Mr. Bishop Robinson
Secretary, Department of Public Safety
- and Correctional Services
6776 Reisterstown Road
Baltimore, Maryland 21215
Re: OBSCIS I Computerized Inmate Reporting System
Dear Mr. Robinson:
The purpose of this letter and its enclosures is to respond to your letter of November 5th, present an analysis of recent OBSCIS data and introduce a proposed new report format for the inmate characteristics report. This is also a follow up to my 1986 Management Analysis which quoted the Department of Budget and Fiscal Planning's comments on the OBSCIS system (relevant page enclosed as Exhibit A). I hope you find this material comprehensive enough for use by your Research and Statistics staff.
Thank you for your November 5th response with regard to the inmate characteristics report and for clarifying that, of the five (5) characteristics reported, the first one (sex) was by manual head count while the remaining four (4) (race, age, sentence and offense) are derived from the OBSCIS I inmate information system.
It is somewhat encouraging to know that even though OBSCIS suffers from data quality problems, the Department is able to derive its projections from head counts done manually at the institutions. Nevertheless, since the OBSCIS system is the only source of descriptive information about the population, it is important to ensure its availability and timeliness. A certain "margin of error" is not, per se, fatal in sociometric analysis but not to determine or disclose it undermines a potentially useful tool.
For this reason, I delved further in an attempt to test the validity of the OBSCIS data against the manual head count output as well as its internal consistency within categories. The results of this inquiry are enclosed as the chart, "Total Prison Inmates in Maryland" (Exhibit B) comparing October, 1991 and July, 1992 against October, 1992 predicted and "actual". "Total Prison Inmates in Maryland" -- Analysis
As can be revealed by the chart, the data by manual head count ("Sex"), extrapolating from the nine month (October, 1991 through July, 1992) rate of change proved to be quite accurate. Notice that the growth predicted for October, 1992 proved to be within 0.3% of the head count measured for that period for the Division and within 5.1% for the Department.
As the analysis will prove, the OBSCIS numbers, whether despite of or because of the "missing" elements you describe, are almost wholly increased over the (assumedly more definitive) head count. I have included "Plus or minus" rows in an attempt to quantify the internal variance for a given period.
Looking at the Actual Percent Growth within categories and the Difference in Percent Growth (last column), it can be seen that OBSCIS had predicted a rate of growth higher than what was eventually reported for the month. The second box from right, "Comparison of Predicted v. Actual", shows that while the head count method yielded an almost exact figure, OBSCIS counts of growth for the same period ranged from 8.2% to 14.7% less than predicted by OBSCIS itself. The range of growth reported by OBSCIS impacts the overall rate by a total of 0.4% to 0.7% (last column).
Probable Causes of OBSCIS Inaccuracy
In summary, regarding this analysis, the OBSCIS system may, as you say, contain "missing" elements but it is apparent that OBSCIS is carrying many more inmate records than head count reveals should exist. For example, the OBSCIS measure of the Department by Age is elevated by 182 over the manual head count for October of this year. Also, as I stated in my letter of October 20, even within the month of July, 1992, the difference between the OBSCIS Age figure and the head count figure is 285, about equal to the population of the SuperMax on any given day.
That is to say, assuming the existence of "missing" elements in inmate records as you explained, it is just as likely (if not more so), that the OBSCIS system is not removing records of inmates who are transferred, released or deceased. I believe you will find that one institution retains the record on its system while another adds the same inmate at the different location. The result is a double-count and/or overcount. The extent to which this is occurring is unknown because the "missing" element records are not disclosed.
However, the fact that on the one hand the OBSCIS counts are elevated in relation to the head count (for all three months studied), while on the other hand, their predictions of rate of growth vary considerably from category to category, points to a not inconsequential degree of internal inconsistency.
While the OBSCIS data is not without flaws, it is of such "ballpark" validity that it is not wholly without utility at least for the characteristics disclosed so far (race, age, sentence and offense). I am also aware that OBSCIS is the only source of information regarding the above. For these reasons, in the interests of optimizing the utility of OBSCIS output and improving the current method of presentation, I herewith present to you a new report format for the OBSCIS system data. New Report Format (Proposed)
The proposed new report format (Exhibit C), inspired in part by the federal government's National Update of the Bureau of Justics Statistics, has a number of features which I hope will be seen as improvements. First, since the structure and contents of the report are set forth on the introduction page, the reader is able to ascertain whether the copy in hand is whole or partial. Second, the metes and bounds of the population under study is delineated. Third, the distinction between the manual head count data and the OBSCIS-derived data is made clear. The manual data then becomes a "baseline" against which to gauge the OBSCIS information. Finally, the amount of uncertainty ("margin of error") as well as the reasons for uncertainty, a vital component of sociometric analysis, are quantified and described to the extent possible.
