Printer Friendly

Opposite of Correct: Inverted Insider Perceptions of Race and Bankruptcy.

ABSTRACT

Previous data collected during the 2007 meltdown of the subprime mortgage market showed that African Americans were approximately twice as lively to file chapter 13 bankruptcy than persons of other races, a significant policy issue given the generally less generous rules in chapter 13. We first update and replicate these findings with new data collected during 2013-2014 as the housing market recovered. Results of the original study were not specific to the subprime crisis as the new data showed the same 2:1 racial disparity as the older data, suggesting that this disparity may be a relatively enduring part of the U.S. bankruptcy system.

To see if insiders were aware of this disparity, we sent surveys to a national random sample of consumer bankruptcy attorneys. They seemed to believe there was a racial disparity in bankruptcy, but they had the disparity exactly backwards. A majority (about 60%) believed that whites were twice as lively to file chapter 13 when it is actually African Americans who are twice as lively to do so. We also report on a convenience sample of respondents from Amazon Mechanical Turk. These respondents have little or no knowledge of the U.S. bankruptcy system but lively hold common American stereotypes about which groups tend to be responsible or irresponsible. Their responses were largely similar to those of bankruptcy professionals. Overall, Mechanical Turk respondents were slightly more accurate than bankruptcy professionals. The lively reason is that Mechanical Turk respondents stuck more closely to the national base rate.

Across studies, it seems that those inside the bankruptcy system have little knowledge of the racial disparities that exist within it. Instead, they rely on common American stereotypes about who are responsible or irresponsible citizens.

I. INTRODUCTION

Previous work conducted during the 2007 subprime mortgage crisis found that African-American bankruptcy debtors are approximately twice as likely to file chapter 13 than their white counterparts. (1) The disparity remains even after controlling for a variety of legal, financial, and demographic factors that should determine bankruptcy chapter choice, (2) a finding we replicate later in this paper with data from 2013-14 as the housing market recovered. A vignette study in the previous work sent to consumer bankruptcy attorneys, where the hypothetical clients differed only on race and chapter preference, found that attorneys were plausibly an important part of creating this effect, by steering African Americans toward the more disadvantageous chapter 13. (3) Our previous work thus suggests that the professionals working in the bankruptcy system may contribute to the racial disparity in chapter choice.

However, are professionals working in a legal system aware of inequities within that system? We accept as natural that people within a system will have different explanations for why inequities occur. For example, prosecutors and defense attorneys may have different opinions on why African Americans are more likely to end up in jail than their white counterparts are. Employment discrimination plaintiff and defense lawyers may have different views about why women are less likely to be in upper management. However, we commonly assume that all participants within a system know that African Americans are overrepresented in prison cells and women are underrepresented in the corner office. If the facts on the ground are plain enough, we assume people will notice them.

In the present paper, however, we show that this is not necessarily the case. Practicing attorneys fail to notice the 2:1 racial disparity in chapter choice between African Americans and whites. Moreover, they in fact believe that the 2:1 disparity runs in the opposite direction. That is, although African Americans are twice as likely as whites to file under a bankruptcy chapter that generally requires them to pay back at least some of their debts, attorneys believe that it is whites who are twice as likely as African Americans to do so.

We argue that the actual racial disparity is not noticed and perceptions are in fact flipped because the beliefs of bankruptcy professionals are driven not by their actual experience but by common American stereotypes about which groups comprise the most responsible citizens. The stereotypes are so common that laypeople outside the system--who know nothing about the bankruptcy courts--have much the same (mistaken) expectations as bankruptcy insiders do. In fact, to the extent that laypeople end up simply using the base rates and ignoring the stereotypes, their predictions become more accurate than those of insiders. The stereotypes and expectations are so powerful that they will not only cause those inside the system to miss a gaping racial disparity, they flip beliefs and cause even insiders to see the world as the opposite of what it is.

Because chapter 7 and chapter 13 debtors have very different experiences in bankruptcy, the disparity matters. The mean attorney's fee in a chapter 7 is currently around $1,200 as compared to $3,200 for a chapter 13. (4) Although chapter 7 debtors tend to have lower incomes than chapter 13 debtors, the attorney's fee is less than 50% of the typical chapter 7 debtor's monthly income as compared to an entire month's income for the typical chapter 13 debtor. (5) When it comes to distributions to creditors, more than 90% of all debtors who file chapter 7 have no assets to turn over to the bankruptcy trustee. (6) In our previous work, the chapter 13 debtors proposed plans paying an average of 28% of their unsecured debt. (7) Nearly every chapter 7 debtor receives a bankruptcy discharge as compared to half of chapter 13 debtors. (8) Chapter 13, however, can certainly be advantageous for some debtors--those with valuable assets to lose if they filed chapter 7, those with substantial mortgage arrearages, or perhaps those needing the somewhat broader discharge available in chapter 13. But, the typical debtor who can file chapter 7 generally will be better off in chapter 7-There is no obvious reason to think chapter choice should break along racial lines, especially after controlling for demographic, legal, and economic factors that might legitimately make one chapter better than the other.

This paper continues as follows. Part II discusses the long-running and well-known divergence in local practices surrounding chapter 13 and discusses the academic literature that has studied chapter 13 usage. In Part II, we also use new data to replicate our previous research showing a racial disparity in chapter 13 usage. These data suggest that the 2:1 racial disparity is a relatively enduring part of the U.S. bankruptcy system, as these new data collected during the housing recovery matched quite well with the earlier data collected during the meltdown. Part III discusses the psychological research that motivates our study of whether insiders recognize this disparity and then provides its methodology. Part IV discusses our results. Part V concludes with a discussion of our thoughts on what our results tell us about the racial disparity in bankruptcy chapter choice and the ability of those within the system to even perceive disparity.

II. DIFFERENCES IN BANKRUPTCY CHAPTER CHOICE IN THE COURTS AND THE ACADEMIC JOURNALS

A. THE MANY CULTURES OF BANKRUPTCY PRACTICE

Although based on a congressional power to pass "uniform Laws on the subject of Bankruptcies throughout the United States," (9) the bankruptcy system is anything but uniform. One of the starkest differences is the percentage of bankruptcies filed as chapter 13 cases in different locales. Figure 1 shows the variation broken down by federal judicial district from 2013-2014. Over this time, the Western District of Louisiana had the highest percentage with 79.6% of all bankruptcies being filed as chapter 13 cases, a rate that was fourteen times as great as the 5.7% of bankruptcies filed as chapter 13 cases in the Northern District of Iowa.

