Appendix: Data and Methods
This appendix contains additional details on the data used for this study and my analyses, including the quality control process, how I identified contingency-fee counties, and tables with more detailed reporting and error rates.
Data sources
Civil case database: Because civil forfeiture cases are civil cases, I requested and obtained a database from the Indiana Office of Court Services of all nonconfidential civil cases in every Indiana county from January 1, 2016, to December 31, 2021. The IOCS provided the data in the form of dozens of text files containing millions of rows of data representing every recorded action across each of the cases, as well as summary information for each case. The summary information includes details like the case number, case title (e.g., “State of Indiana v. $300”), the date the case was opened, and so forth. I identified forfeiture cases by searching text entries in certain fields—such as the case title or defendant field—for characters or parts of words often found in forfeiture cases. My search terms included a dollar sign, the term “VIN” for an automobile’s vehicle identification number, and variations on parts of the word forfeiture, such as “forf” and “foref,” the latter included in case forfeiture was misspelled in the data. I then checked the cases to confirm they were indeed forfeiture cases. False positives tended to be land disputes between private parties, cases involving real estate foreclosures, and non-forfeiture cases involving the state Bureau of Motor Vehicles.
Case-level reported forfeiture data: Each year, the Institute for Justice obtains an Excel workbook from the Indiana Prosecuting Attorneys Council containing information compiled from all forfeiture cases reported to IPAC for the prior fiscal year. The data include case numbers, number of items forfeited, amount of property forfeited, amounts returned to property owners, whether a case was settled, and dollar-amount distributions to the Common School Fund, attorney fees, prosecutors, the Indiana State Police, sheriffs’ offices, police departments, and other law enforcement agencies. I used the IPAC data for two main purposes. First, I cross-referenced cases in the IPAC data with cases in the civil case database described above to identify unreported forfeiture cases in the latter. Second, I compared the reported information within the IPAC case data to the information observed in court records to assess the accuracy of the IPAC case data.
Aggregate forfeiture reports: Each year, IPAC uses the case-level reported forfeiture data described above to produce an aggregate forfeiture report for the legislature. These reports contain state-level summary information for most of the information compiled in the case-level data. I used these reports to examine how certain aggregate statistics reported within them—such as the frequency of settlements in forfeiture cases—compared with the same statistics calculated from the random sample of forfeiture cases in the IPAC data over the same period.
Court documents and case information: The Indiana Supreme Court’s Office of Judicial Administration operates an online search engine called MyCase that allows users to examine individual case information and download court documents for nonconfidential civil and criminal cases in the state. My team used MyCase to download forfeiture case documents that we then coded to compare with the case-level information contained in the IPAC datasets.
Identifying and checking reporting errors
Because of the large number of unreported cases to code and the large number of randomly selected reported cases to review for errors, I benefited from the help of a research assistant who coded information about unreported and reported cases. For both types of coding work, the assistant worked in several phases to allow for trainings and multiple rounds of reviews and quality control checks. This process transpired as follows.
First, for the coding of unreported case information—which included the amount of any currency forfeitures, the number of vehicles forfeited, and whether seized property was transferred to the federal government—I conducted a thorough training to ensure that the assistant and I both understood where to find the information we needed to code and how to record the information in the codebook. While these were relatively simple tasks, I wanted to minimize uncertainty, come to a mutual understanding on terminology, and reduce the likelihood of making errors during the coding process.
Following the training, the assistant coded information for an initial set of 100 cases, after which a colleague with extensive experience working with forfeiture data and I reviewed all coding decisions to identify any errors and the reasons they were errors and to ensure any errors were corrected. We identified only one error—a case mistakenly coded as involving a federal transfer. All other coding decisions were correct. I shared the information on the incorrect coding decision with the assistant and provided guidance on how to correctly identify federal transfers for the remaining cases. For the rest of the unreported case coding, I directed the assistant to flag any coding decisions she was unsure about and to keep detailed notes on particularly difficult to understand cases so that I could review them for accuracy. In addition, my colleague coded 566 of the 592 unreported forfeitures in Marion and Vanderburgh counties.
Upon completion of coding for the approximately 1,998 total unreported cases, my colleague and I first reviewed all cases flagged by the research assistant while I reviewed any cases my colleague marked for review among the Marion and Vanderburgh county cases he coded. Coming out of this review, I deleted 12 cases as duplicative or otherwise having key fields coded correctly in a merged case elsewhere in the dataset. For instance, I identified case 22D01-1811-MI-001605 as a forfeiture case in the IOCS data. However, an entry toward the end of the case history on MyCase noted that the case was closed and said to “see case 22D01-1811-PL-1736,” which also existed within the dataset as an unreported forfeiture case. After confirming these were duplicates, I deleted case 22D01-1811-MI-001605 from the unreported forfeiture case dataset.
