 8 filers out of the total of 244 shown in table 2 the time to filing could not be calculated due to missing information on either the date of filing or the admission date. Since the probabilities were calculated upon the remaining 236 filers, the figures have to be inflated by a factor of 244/236, making the probability of filing within 54 months equal to 244/236 ×.093 =.096 and the probability of ever filing equal to 244/236 ×.103 =.106. Carrying through the corresponding evaluation of the error term, the final figures for the two estimated probabilities in Illinois, presented in terms of the significant digits that the error term permits, are.10 ±.02 and.11 ±.02, respectively.Google Scholar

Comparing the last figure, or better, the corresponding one for the probability of filing within 54 months (.10 ±.02), with the probability of filing for the 1967 generation for which there are 49 filers attributed to 583 prisoners, one finds a probability of 49/583 =.08 with a margin of error of ±.02. The two figures are within the margin of error of each other. The difference indicates that the 1967 generation might have been slightly below the others in the probability of filing and thus does not represent all generations well. It is also interesting to note the irregularity in the 30-36-month period in the sequence of probabilities in col. 4 of table 3: the higher than expected probability is a result of the lack of uniformity of behavior over time among the various generations.Google Scholar

Finally, note that the 236 filers used in this last correction include the 211 filers in table 3. The difference between the two figures of 25 filers is created by the method of estimation: these 25 filers include, for example, those from the 1971 generation filing before the cutoff date but more than 6 months after admission who were not counted because not all prisoners admitted in 1971 were at risk for more than 6 months for their first filing; the same is true for the other generations. One could refine the method used to utilize the full 236 cases by tediously identifying within each generation of admissions those individuals who were at risk for the corresponding 6-month period, adding them to the population at risk and adding to the count of filers those among them who filed in the period. The difference between this refinement and the method used is inconsequential.Google Scholar

30 Tables 2 and 3 give data for Illinois; tables A1 and A2 give corresponding data for the other three states. The fitting of the exponential distribution to the filing data for California and Texas is presented in table A3. The exponential distribution does not fit the filing data for Colorado because of the irregular pattern of renewed activity revealed in the Colorado data for the 42-54-month period (table A2). This fact prevents any attempt to estimate the probability of ever filing in Colorado and serves as an early warning of the unusual behavior of filing over time, as will become evident in sec. VI.Google Scholar

a The probability of ever filing was not estimated for Colorado as explained in note 25.Google Scholar

NOTE: The figures in table 4 represent the final estimates derived while utilizing the correction factors presented in the following table:Google Scholar

32 See note 11 supra.Google Scholar

33 Similar figures based upon the probability of ever filing for Illinois, California, and Texas are presented in table A4.Google Scholar

34 For the last three states, the corresponding figures based on the probability of ever filing presented in table A4 should be used whenever absolute rather than comparative figures are needed.Google Scholar

35 “Intensity” is used in this paper to mean “degree or amount of a quality or condition.”Google Scholar

36 The curve fitted is of the form y = ae^{b(1972 – x)} , where y represents the mean number of filings for the corresponding year of entry, x. The transformation 1972 –x expresses the average time allowed for filing, which is 1 year for the 1971 generation, up to 5 years for the 1967 one. For Illinois the curve of best fit was found to be the one with a= 1.07 and b=.05, or y = 1.07e^{.05(1972 – x)} , with a coefficient of determination of r ²=.65, a measure of the goodness of the fit where 1 indicates a perfect fit. From this curve, the extrapolated mean number of filings over 6 years (1972 –x= 6) was found to be 1.41. The 6-year period was adopted to allow for some possible additional undetected filing, even at the risk of slight overestimation. However, the figures for 5 (1.35), 6 (1.42), or 7 (1.48) years do not vary much, a good indication of the usefulness of the exponential curve.Google Scholar

37 The full information for each state is given in table A5, from which the figures in the first row of table 6 are taken.Google Scholar

38 Figures are based on the 1967 generation of entry except for Illinois, where the 1967 figure was aberrational and the mean for the 1967 and 1968 generations was taken; and Colorado, for which 1968 was used.Google Scholar