The variation in chapter 13 rates is long-running and long-known. Stanley & Girth's Brookings Institution study, intended as a comprehensive statistical portrait of the bankruptcy system in the 1960s, noted the strong regional differences in chapter choice. (10) At a time when the national rate was 17%, Stanley 8c Girth observed chapter XIII rates varying from 4% to 76% in the seven judicial districts they studied. In trying to explain the regional disparities they found, Stanley & Girth attributed the differences to attitudes by local legal professionals--both lawyers and judges. They quoted one attorney saying chapter XIII was "morally correct," but they cited instances in other districts where judges and lawyers opposed chapter

XIII on principle as a bad deal for debtors. (11)

After the passage of the current bankruptcy law in 1978, several studies examined how the decision to file chapter 13 is made. As part of a sociological study on attorney-client interaction more generally, Neustadter sat in on consumer debtors consulting with lawyers on the decision to file bankruptcy and concluded that the decision-making process on chapter choice was decidedly attorney-centered. (12) Braucher explored local variations in a theoretically uniform federal bankruptcy system through a series of interviews with bankruptcy attorneys in Texas and Ohio. The differences in chapter choice were a major focus of her work. Summarizing her findings, Braucher wrote:

In a large, far-flung, locally administered system such as consumer bankruptcy it is inevitable that there will be local variation in practice and attitudes. But the current state of consumer bankruptcy is very far from the constitutional ideal of uniformity, and takes the form of a network of fiefdoms with significantly different customs.... Some lawyers use counseling to try to show most debtors that they could and should pay under chapter 13. Others influence debtors to use chapter 7 by emphasizing the debtors' lack of disposable income after realistic budgeting and the absence of material rewards for making repayment. While some debtors surely make their choices with some degree of autonomy, a substantial portion of them end up in one chapter or the other primarily because of the views of their lawyers. (13)

As Braucher was doing her work, Sullivan, Warren & Westbrook were exploring variations in the bankruptcy system, including chapter choice, using data from the 1981 and 1991 versions of the Consumer Bankruptcy Project ("CBP"). (14) With data from bankruptcy court records, they found that debtors' economic situation only partially explained differences in chapter choice. They hypothesized that local legal culture as well as social and demographic factors played a substantial role in affecting chapter choice. (15) Attorneys, judges, and other actors such as bankruptcy trustees perpetuated a local legal culture that either made chapter 13 relatively commonplace or relatively unusual. (16)

Thus, the first generation of scholarship documented the differences in chapter 13 filing rates and offered "local legal culture" as the explanation for why these differences occurred, but this scholarship did not explore what that "culture" might be. There were a few observations of possible racial disparity in chapter choice, but these mentions were anecdotal in nature, observed in passing in a paper focused on other matters, or not based on nationally representative samples. (17) In recent work, we examined racial disparity in chapter 13 use--as well as the role attorneys might be playing in it--through two national random samples, one of consumer bankruptcy debtors (taken from the 2007 CBP) and one of consumer bankruptcy attorneys. (18)

The first national random sample came from the 2007 version of the CBP, which collected information on 2,437 bankruptcy debtors through court records, written questionnaires, and telephone interviews. In this sample, African-American filers ended up in chapter 13 about 55% of the time, as compared to white, Latino, and Asian American filers who ended up in chapter 13 about 29%, 22%, and 24% of the time, respectively. (19) In fact, although protection of a home is widely believed to be the primary motive for filing chapter 13, African-American filers had higher chapter 13 rates than did homeowners (47%).

The disparity of African Americans being more likely to file chapter 13 remained even after statistically controlling for a variety of (1) financial variables such as a filer's monthly income, assets, and debts, (2) legal variables such as attorney representation, prior bankruptcy filings, and chapter 13 filing rates in the district, and (3) demographic variables such as education, occupational prestige, marital status, and so on. The raw difference in African American vs. non-African American chapter 13 rates is thus quite large and unlikely to be explained away by confounding factors. (20)

Results from the second study came from a national random sample of 262 consumer bankruptcy attorneys to whom we had sent a hypothetical vignette about a couple wanting to file bankruptcy. Findings from the sample of consumer bankruptcy attorneys suggested that they were plausibly an important part of creating this effect, by steering African Americans toward the more disadvantageous chapter 13. The attorneys received a hypothetical fact pattern describing a couple who were considering filing bankruptcy. The fact pattern was pretested to ensure it was a tough professional call as to which chapter was most appropriate. The attorneys were asked which chapter they would recommend to the couple and to give their general impressions of them. The fact pattern portrayed the couple as sympathetic, church-going people who were in the work force but also had to cope with unemployment, medical expenses due to the husband's diabetes, and a few perhaps unwise purchases.

When the survey named the couple filing for bankruptcy as Reggie and Latisha, attorneys recommended chapter 13 about 47% of the time as compared to 32% of the time when the survey identified the couple as Todd and Alison. (21) These figures were not strikingly different from (a) the overall real-world chapter 13 rates of African Americans (55%) and whites (29%) and (b) what a statistical model of the CBP data would predict for an African-American couple (50%) and white couple (28%) who had characteristics matching the given fact pattern.

Why attorneys would recommend chapter 13 more for African-American clients is an open question. On the one hand, attorneys might "know" that African Americans are twice as likely as whites to file chapter 13 and assume this reflects some sort of cultural preference--perhaps a greater preference to save their home. There is some evidence for this, as there was an overall tendency for attorneys to presume Reggie and Latisha would prefer chapter 13 more than Todd and Alison would. On the other hand, their steering might reflect some sort of unconscious bias about African Americans, who might be presumed irresponsible/untrustworthy--and hence would otherwise try to take the "easy way out" with chapter 7. There is some evidence for this last proposition as well. That is, attorneys were likely to rate African American couples who expressed a preference for chapter 7 as lacking in good values and as less competent, relative to their ratings for other couples. Thus, Reggie & Latisha had to "earn" the respect of the attorneys by expressing a preference for chapter 13 in a way that Todd and Allison did not.

B. REPLICATION OF EARLIER RESULTS

The results from the consumer bankruptcy debtors in the 2007 CBP data along with the survey of attorneys provide converging evidence of racial disparity. Still, the 2007 real-world data occurred as the subprime mortgage market melted down. Thus, in addition to the usual sound social-science practices that argue in favor of replication whenever possible, there may be added reasons to ensure that results from the 2007 data would replicate with more recent data as the housing market recovered. One of this paper's authors helped to relaunch the CBP in 2013 (the "Current CBP"), giving us an opportunity to see if the previous racial disparity in chapter 13 rates would replicate in a more "normal" (non-crisis) period, when housing prices were not crashing. By 2013, national home prices had begun to rise again, (22) and delinquency rates on home mortgages already were falling. (23)

The Current CBP began in February 2013. Every three months, 200 bankruptcy cases are drawn from a national random sample of bankruptcy filers using the federal courts Public Access to Court Electronic Records (PACER) system. Data from the first eight rounds of data collection (ending November 2014) were gathered for the present paper (n = 1,600). Legal and financial variables were collected from court records for all cases. In addition, filers received an eight-page survey in the mail that they could also complete online. The survey collected demographic information, reasons for filing, emotional reactions to bankruptcy, and hardships suffered before bankruptcy. Filers received a $50 gift card for completing the survey. The response rate for the surveys was 28.8% (n = 461), down from the 2007 response rate of 48.1%, perhaps due to the lack of funding to follow up with telephone calls to nonrespondents. Respondents and nonrespondents did not exhibit statistically significant differences across the major financial variables collected from court records (e.g., debts, assets, income, debt-income ratios), reducing the possibility of response bias, though we cannot rule it out completely. In the statistical analyses reported below, all legal & financial variables (except "foreclosure as a reason for bankruptcy'1) were collected from court records. The district chapter 13 rate was collected from official statistics of the federal courts. All other variables were collected from the surveys.

The new data revealed that African Americans (n = 120) again filed chapter 13 (vs. chapter 7) at much higher rates (55.0%), compared to people of other groups (26.1% of whites (n = 287) and 28.8% of Latinos (n = 52)) who filed under chapter 13. There were only fifteen Asian-American households. The African-American chapter 13 rate of 55.0% is extremely similar to the 54.7% rate from the 2007 data. In addition, the chapter 13 rate of 26.3% for households where no one identified as African American is similar to the 28.2% figure from 2007-The 2:1 difference between African American and non-African American households in filing chapter 13 was again statistically significant (chi-squared = 32.51, p < .001).