In addition, I deleted 19 cases that had been flagged as potentially not being forfeiture cases. Upon closer inspection, they, indeed, turned out not to be forfeiture cases. These included cases dealing with private foreclosures, cases dealing with reinstatement of driving privileges, cases we could not find on MyCase, and cases the state had opened in error. Such cases initially appeared in the dataset because certain fields contained strings or substrings used to identify likely forfeiture cases in the IOCS data.
This left a total of 1,967 unreported forfeiture cases that were opened between calendar years 2016 and 2021 and closed by October 31, 2022.
In addition, the assistant flagged 150 unreported forfeiture cases as not having enough information in MyCase to allow coding of some or all of the fields in the dataset. For instance, some cases did not have downloadable orders but instead had only brief text entries allowing observation of some characteristics (such as the fact that the case was indeed a forfeiture case) but not others (such as the value of the forfeited property). These cases remain in the dataset as unreported forfeiture cases, but any additional information on those cases that could not be coded obviously does not appear in the reported results. This is a key reason the amount of unreported cash and vehicles forfeited are conservative, lower-bound numbers.
Following the reviews of all cases flagged by the assistant, I conducted three additional quality control checks:
Check 1: A non-random check of the highest and lowest coded forfeiture values. To ensure the largest forfeiture values and the smallest non-zero forfeiture values were entered correctly, I checked the coding for the 23 unreported cases with the highest amounts (values of at least $30,000; the highest-value forfeiture coded was $525,000) and the 18 unreported cases with the lowest forfeiture amounts ($100 or less). Among these 41 cases, I corrected four high-value cases and one low-value case.
Check 2: A random spot-check of coded cases in Marion and Vanderburgh counties. Among the 566 unreported forfeiture cases in Marion and Vanderburgh counties coded by my colleague, I randomly spot-checked 45 that had not been previously reviewed or corrected. I did not observe any errors pertaining to reported fields.
Check 3: A random spot-check of the remaining coded cases. Finally, among the remaining 1,199 unreported forfeiture cases, I randomly spot-checked 150 that I had not previously reviewed or corrected as part of my process for checking cases that were flagged for review or on my own initiative after reading notes left by the assistant in the coding notes field. Using this 150-case sample, I calculated the following accuracy rates:
- Amount of cash forfeited: 145/150 = 96.7%
- Number of vehicles forfeited: 147/150 = 98%
- Whether a federal transfer occurred: 150/150 = 100%
I corrected all errors identified before proceeding with my analysis.
Ensuring quality in coding the accuracy of reported cases
As with the process for the coding of unreported forfeiture cases, I began with a training on where to find the applicable coding items within the court documents and how to record the information. Similarly, I checked all applicable fields among each of the first 100 cases coded by the assistant to ensure she was making correct decisions and to create an opportunity for additional guidance. My colleague and I also checked 47 cases that the assistant flagged for review out of the remaining 315 cases. These cases tended to be complicated forfeiture cases with many assets or unusual circumstances. This means 147 out of 415 cases were already checked and corrected, if necessary, prior to the random spot check of 50 of the remaining 268 cases.
Using this 50-case sample, I calculated an accuracy rate for the coding decisions of 100% for 10 of the 14 fields examined. The fields with an error were:
- Whether a settlement occurred: 47/50 = 94%
- Number of properties returned: 48/50 = 96%
- Amount of cash forfeited: 48/50 = 96%
- Amount of cash returned to claimants: 49/50 = 98%
Here again, I corrected all errors before proceeding with my analysis.
Identifying contingency-fee counties
Because Indiana’s contingency-fee forfeiture scheme creates an additional conflict of interest in counties that take advantage of it, I explored whether reporting problems were particularly bad in those counties. Since no definitive source identifying such counties exists, I had to create such a list. To do that, I compared the plaintiff attorneys appearing in forfeiture cases in the IOCS civil case database to the attorneys serving in a given county prosecutors’ office. The list of attorneys serving in county prosecutor’s offices came from two sources: (1) the IPAC “Find Your Prosecutor” website 1 and (2) a public employee compensation database containing the full names of employees, which I searched for employees where “prosecut” appeared in their department or job title/duties fields between 2016 and 2021. 2 If any of the plaintiff attorneys for a given county did not appear on the list of attorneys serving in that same county prosecutor’s office, I deemed the county to be a contingency-fee county.