Part I. Introduction and Head Count Year-to-Year Summary presents a capsulate overview in form and format heretofore not evident in the Department's population disclosures. With the simple addition of "One Year Ago" and "One Year From Now" figures, the data can be viewed in a more meaningful context and their importance for the future demonstrated.
Part II. Description of the Data is a necessary component of a sociometric analysis. To omit the caveats it discloses undermines the veracity, reliability and utility of the remainder of the presentation. Notice the use of the "Sex" parameter as a benchmark against which the manual head count and OBSCIS data can be evaluated in relationship to each other.
Part III. List of Institutions Analyzed contains the same information as has been disclosed previously but with the addition of the "Predominant Security Level" column. This simple addition to the existing record provides an added dimension and quality of analysis without contributing to bulk or complexity. A simple footnote provides relevant description not contained elsewhere on the page.
Part IV. Graphical Overview of Findings is wholly new and easily accomplished using minimal personal computer equipment. The computations and graphing were done in spreadsheet software and converted to word processing with a simple utility included with the software. For the reasons seen below, graphical analysis of data can be very instructive and may reveal aspects not easily evident from data in chart form. For example: Sex: From the appearance of this "pie" graph, it can be seen with much more impact that women are a very small part of the Division's operation. However, the ability to compare such "pie" charts over time would presumably illustrate an increasing trend. Age: It can be seen that the vast bulk of the population is in their mid to late 20s (rising sharply in the early 20s), a fact not belied by using the average age figure of 31.3. Race: The "Indian" and "Other" categories are vanishingly small. Perhaps for the sake of the report, they should be combined. Sentence: The very large column to the right would tend to suggest that the category of "Over 180 Months" (i.e., "over 15 years") ought to be further defined. Surely, it would be in the Department's best planning and management efforts to measure and report the following additional categories:
Especially in the "Life Without Parole" category, with many more of these inmates entering the system every year, it is advisable that the Department create and maintain accurate records beginning now. They are certain to result in fiscal impact later.
Part V. Detailed Presentation of Data merely reproduces the current OBSCIS report with the exception that the former footnote on the "Sex/Race" page has been more fully elucidated in Part II and elsewhere in the report. I have photocopied it onto letter-length paper for ease of handling. The advantage of Part V is that the reader has the opportunity to evaluate individual institutions against others of the same security level, region, comparable size and so forth. (Perhaps a further improvement would be to put the abbreviations and the institutions in some kind of order: alphabetical, by region, by security level or by some other easily referenced parameter).
Additional Chart(s) - Length of Stay
I would suggest that at one time or another the Department consider adding additional charts to its Part V analysis. For example, it would be illuminating to get an idea of the time each inmate has spent in the system since his most recent incarceration (ignoring diminution credits). The present "Sentence Distribution" report does contain a column (to the far right) of "Average Stay" but it does not show supporting figures as does the rest of the chart. A "stay" analysis would reveal factors such as amount of turnover, the blend of "stays" within a particular institution, the "greying" of population such as is occurring with long-term inmates at the Penitentiary, the effect of diminution credits on release rates, the effect of changes in parole policy, and so forth.
The Hidden Dimension - Releases for the Same Period
Neither the present report nor the report I have proposed contend with the issues of how many inmates were transferred during the period and, especially, how many inmates were released (even to the custody of the Division of Parole and Probation). Perhaps it would be possible for the Department to produce a separate report detailing same, and in the case of released inmates, a profile (including sex, race, age, sentence, offense and "stay") similar to that described here. Such as report would be useful to both aftercare and law enforcement agencies and would be helpful in resolving public policy issues.
Thank you for your attention to this letter. I am looking forward to improvements in the OBSCIS I reporting system and in its report and report format. I hope you or members of your staff find the comments and suggestions of this letter and its enclosures of assistance. For the convenience of your staff, I am enclosing a computer diskette containing the templates for the graphs and proposed report format. Please let me know if you have any questions or comments or require further information. You or a designee may call me at my home number.
Sincerely,
Nancy Moran
Volunteer
nm
enclosures
cc:
Governor William Donald Schaefer
Senator Charles H. Smelser
Delegate Timothy Maloney
Ms. Nancy Nowak (w/o Part V)
Commissioner Richard A. Lanham, Sr. (w/o Part V)
Director Joseph Henneberry (w/o Part V)
Mr. Richard Tamberrino (w/o Part V)