Also, as in 2007, it is possible that race merely covaries with other attributes that affect whether one files chapter 13. Importantly, it could be the case that African Americans merely happen to live in places where the local legal culture leads to predominantly chapter 13 filings. To rule out these possibilities, we conduct a logistic regression where the dependent variable is whether the debtor filed chapter 13. (24) The regressions are as similar as possi' ble to our regressions for the 2007 CBP with a few exceptions necessitated by the data. First, smaller sample sizes make it impractical to compute a non- African American chapter 13 rate for the judicial district, and instead we use the overall chapter 13 rate for that district (as recorded by the Administrative Office of the U.S. Courts). Second, occupational prestige score and credit card debt are not computed in the new data. Third, the Current CBP questionnaire has slightly different wording for where the debtor self reports "attempts to work with creditors" prior to bankruptcy. (25) Table 1 reports an abbreviated form of the regression results with the full results reproduced in Appendix 1.

The 2007 data collected as the housing crisis took off and the post-crash 2013-14 data give remarkably consistent results. In fact, the odds ratio for the African American household variable in the full model from the Current CBP, 2.39, is extremely close to the odds ratio for this variable in the full model from the 2007 CBP, 2.46. These odds ratios imply that, with other variables at their mean values, in both periods, African Americans were twice as likely as people of other races to file for chapter 13 (vs. chapter 7), even after controlling for possibly confounding factors.

The survey response rates are lower in the Current CBP. Again, however, respondents and nonrespondents did not differ on major financial variables collected from the court records. Further, we are reassured that the estimates for African-American and non-African-American filing rates in the more recent data are so strikingly similar to those gathered in the 2007 survey that had a much higher response rate. Overall, the results strongly suggest that the 2:1 racial disparity found in the 2007 data was not a product of the particular circumstance of a subprime housing market crumbling or some other artifact of when the data were collected. Rather, the Current CBP data suggest that the 2:1 racial disparity persists even as housing prices are recovering. Of course, data from more periods is needed to establish further continuity, but it is clear that the 2:1 racial disparity in chapter 13 rates was not a one-off event.

In summary, the prior literature established wide variation in the rate of chapter 13 cases across the country and found that African Americans are twice as likely to file chapter 13 than people of other races. In our prior research, this difference did not go away after controlling for possibly confounding factors that might influence chapter choice. These results replicated with more recent data from 2013-14-Furthermore, previous literature and our own research found that attorneys play a significant role in helping to guide clients in their chapter choice. In an experimental vignette study, we found that attorneys were more likely to recommend chapter 13 when the client was perceived as African American as opposed to white. The insiders to the bankruptcy system thus have a "front-row seat" to a stark racial disparity. The remainder of this paper explores whether they recognize it.

III. MEASURING ATTORNEY PERCEPTIONS OF DISPARITIES IN CHAPTER CHOICE

It was an open question whether the insiders would see the disparity. On the one hand, we assume that people generally have perceptions that coincide with reality--or at least if these perceptions are inaccurate, random errors should average out when aggregated. Any given attorney may not have a good picture of the system overall, but aggregated together, their judgments should be relatively on target. Our previous random sample of attorneys making recommendations for an African-American couple versus a white couple reproduced the racial disparity in the CBP data reasonably closely. Attorneys might indeed be aware of the racial difference in chapter 13 rates and have simply been reflecting it back in their recommendations to the couple.

There is, however, a more troubling possibility. That is, it is possible that attorneys--despite having a front row seat to the disparity and recreating this disparity in recommendations to the hypothetical African American vs. white couples--are unaware that the disparity exists. Although there is a real-world 2:1 racial difference, attorneys simply may not pick it up. Stereotypes can be quite powerful and can prevent us from detecting even decent-sized patterns of covariation if those patterns do not align with expectations. (26)

To the extent that attorneys do not know if there is a racial disparity and are aware that they do not know, we would expect them to simply go with the base rates and assume people from all racial groups file at approximately the national average. However, stereotypes are seductive and can provide the illusion of knowing. Early demonstrations about how stereotypes can overpower base-rate information came from Kahneman and Tversky. In one of the first controlled experiments on base-rate neglect, participants were given a series of short personality descriptions and asked to predict which sub-population the described person belonged to. Regardless of whether they were told the person was sampled from a population of 70 engineers and 30 lawyers or 70 lawyers and 30 engineers, participants' guesses reflected the stereotypes of what lawyers and engineers are thought to be like, with minimal (though statistically significant) consideration of whether the description was taken from a sample that was mostly engineers versus mostly lawyers. (27) There have been a number of explanations proposed for why individuals favor stereotypical information over simply going with the base rates. According to Kahneman and Tversky, "people predict by representativeness, that is, they select ... outcomes by the degree to which (they) represent the essential features of the evidence." (28)

One common stereotype is that African Americans are untrustworthy or irresponsible. Though not as openly endorsed as it once was, the stereotype does seem to still have some place in contemporary culture even if it is not flagrantly vocalized or even consciously thought. (29) In the previous study, there appears to be some hint of this stereotype's influence among bankruptcy attorneys who seemed to view Reggie and Latisha as less competent and less likely to have good values if they expressed a preference for filing chapter 7, in a way that did not occur when Todd and Alison expressed a preference for chapter 7-30 If--like most Americans--some attorneys do have unconscious stereotypical beliefs about African Americans as irresponsible or untrustworthy, we might expect that this would influence their judgments about which groups are likely to file under chapter 13. If this holds, we would expect to find that attorneys (a) do not apprehend that that there is a 2:1 racial disparity, (b) do not recognize their own ignorance and simply go with the national base rate, but instead (c) use common stereotypes about who is responsible/trustworthy vs. irresponsible/ untrustworthy to underpredict African-American chapter 13 filings and overpredict white filings.

Using Westlaw, we generated a list of bankruptcy cases filed on five different days of the week in late July and early August 2011 (approximately 1 year before our previous research was published). The result was a list of 23,341 bankruptcy filings from which a random sample of 400 attorneys were selected. Attorneys who had been respondents in our previous study or who were in the same law firm as another survey recipient were excluded.

The survey informed respondents that "we are interested in the perceptions of bankruptcy professionals about how often people from different groups file chapter 13 (vs. chapter 7)." The cover letter informed respondents that they were randomly selected from a group of cases where they had appeared as an attorney of record and contained disclosures typical for human-subjects research. A follow-up letter and another copy of the survey were sent approximately four weeks after the initial mailing.

To minimize the possibility of confusion and to give respondents the national chapter 13 rate, survey instructions told respondents the first question had been completed for them based on the then-current data: "Because 28% of the nations bankruptcy cases currently are chapter 13, we wrote 28 for the 'nation as a whole.'" Finally, we informed respondents that we were interested in their estimates to the best of their knowledge and asked to make their best guess even if they were unsure.

The survey asked respondents to estimate the percentage of cases that were filed as chapter 13 for:

* Men, filing alone

* Women, filing alone

* Persons 65 years old and over

* Homeowners

* Persons with a prior bankruptcy filing in the last 8 years

* Whites

* African Americans/blacks

* Hispanics/Latinos

* Asian-Americans/Asians

* Pro se debtors

* Southerners (from the states of AL, AR, FL, GA, LA, MS, NC, SC, TN, TX, VA)

The survey concluded by asking for basic demographic information: the federal judicial district in which the respondent practiced, year of birth, gender, and race/ethnicity of the respondent.