39 In the whole sample only 9 individuals had 3 filings, only 1 had 4, and none had more than 4.Google Scholar

40 As evidenced also by the new legislation introducted in 1972; see note 11 supra.Google Scholar

41 The mean of the exponential distribution is mathematically equal to the reciprocal of the parameter of the distribution, θ. Recalling the corresponding parameters for the states from sec. II, one finds for Illinois 1/.0435 = 23 months, for California 1/.050 = 20 months, and for Texas 1/.071 = 14 months as the estimated mean times to filing the first petition. But the mean is not, in general, a good representative of an exponential distribution. Better statistics are presented in table 8. These figures are presented here for the reader who might find them useful for comparison with other information at hand for which only means may be available. Since the figures in table 8 as well as those presented above rely heavily on the exponential distribution, it is instructive at this point to find reassurance in the fact that these figures bear a close relationship to figures acceptable intuitively, as can be seen by resorting to the same technique utilized in estimating the mean number of filings in the preceding section, this time applied to the length of time from admission to the filing of the first postconviction motion. Specifically, an exponential curve is fitted to the mean length of time to filing for the five generations of entry in each state. The estimated mean time to filing for each state is then read off the fitted curve. The detailed figures are presented in table A6 and contrasted with the figures based on the exponential distribution in the following table:Google Scholar

Comparison of Mean Time to Filing by Different Methods of EstimationGoogle Scholar

Although both methods are applied to the same set of data and have the word “exponential” in common, they are completely different and independent of each other, which is why one can serve as a check on the other. The results are mutually reassuring to the point that no explanation of divergences seems necessary; they thus justify the further utilization of the exponential distribution.Google Scholar

42 Calculated by substituting t= 6; 12; 24 in the formula for the probability of filing within time t, which for the exponential distribution is P_r ±= 1 –e^−θt .Google Scholar

43 This is achieved by solving the equation 1 –e^−θt= p each time for a certain value of p given in the table.Google Scholar

44 These petitions are filed after sentencing but before a mittimus to state prison. For example, the petitioner may be in county jail awaiting appeal, or awaiting trial in another county, or possibly on appeal bond.Google Scholar

45 Similar calculations for the other states are presented in tables A7 and A8.Google Scholar

46 By comparing the X ² calculated in the table to the value of the chi-square distribution with the corresponding number of degrees of freedom at the 5 percent level of significance. The results of these comparisons are: for the total number of filings, X ²= 11.44 is greater than X ² _.05(5) = 11.07. For the filers within the 6-month period, X ²= 8.58 is greater than X ² _.05(3) = 7.81.Google Scholar

For the total number of filers the number of degrees of freedom is 15 (number of cells in table 10) minus 5 (number of parameters estimated from the table–the five probabilities used in the computation) minus 5 (number of restrictions since the sums of the observed–and expected–number of filers in each time period is predetermined) to a total of 5. Similarly, for the filers within the 6-month period, the number of degrees of freedom is 5 – 1 – 1 = 3.Google Scholar

47 By comparing the X ² calculated in the table to the value of the chi-square distribution with the corresponding number of degrees of freedom at the 5 percent level of significance. The results of these comparisons are: for the total number of filings, X ²= 11.44 is greater than X ² _.05(5) = 11.07. For the filers within the 6-month period, x²= 8.58 is greater than x².05(3) = 7.81.Google Scholar

48 The calculations for Colorado exclude the 1970 generation of entry; consequently the degrees of freedom are different for the chi-square test compared with the other states; they are 5 – 3 = 2 for the total filers because there are 3 less cells in the Colorado table (table A8) than in table A7. Similarly there are only 3 – 1 = 2 degrees of freedom for 6-month filers. The numbers of filers in the Colorado section of this table were derived by weighting the observed and expected number of filers for Buena Vista and Cañon City to arrive at correct estimates for the state. The numbers are still referred to, for uniformity, as the “observed” and “expected” ones, but the X ² values are corrected accordingly, to account for the weighting.Google Scholar

49 The degrees of freedom are different; thus the X ² is not comparable.Google Scholar

50 This is true for each of the four states with regard to the 1967 year (1968 in Colorado) since only the first generation could contribute to it. For this reason the discussion focuses mainly on the later years for which the contrasts are more reliable. Similarly, looking at figure 1 with and without the first year may be helpful in understanding it. However, because of the consistency across states it is possible to speculate that the first year truly has a lower probability of filing.Google Scholar

51 Supra note 11.Google Scholar