I then checked the contingency-fee designations in two ways. First, I examined final forfeiture orders for each contingency-fee county, confirming that status if the attorney for the state was indeed a private attorney or there was a distribution of proceeds in the form of attorney fees to a private attorney. All counties that were initially classified as contingency-fee counties were confirmed using this approach. Second, to see if any additional contingency-fee counties could be found, I examined reported forfeiture cases in the IPAC data that came from counties classified as non-contingency-fee counties but that included distribution amounts for attorney fees, as such fees should only appear if a private attorney was involved in the case. All instances of attorney fees in such cases were found to be incorrect except for those in two additional counties—Johnson and Washington counties. As such, they are classified in this study as contingency-fee counties.
Reporting and error rates in greater detail
Table A1: Reporting rates and amounts of unreported currency forfeitures for the 20 counties with the lowest reporting rates*
| County | Contingency fee? | Unreported forfeiture cases | All known forfeiture cases (reported and unreported) | Percent of forfeiture cases reported | Dollar amount of unreported currency forfeited |
|---|---|---|---|---|---|
| Elkhart | 76 | 76 | 0.0% | $246,440 | |
| Clark | 49 | 49 | 0.0% | $96,128 | |
| Warrick | 14 | 14 | 0.0% | $13,734 | |
| Franklin | Yes | 11 | 11 | 0.0% | $18,392 |
| Vigo | Yes | 263 | 277 | 5.1% | $737,742 |
| Jefferson | Yes | 18 | 19 | 5.3% | $7,975 |
| Boone | Yes | 17 | 18 | 5.6% | $28,250 |
| Dearborn | Yes | 38 | 41 | 7.3% | $366,637 |
| Vanderburgh | Yes | 403 | 436 | 7.6% | $908,700 |
| Monroe | Yes | 20 | 22 | 9.1% | $204,087 |
| Madison | Yes | 70 | 81 | 13.6% | $369,110 |
| Floyd | 16 | 21 | 23.8% | $32,305 | |
| Hamilton | Yes | 91 | 145 | 37.2% | $472,391 |
| Tippecanoe | 13 | 21 | 38.1% | $147,759 | |
| Lawrence | Yes | 17 | 28 | 39.3% | $21,544 |
| Fulton | Yes | 10 | 18 | 44.4% | $26,393 |
| Miami | Yes | 21 | 40 | 47.5% | $685,776 |
| Shelby | Yes | 12 | 26 | 53.8% | $65,578 |
| Newton | 6 | 13 | 53.8% | $500 | |
| St. Joseph | Yes | 87 | 189 | 54.0% | $172,894 |
| State total (all counties) | 1,967 | 6,731 | 70.8% | $6,252,565 | |
| Total contingency- fee counties | 1,418 | 2,763 | 48.7% | $4,936,105 | |
| Total non-contingency- fee counties | 549 | 3,968 | 86.2% | $1,316,460 |
*Among counties with at least 10 known forfeiture cases
Table A2: Errors, types of errors, and rates of errors observed in the random sample of reported forfeiture cases
| Number and percent of entries with an error | Number and percent of errors that were overcounts (or false positives) | Number and percent of errors that were undercounts (or false negatives) | ||||
|---|---|---|---|---|---|---|
| Full sample (n=415 cases) | ||||||
| Vehicles forfeited | 40 | 9.6% | 35 | 87.5% | 5 | 12.5% |
| Firearms forfeited | 6 | 1.4% | 2 | 33.3% | 4 | 66.7% |
| Real property forfeited | 7 | 1.7% | 0 | 0% | 7 | 100% |
| Cash property forfeited | 60 | 14.5% | 46 | 76.7% | 14 | 23.3% |
| Property returned to claimants | 94 | 22.7% | 7 | 7.4% | 87 | 92.6% |
| Settlement | 132 | 31.8% | 13 | 9.8% | 119 | 90.2% |
| Cases with at least one error | 177 | 42.7% | ||||
| Cash-only sample (n=340 cases) | ||||||
| Amount forfeited | 107 | 31.5% | 101 | 94.4% | 6 | 5.6% |
| Amount returned to claimants | 51 | 15.0% | 9 | 17.6% | 42 | 82.4% |
| Amount distributed to Common School Fund | 45 | 13.2% | 35 | 77.8% | 10 | 22.2% |
| Amount distributed to attorney fees | 33 | 9.7% | 19 | 57.6% | 14 | 42.4% |
| Amount distributed to prosecutors’ offices | 68 | 20.0% | 51 | 75.0% | 17 | 25.0% |
| Amount distributed to Indiana State Police | 6 | 1.8% | 5 | 83.3% | 1 | 16.7% |
| Amount distributed to sheriffs’ offices | 18 | 5.3% | 15 | 83.3% | 3 | 16.7% |
| Amount distributed to police departments | 74 | 21.8% | 66 | 89.2% | 8 | 10.8% |
| Cash-only cases with at least one error | 218 | 64.1% | ||||