We heard back from 38.3% of our attorney sample. Of the nine surveys returned blank, seven had written comments indicating the attorney was not providing any estimates because his or her practice involved no or very few chapter 13 cases. Our respondents were 84.3% male with a racial breakdown of 91.0% white, 1.5% African American, 6.0% Hispanic/Latino, and 1.5% Asian-American. The mean age was 49.4 years with a median of 52 years. These characteristics were very similar to the respondent pool of bankruptcy attorneys in our previous work.

IV. PERCEPTIONS OF RACIAL SORTING IN BANKRUPTCY CHAPTER CHOICE

A. Attorney Responses

Our research question asks what bankruptcy professionals knew about the racial disparity in bankruptcy chapter choice. By gathering estimates for the percentages of bankruptcies that were filed as chapter 13 cases, the survey instruments captured one side of this question, but what we want to know is the extent to which these estimates were accurate. To assess the estimates' accuracy, we need to know the real-world percentage at which different groups choose chapter 13.

No government agency regularly collects demographic information on bankruptcy filers. Therefore, we use the detailed demographic data available from the 2007 CBP collected from a national random sample of 2,437 bankruptcy cases. The drawback of using the CBP data is the passage of time from its data collection in 2007 to our surveying bankruptcy attorneys in 2011. (31) At the time of the CBPs data collection, 38.7% of all bankruptcy petitions were filed as chapter 13 cases, but the chapter 13 rate had fallen to 27-9% at the time of the data collection with attorneys in 2011. Also, for our respondents, we provided a baseline estimate that 28% of bankruptcy petitions for the nation as a whole were chapter 13 cases. Although the overall rate had fallen, there is no reason to think that the demographic composition of chapter 13 filers (which was similar in the 2013-2014 CBP to what it was in the 2007 CBP) was different in 2011, though we cannot rule this out. Given the drop from the 2007 measurement of actual usage to the 2011 survey and national base rate we provided, we would expect attorney estimates to be slightly lower than they might have been had attorneys been surveyed in 2007.

Table 2 compares the estimates from bankruptcy attorneys with the realworld chapter 13 data from the 2007 CBP. The accuracy of estimates (column c) shows three clusters, with greater variation between clusters than within clusters. A large cluster of groups in the middle has mild underestimates of chapter 13 usage. On average, the probabilities given for the middle cluster of groups underestimate chapter 13 usage by 12%--which is extremely close to what one would expect given the 11% drop in overall chapter 13 use from 2007 to 2011. The groups in this middle cluster are: Hispanics, women filing alone, senior citizens, men filing alone, southerners, accuracy rather than those from the more recent study because of the larger sample size in the 2007 data. Results are similar if we use either the more recent data alone or average it with the earlier data. Asian Americans, and pro se debtors. Four groups fall outside this middle cluster, though.

Among all groups we asked about, for only two groups is there an overestimate of chapter 13 use: whites (+17.8%) and homeowners (+7.3%). At the other extreme, attorneys substantially underestimated chapter 13 usage for African Americans (-32.6%) and prior bankrupts (-47-4%).

Considering the results of the previous studies, the attorney underestimate for African Americans stands out. From the real-world data in both 2007 and 2013-14, we found a large, 2:1 racial disparity in chapter choice, which one would think would be difficult to miss. In aggregate, attorneys do miss it. Instead, their perceptions of who uses chapter 13 seem to follow general cultural beliefs about which groups are responsible. In the case of race, this means an actual 2:1 racial disparity of African Americans being more likely to file chapter 13 as opposed to whites becomes flipped to a perception of a 2:1 difference that whites are more likely to file chapter 13 as compared to African Americans.32

Breaking the data down a bit differently, we can compute an "ipsative" (within-person) score for each attorney as the attorney's estimate for African Americans divided by the attorney's estimate for whites. Only 6% of attorneys estimated that the ratio of African American to white chapter 13 filings was 2:1 or greater, whereas 59% of attorneys had flipped perceptions of the world, believing that the white to African American ratio was 2:1 or greater. Overall, 71% of attorneys thought the chapter 13 rate was higher for whites, 13% thought it was equal, and 16% thought it was higher for African Americans.

The attorney estimates appear to line up with American stereotypes of who is a responsible person. Attorneys see homeowners and whites as more likely to attempt repayment to creditors through chapter 13 whereas they see African Americans and prior bankrupts as more likely to seek a full discharge in chapter 7-There is no reason to think that attorneys, as a group, should not hold prevailing American cultural biases about who is a responsible person. That the attorneys are the system's insiders and have a front-row seat to witnessing a 2:1 racial disparity is not enough to overcome the bias.

If attorneys are largely unaware of the systematic tendency of African Americans to end up in chapter 13 and tend to counsel African Americans into chapter 13 (unaware that others are doing the same thing), the attorney estimates fit well with the findings in our previous work. "Reggie and Latisha" had to earn the right to be viewed as good, competent people by expressing a desire to file chapter 13. If attorneys believe that most African Americans irresponsibly file chapter 7, then "Reggie and Latisha" who want to file a chapter 7 are just one more couple trying to get off the hook. However, if they express a desire for chapter 13, the attorneys will view them as competent people of good values.

The underestimation for prior bankrupts is also quite interesting. On this point, the language in the survey instrument was very specific, asking attorneys to estimate the percentage of bankruptcies that were chapter 13 cases for "persons with a prior bankruptcy filing in last 8 years." In the 2007 CBP data, persons filing for bankruptcy in the previous eight years had the highest percentage of chapter 13 use of any group we identified in the data. The reason for high chapter 13 usage has a legal explanation behind it. A person receiving a chapter 7 discharge is ineligible for another discharge for eight years, and a person receiving a chapter 13 discharge is ineligible for another discharge for six years (with some exceptions). (33) Because the primary benefit from chapter 7 for a consumer debtor is the discharge, a person with a discharge in his or her immediate past will find little benefit in a chapter 7 filing. (34) Although a chapter 13 discharge might be similarly barred for the same person, persons trying to save a home or other asset might find use for chapter 13 rules that allow the payment of arrearages over time or the curing of defaults that occurred before bankruptcy. The real-world data bear out these advantages. Although only 11.8% of the debtors in the 2007 CBP reported a prior bankruptcy in the eight years before filing, 83.2% of the prior bankrupts filed in chapter 13. Persons with a prior bankruptcy are likely to be considered irresponsible. When combined with the large underestimation of chapter 13 usage for African Americans, the similarly large underestimation for prior bankrupts supports the idea that attorney estimates appear to reflect American cultural biases about who is a responsible person.

A consumer bankruptcy attorney who reflected on his or her estimates might be likely to realize that prior bankrupts must have high chapter 13 usage. Twenty-four or 17-8% of our respondents overestimated the 83.2% chapter 13 rate for prior bankrupts. These respondents provide a useful subgroup for analysis because they demonstrated both very careful attention to the survey instrument and, given their estimate for the group with the highest chapter 13 rate, possibly a propensity to overestimate chapter 13 use. Yet, these respondents displayed a pattern similar to the overall responses. They overestimated the chapter 13 rate for whites--the only other group for whom they overestimated chapter 13 rates--by a mean of 10.8%. At the same time, they underestimated the chapter 13 rate for African Americans by a mean of-30.7%. The mean underestimation for all other groups ranged between-0.6% and-8.9%. Thus, although this subset of respondents recognized the obvious legal implications for chapter 13 rates of prior bankrupts, they still displayed attitudes about chapter 13 that seem to reflect the same broad American cultural biases. (35)

B. EXPERIENCE: HIGH CHAPTER 13 DISTRICTS AND AGE

One question is whether experience might improve the accuracy of attorneys' estimates. We addressed this question by examining two variables: the chapter 13 rate in the district and the age of the attorney (used as an imperfect proxy for years of legal experience). As will be seen below, the evidence about whether experience improves accuracy was mixed. At best, it only slightly lessened attorneys' use of stereotypes, still leaving them with an inverted perception of the difference between African Americans and whites.

Age. Because we did not ask about years of practice, we could only use attorney age as a proxy for this variable. Surprisingly, age had a negative relationship to overall accuracy, as younger respondents gave more accurate estimates across groups (r = .20, p = .01). It appears that younger respondents' greater accuracy derived from their greater reliance on the national base rate. That is, younger attorneys were more likely to stick closer to the national base rate of 28 percent in their guesses. When we control for how closely attorneys stuck to this base rate, the effect of age is not statistically significant (p = .47). The ultimate interpretation of the age effect is still unclear, though. It could be that newer attorneys preferred to use the national base rate because this younger cohort is more reluctant to use group-based stereotypes. Or it could be that younger attorneys stuck to the base rate simply because they knew how much they did not know and were hence reluctant to guess. Regardless of the reason, however, the implication is clear: experience (as proxied by age) does not seem to lead to greater accuracy across groups.

As with accuracy in general, age did not lead to greater accuracy with respect to the four groups for whom stereotypes dominate perceptions. The correlations of age with accuracy of estimates were -.1 (p = .25), -.17 (p = .05), .01 (p = .92), and -.04 (p = .67) for whites, homeowners, African Americans, and prior bankrupts, respectively.

High chapter 13 districts. Inferences about the effects of experience become slightly more optimistic when we examine the effect of practicing in a district with a high rate of chapter 13 cases. However, even here, experience seems to be no match for stereotypes.

The survey instrument asked respondents to estimate groups' rates of chapter 13 filing across the nation as a whole. Many attorneys might not have experience with chapter 13 beyond their locale. Other consumer bankruptcy attorneys might specialize in chapter 7 and thus not have much knowledge about chapter 13 even in their own locale. Seven attorney respondents returned blank surveys with comments that they did not do chapter 13 bankruptcy.

Attorneys practicing in areas with the highest usage of chapter 13 should be the attorneys most knowledgeable about it and hence less prone to errors when estimating chapter 13 rates. To assess whether attorney familiarity with chapter 13 led to better estimates of chapter 13 usage patterns, we divided the sample into two groups. The first group contained attorneys who practiced in federal judicial districts where the percentage of bankruptcies filed as chapter 13 cases was below the national average. The second group contained attorneys who practiced in federal judicial districts with chapter 13 rates above the national average. Table 3 reports the results.

Across most groups, attorneys in districts above the mean chapter 13 rate tended to make better estimates than attorneys in districts below the mean. The raw correlation between the chapter 13 rate in a district and the accuracy of attorneys' answers across groups was r = .29, p=.001. This accuracy cannot be accounted for by attorneys sticking close to the national base rate, because the correlation between district rate and accuracy remains even after controlling for any such tendency (partial r for district rate = .36, p = .001). We interpret this correlation between district rate and accuracy as some indication of the survey instrument's validity in that familiarity with chapter 13 did lead to more accurate answers.

As with accuracy in general, district rate of chapter 13 use led to greater accuracy with respect to the four groups for whom stereotypes dominate perceptions. The correlations of district rate with accuracy of estimates were .32 (p = .001), .30 (p = .001), .45 (p = .001), and .14 (p = .11) for whites, homeowners, African Americans, and prior bankrupts, respectively.

Still, the same basic pattern as with the overall data emerges for those practicing in both judicial districts above and below the mean. Relative to their estimates for other groups, attorneys overestimated chapter 13 filing rates for stereotypically responsible groups (whites and homeowners) and underestimated the rates for stereotypically irresponsible groups (African Americans and prior bankrupts).

Again, breaking the data down ipsatively (within-person), in districts where the chapter 13 rate was below the national mean, 70% of attorneys thought the white chapter 13 rate was at least twice as large as the African- American rate, whereas only 3% thought the African-American rate was at least twice as large. In districts where the chapter 13 rate was above the national mean, 43% of respondents thought the white rate was at least twice as large as the African American rate, whereas only 12% thought the African American rate was at least twice as large as the white rate. In these high chapter 13 districts, overall, 57% of attorneys thought the white rate was higher than the African American rate by any amount, 12% thought the rates were equal, and 31% thought the African American rate was higher. Thus, regardless of the local chapter 13 rate, attorneys seem to miss glaring racial differences and seem instead to be guided by American cultural stereotypes about who is responsible or irresponsible. The differences in estimate errors remain large such that a broad range of attorneys have an inverted perception of the world.

C. ROBUSTNESS CHECK: CONTROLLING FOR "INNUMERATE" RESPONSES

We asked our respondents to "estimate the percentage of bankruptcies in that group that were filed as chapter 13s." A concern was that respondents would actually perceive a different question, namely what percentage of chapter 13 bankruptcy filers have the characteristic listed. For example, what percentage of chapter 13 bankruptcy filers are men filing alone, homeowners, African Americans, and so forth? To minimize the possibility of miscommunication, we labeled the column of estimates respondents needed to complete as "Percentage of bankruptcies that were chapter 13." We also completed the first entry for the "Nation as a whole" with an estimate of 28% with a textual explanation that we had completed the first entry this way "because 28% of the nation's bankruptcy cases currently are chapter 13s."

Despite these precautions, it is possible that some respondents still misunderstood these instructions. To identify responses where there was a possible misunderstanding of the survey question, we focused on the four racial categories and adopted two strategies. First, we identified any response where all four racial categories were estimated above or below the stated national average of 28%. (36) Second, we identified any survey response where the estimates for the four racial categories summed to between 90% and 110%. (37) For short-hand reference, we refer to responses not exhibiting these characteristics as "numerate" responses and then separately examine the results for these "numerate" responses. We recognize that some persons identify as both Hispanic/Latino and white or African American as well as the existence of racial identifications other than those provided to respondents. A response exhibiting an "innumerate" pattern was not necessarily logically impossible. Rather, we use these heuristics to identify respondents that were the most likely to have misunderstood the survey question, and we exclude them from analyses as a type of robustness check on our data.

Table 4 reports the results excluding responses exhibiting our first definition of "innumeracy," namely responses where the racial categories were all either above or below 28%. The same general pattern emerges as in the overall data. Whites and homeowners remain the only two groups whose chapter 13 rates are overestimated. Using just the responses from attorneys who were most likely not to have misunderstood the survey question actually produces an increase in these overestimation errors. Similarly, African Americans and prior bankrupts remain the groups with large underestimations of chapter 13 usage. For space, we have not separately reported the results using the second definition of "innumeracy," where the racial categories summed to between 90% and 110%. Using this definition, the basic conclusions stay the same. Whites are now the only group overestimated at +10.0%; homeowners are slightly underestimated at -0.5%. African Americans (-34.6%) and prior bankrupts (-42.7%) remain the only groups with vast underestimates.

D. CONVENIENCE SAMPLE OF MECHANICAL TURK WORKERS

We believe that the pattern of overestimates and underestimates from the attorneys reveals that they are overwhelmingly unaware that African Americans are twice as likely as whites to file chapter 13 versus chapter 7Further, their perceptions have flipped reality on its head, and a majority (about 60%) guess that whites are twice as likely as African Americans to file chapter 13. Although attorneys are insiders in the bankruptcy system, they perceive exactly the opposite of what is happening in the system-at least in terms of African-American and white chapter 13 filing rates. Our hypothesis is that attorneys-though part of the system-base their perceptions not on actual experience or inside knowledge, but they instead rely on common American stereotypes about which groups are responsible and which groups are not.

If it is indeed true that attorneys' inaccurate estimates are produced by the application of popular stereotypes (rather than, say, beliefs specific to the legal community, some source of error common only to bankruptcy attorneys, or some other source of insider "knowledge"), then it should be possible to show that the attorneys' errors can be reproduced by a completely naive population that has no knowledge of the bankruptcy system and can only make guesses based on their stereotypes. We explore this prediction here, using a convenience sample of workers on Amazon Mechanical Turk ("MTurk"), a service that identifies persons who are willing to perform online human intelligence tasks such as taking part in university research.

There is no reason to think MTurk workers are representative of the U.S. population. In fact, they are not. Among other differences, they tend to be more politically liberal (and thus likely more similar to debtor's attorneys) and younger (and thus less likely to use stereotypes), compared to the U.S. population as a whole. (38) Nevertheless, we suspect that the racial stereotypes leading to the mistaken estimates in Study 2 are suitably widespread that MTurk workers will still have the same erroneous belief that whites are much more likely to file under chapter 13 than African Americans are.

We administered a survey to 201 MTurk workers. The survey began with a brief explanation of the difference between chapters 7 and 13. Respondents were then asked to give probabilities of filing chapter 13 for people in various groups. Half the participants were asked in an open-ended format to give the percentage of filers in that group who filed chapter 13. The question format here was equivalent to the question format given to attorneys. Additionally, in an attempt to reduce any innumeracy problems, the other half of the respondents were given sliding scales. For each group, respondents used sliders to indicate the percentage of each group filing chapter 13 and the percentage of each group filing chapter 7-They were told that the two sliders should add up to 100 percent (the sum was automatically calculated by the survey program).

Orthogonal to the response format manipulation, half the respondents were also incentivized to give correct answers, being told that "if your guesses for all the groups were within 10 percentage points of the correct answer, we will double the payment to you." The other half of respondents were not given this incentive (similar to the attorneys, who were not given incentives). Overall, the scale format did not appear to improve accuracy (p = .21), though the incentive slightly did (p = .06, average distance from true value in incentive condition = 21.94 vs. average distance from true value incentive condition = 23.49). However, for the key questions asking about African-American and white chapter 13 rates, neither format nor incentive had a statistically significant effect on accuracy (both p's > .24). In Table 5, we collapse across question formatting and incentivizing condition.

As may be seen in Table 5, MTurk workers gave estimates for whites and African Americans that were not significantly different from those given by the attorneys in the first study (MTurk vs. attorney estimates for whites, p = .18; MTurk vs. attorney estimates for African Americans, p = .11). The MTurk workers on average overestimate the chapter 13 filing rates for whites (a stereotypically responsible group) and underestimate the filing rates for African Americans and prior bankruptcy filers (two stereotypically irresponsible groups). They do not seem to overestimate the chapter 13 filing rate for homeowners (a stereotypically responsible group), though they do seem to underestimate the filing rates for Southerners (a group that is also sometimes stigmatized by outsiders).

Ipsatively (within respondent), 75% of MTurk respondents thought the white chapter 13 rate was higher than the African-American rate, with 48% believing the white rate was at least twice as high as the African-American rate. On the other hand, 21% of respondents believed the African-American rate was higher than the white rate, with only 5% of respondents believing the African-American rate was twice as high.

The wisdom of the crowd and the wisdom of knowing one's ignorance. For seven categories, the average MTurk estimate was significantly different from the average attorney estimate. For six of these seven differences, the aggregated estimates of MTurk workers were closer to the actual data from the CBP than were the aggregated estimates of attorneys. The exception was the estimate for prior bankruptcy filers where the legal nuances of the prohibitions against successive bankruptcy filing would be most likely to give legal professionals an advantage. (39)

The overall superiority of our relatively naive MTurk workers to the bankruptcy attorneys is quite surprising. (40) One could perhaps attribute this result to the "miracle of aggregation" or the "wisdom of the crowd," in which aggregated judgments are accurate because the random errors of relatively uninformed people cancel each other out. However, one has to wonder why the wisdom of a crowd of MTurk workers exceeds the wisdom of a crowd of experienced bankruptcy attorneys. It is possible this is due simply to the size of the crowd being bigger for MTurk workers (n = 201) than for attorneys (n = 144). The bigger the crowd, the more likely random errors will cancel out and the estimates will converge on the true number.

However, we think another answer is likely. It was not simply the case that the aggregate judgment of MTurk workers was better than the aggregate judgment of attorneys. It was also the case that the individual judgments of MTurk workers were better than the judgments of individual attorneys. For each individual respondent, we looked at the absolute distance between the respondent's estimate and the CBP data. Using absolute distance does not allow random errors to cancel out, because absolute distance is a function of the size of the error rather than size and direction. Across the eleven categories, our relatively naive MTurk respondents were significantly more accurate than the attorneys were (mean absolute difference for MTurk workers = 22.72 vs. mean absolute difference for lawyers = 24.11, p = .05).

Why individual MTurk workers would do better than individual bankruptcy attorneys seems clear from looking at another aspect of the data. Individual MTurk workers were better than individual bankruptcy attorneys because they stuck closer to the base rate of 28%. For each individual, we looked at the absolute distance between their guess for each category and 28%. Each individual's score was the average of the distances across the eleven categories. For MTurk workers, this average distance from the overall base rate of 28% was 17%, whereas for bankruptcy attorneys it was 21% (p = .001). This difference largely explains why MTurk workers were more accurate. Using a simple regression analysis, MTurk workers are more accurate (p = .05). However, when we add to the regression how far the estimate was from the base rate, the distance from the base rate is significant (p = .001), whereas the effect of being an MTurk worker completely disappears. In fact, it reverses. Thus, MTurk workers stuck to the base rate more than the attorneys did. Once we account for this tendency to stick to the base rate, the MTurk users did no better and in fact marginally worse than bankruptcy attorneys. Thus, it appears that our relatively naive judges (MTurk workers) outperformed our experts (bankruptcy attorneys) solely because they realized how much they did not know and just stuck to the national base rate.

This finding resonates with research in psychology on the fallibility of clinical judgment as well as the fallacy of base-rate neglect (a fallacy that MTurk workers were less likely to commit than our experts were). (41) More generally, the findings augment the larger point we make in this article about the need for objective data on disparate effects within the justice system and how, without such data, peoples' preconceptions bias not simply their understanding but also their very perception of the facts on the ground.

Summary. In sum, MTurk workers had relatively less knowledge of the bankruptcy system than bankruptcy attorneys did. Yet, their responses stuck closer to the national average and consequently were generally more accurate than those of attorneys. More importantly for the present purposes, the same inverted perception of reality shown by attorneys was shown by our relatively naive MTurk workers: Although African American bankruptcy debtors file chapter 13 at a rate double that of whites, MTurk workers estimated that the white rate was almost double that of African Americans. As such, reliance on common American stereotypes about whites and African Americans seems to go a long way toward explaining the inverted perceptions of the bankruptcy system.

A general caveat. We have demonstrated that attorneys and MTurk workers all make similarly inaccurate judgments about the racial disparity in chapter 13 filing. That is, those inside the system have the same misperceptions that relatively naive outsiders do. However, there is a general caution that must be heeded. Regarding the bankruptcy professionals, it is possible that "that those who know (that the system has inequities), won't tell, whereas those who will tell, don't know."

We cannot rule out the possibility that our sample of bankruptcy professionals is influenced in this way by nonresponse bias. Our response rate (36%) was decent for a mail survey, but not so high that concerns about nonresponse can be dismissed. It remains a potentially important artifactual explanation for the results. However, we think a nonresponse bias explanation will only go so far. There are a few reasons for this.

First, even if all our nonrespondents had perceptions that were completely accurate representations of reality, it is still a bit disturbing to know that 36% of attorneys have, collectively, a flipped perception of reality. Thirty-six percent of attorneys believing that a 2:1 difference "favoring" African Americans is actually a 2:1 difference "favoring" whites represents a sizeable percentage of attorneys who do not know about the racial disparity that seems to exist within the system.

Second, the absolute size of the African American/white misperception effect is so large that the direction of the effect is likely to be reasonably robust to nonresponse biases. In our sample, attorneys estimated that 46% of white debtors filed chapter 13 whereas only 29% actually did so, according to CBP data. They also estimated that 22% of African American debtors filed chapter 13, whereas 55% did so according to the CBP data. What would the answers of the nonrespondents have to be to make the attorneys' collective estimates accurate? Simple algebra reveals that to make the sample as a whole accurate, all of the remaining 64% of nonrespondents would have to collectively estimate that whites file chapter 13 at a rate of 19% whereas African-Americans file chapter 13 at a rate of 74%. That is, the remaining nonrespondents would have to underestimate the actual white rate (29%) by one-third whereas they would have to overestimate the actual African American rate (55%) by one-third. While this might be the case, we think the possibility of such estimates, particularly the African-American overestimate, is not likely. In the surveys of the attorneys and MTurk workers, there is no filing rate estimate for a group that comes close to 74% (the estimate required to bring the entire sample of attorneys' perceptions of African Americans in line with reality).

Of course, it might be argued that attorneys' perceptions in the present paper are actually accurate and the estimates generated by the CBP data are off. We think this is unlikely. The 2007 CBP survey had a response rate of 47%-To bring the CBP data in line with attorneys' estimates (and assuming that the response rate of African Americans was no different than the general response rate in the CBP), the nonresponding African Americans from the CBP would have had to file chapter 13 at a rate of-7% (an impossible value). For whites, to bring the CBP data in line with attorneys' estimates (and assuming the response rate of whites was no different than the general response rate in the CBP), the nonresponding whites from the CBP would have had to file chapter 13 at a rate of 61%--not an impossible figure but one higher than that of any other large demographic group we have examined. The court-record data from the CBP respondents also was cross-checked against court records from the CBP nonrespondents. No major differences were found in their financial conditions, making it less likely there was any easily identifiable response bias in the CBP data.

It is also possible that there is some nonresponse bias in the CBP data and some nonresponse bias in the attorney perceptions data that would bring attorney perceptions in line with reality, but assumptions now begin to accumulate. That is, there would have to be a particular sort of nonresponse bias in the 2007 CBP data and a particular nonresponse bias in the 2013-14 CBP replication and a particular nonresponse bias in the Reggie & Latisha vs. Todd & Allison data and a particular nonresponse bias in the present survey of attorneys to make this alternate explanation plausible.

Here another caveat is in order. We have argued that our system insiders departed from reality-based judgments because they relied on biased perceptions about who is responsible and trustworthy and who is not. We draw our conclusions from the four groups for whom attorneys most diverged from reality (whites and homeowners on the one hand and African Americans and prior bankruptcy filers on the other). There could be other factors attorneys were thinking about, some of which would have led to greater accuracy and some of which would have led to less accuracy. For example, attorneys could have thought that African Americans were disproportionately likely to be financially-strapped homeowners caught in expensive subprime mortgages and desperately trying to save their home. That would likely have led them to greater accuracy in their judgments. They could have thought that African Americans were disproportionately cash-strapped and needed to file chapter 13 because they could pay their attorneys over time. That too would have led to greater accuracy. They could have been thinking that African Americans had fewer assets to protect. That would have led them to less accuracy in their judgments. If, however, attorneys thought about who had assets to protect--instead of who was responsible and trustworthy --they would have also likely given much higher estimates for prior bankruptcy filers (who are likely to use chapter 13 to pay arrearages on secured assets) and Asian Americans (who have comparable wealth and higher income than whites, though show more variability as a group). (42)

Further, the supposition that results are likely driven by stereotypes about who is responsible and trustworthy is most consistent with the data from previous studies. That is, it is consistent with attorneys thinking particularly poorly of Reggie and Latisha (vs. Todd and Alison) if they express a preference for chapter 7-It further helps explain why the church-going, relatively responsible Reggie and Latisha are generally presumed to prefer chapter 13 more than Todd and Alison--it plausibly derives from a contrast effect between the reasonably trustworthy exemplar couple and the stereotype about the general category of African Americans as untrustworthy. (43)

Though the "responsible and trustworthy" stereotype seems most plausible, the exact content of the stereotype attorneys were using could benefit from further research. What seems beyond question is that attorneys were using some common stereotype (at least, one they shared with MTurk workers) and that this stereotype prevented them from recognizing even a glaring 2:1 racial disparity that appears in real life.

Preconceptions are important for perceptions. (44) People see what they expect to see and do not see what they do not expect to see.

V. CONCLUSION

In this paper, we replicated our earlier finding on racial disparity in bankruptcy chapter choice and explored what bankruptcy professionals know about this disparity. Because of the serious financial consequences for the many people in bankruptcy, the overrepresentation of African Americans in chapter 13 bankruptcy should be a major policy concern. As compared to chapter 7 bankruptcy, African Americans tend to end up in a bankruptcy proceeding that takes longer, costs more, and usually leads to less bankruptcy relief. When they arrive in bankruptcy court, society appears to give African Americans less forgiveness than persons of other races.

It is relatively easy to imagine that various actors within the system would have differences of opinion as to why African Americans would be more likely to file chapter 13. They might believe that it resulted from the cultural preferences of African Americans, biases of attorneys, biases of other people in the system, differences in up-front costs, the geographic concentration of African Americans in high chapter 13 regions (the South), previous bankruptcy filings that would force one into chapter 13, or the confounding of any number of demographic or economic factors that happen to be related to race.

What is most striking in this paper, however, is not that various actors disagree on the cause of the difference. What is striking is that they do not even see the difference. Actors within the system do not even notice a 2:1 racial disparity in chapter 13 choice--and most in fact believe the disparity runs in the opposite direction. Although African Americans in bankruptcy are twice as likely to file under chapter 13 as whites are, bankruptcy attorneys believe that it is whites who are twice as likely as African Americans to do so. Their perceptions of racial disparity and the actual racial disparity are completely inverted.

Such is the power of expectations to drive perceptions. In general, it seems that stereotypically irresponsible groups (African Americans, prior bankrupts) are believed to file for the immediate discharge of a chapter 7, whereas stereotypically responsible groups (whites, homeowners) are believed to file for the repayment plans of chapter 13. Given only a brief introduction to what chapters 7 and 13 involve, relatively naive raters (on Amazon's Mechanical Turk) make guesses as to which groups file under which chapter that are strikingly similar to the guesses of professionals within the system (bankruptcy attorneys). If anything, the naive MTurk workers tended to be more accurate than system insiders about which groups filed what, solely because the MTurk workers knew how much they did not know and stuck closer to the national base rates.

The results have implications for general beliefs about how well insiders can know their system. It is generally believed that people with a front-row seat in the justice system will know the justice system and have privileged insight into how it works. This is undoubtedly the case in many matters. However, when it comes to matters in which there are powerful social stereotypes, we simply fail to notice (or mentally invert) the inequities appearing before us.

The data have implications for reform of the system in that they make clear that good measurement is needed. In other papers, we have argued that the Department of Justice or the federal courts should explicitly keep track of the race and other demographic characteristics of bankruptcy filers. We believe this paper also illustrates this point--actual data on who files what, rather than general impressions or subjective judgments are likely to tell us what is happening on the ground and whether improvements are being made toward a more equitable system.

As noted, the implications go beyond just bankruptcy. Any system is subject to effects in which insiders fail to notice disparities or even have perceptions that are the inverse of reality. This is not to say that outsiders must always police a system. Insiders may also be able to effectively monitor their own systems. However, they will need hard data, rather than impressions and guesses based on preconceptions. When it comes to forming (false) beliefs and perceiving inequities, what is right before our eyes is often no match for what we already "know."
APPENDIX 1

Table 6: Complete Logistic Regression Results (Expansion of Table 1)

                             (1)          (2)              (3)
                            Race      (1)+ Legal &    (2)+ Attempts
                            Only       Financial        to Handle
                                     Circumstances         Debt

African-American           3.51 *        3.54 *           3.47 *
  household
Prior bankruptcy                        16.06 *          16.84 *
Represented by                          11.36 *           12.0 *
  attorney
Foreclosure as reason                    2.62 *           2.81 *
  for filing bankruptcy
Homeowner                                 0.86             0.90
Monthly income (In)                      2.10 *           2.11 *
Total assets (In)                         1.12             1.12
Total debt (In)                          0.45 *           0.44 *
Priority debt (In)                       1.07 *            1.07
Secured debt/total debt                  8.40 *           8.16 *
Tried a loan mod, home                                     0.96
  refi or home equity
  loan
Attempted to "work                                         1.56
  with" creditors
Sell house or give it                                      0.79
  back to lender
Bachelor's degree or
  higher
Number of dependents
Live with a spouse or
partner
Female head of
household
Overall chapter 13 rate
in district
Constant                    0.36          0.07             0.05
Model Statistics
N                            428          428              428
Wald chi-squared           29.9 *       160.9 *          163.7 *
McFadden pseudo R-           .05          .29              .30
  squared
Correctly classified        69.2%        79.4%            79.7%

                               (4)           (5)
                               (3)+        (4) Loca
                           Demographic      Legal
                           Information     Culture

African-American              3.24 *        2.39 *
  household
Prior bankruptcy             17.35 *       15.65 *
Represented by               10.23 *         7.41
  attorney
Foreclosure as reason         2.85 *        2.98 *
  for filing bankruptcy
Homeowner                      0.87          0.79
Monthly income (In)           2.52 *        2.66 *
Total assets (In)              1.13          1.15
Total debt (In)               0.40 *        0.48 *
Priority debt (In)             1.07          1.06
Secured debt/total debt       9.30 *        7.19 *
Tried a loan mod, home         0.94          0.80
  refi or home equity
  loan
Attempted to "work             1.71          1.58
  with" creditors
Sell house or give it          0.83          0.88
  back to lender
Bachelor's degree or           1.65          1.50
  higher
Number of dependents          0.77 *        0.77 *
Live with a spouse or          1.06          1.10
partner
Female head of                 1.35          1.55
household
Overall chapter 13 rate                    31.61 *
in district
Constant                       0.04          0.00
Model Statistics
N                              428           428
Wald chi-squared             172.4 *       191.6 *
McFadden pseudo R-             .32           .35
  squared
Correctly classified          80.1%         80.4%

NOTES: The odds ratios and statistical significance for all
variables from the logistic regressions in Table 1 are reported
above. The dependent variable is whether the bankruptcy was filed
as a chapter 13 as opposed to a chapter 7 (1 = "yes"). In all
models, the filers race is a statistically significant
determinant of selecting into chapter 13. Data are from the
Consumer Bankruptcy Project. "Correctly classified" is the number
of observations correctly classified by the model. Across all
observations, the percentage of chapter 7 cases is 64-7%. Thus, a
naive guess that a particular case was not a chapter 13 would be
correct 64.7% of the time.


APPENDIX 2: SURVEY INSTRUMENT

A national random sample of attorneys were asked to estimate the rate at which various groups chose chapter 13 after deciding to file bankruptcy. The survey instrument is reproduced below.

We are interested in the perceptions of bankruptcy professionals about how often people from different groups file chapter 13 (vs. chapter 7). For each group below, please estimate the percentage of bankruptcies in that group that were filed as chapter 13s.

The first question has been completed for you as an example. Because 28% of the nation's bankruptcy cases currently are chapter 13s, we wrote 28% for the "nation as a whole."

Please complete the questions to the best of your knowledge. Even if you are not sure, please make your best guesses. We are interested in your estimates.
                                                     Percentage of
                                                   bankruptcies that
Group                                               were chapter 13

Nation as a whole                                         28%

Men, filing alone

Women, filing alone

Persons 65 years old and over

Homeowners

Persons with a prior bankruptcy
  filing in last 8 years

White

African-American/black

Hispanic/Latino

Asian-American/Asian

Pro se debtors

Southerners (from the states of AL, AR, FL, GA,
LA, MS, NC, SC, TN, TX, VA)


Background information about yourself

Judicial district in which you most often work:--

Year of birth:--

Gender (circle): M F

Race/ethnicity (circle all that apply):

White Black/African-American Latino Asian Other (please specify)--

by Dov Cohen *

Robert M. Lawless **

Faith Shin ***

* Professor of Psychology, University of Illinois

** Max L. Rowe Professor of Law, University of Illinois.

*** Graduate student, Department of Psychology, University of Illinois.

The late Jean Braucher, Roger C. Henderson Professor of Law at the University of Arizona, assisted in the data collection for this paper, but her untimely passing prevented her from being able to participate in the development of this manuscript.

(1) Jean Braucher, Dov Cohen & Robert M. Lawless, Race, Attorney Influence, and Bankruptcy Chapter Choice, 9 J. Empirical Legal Stud. 393, passim (2012). Other work has reported the disparity without controlling for other factors and used data from the Consumer Bankruptcy Project prior to its move to national random samples. See Robert B. Chapman, Missing Persons, Social Science and Accounting for Race, Gender, Class and Marriage in Bankruptcy, 76 Am. Bankr. L.J. 347, 387 n.226 (2002) (reporting a "relatively uniform" effect from the 1991 CBP data of "stark differences" in chapter choice between African Americans and debtors from other races); Rory Van Loo, A Tale of Two Debtors: Bankruptcy Disparities by Race, 72 Alb. L. Rev. 231, 232-34 (2009) (reporting from the 2001 CBP data that African Americans were more likely to enter chapter 13 than debtors of other races).

(2) Braucher, Cohen & Lawless, supra note 1, at 397-405.

(3) Id. at 405-20.

(4) Pamela Foohey, Robert M. Lawless, Katherine Porter & Deborah Thorne, "No Money Down" Bankruptcy, 90 S. Cal. L. Rev. 1055 (2017).

(5) Id.

(6) Dalie Jimenez, The Distribution of Assets in Consumer Chapter 7 Bankruptcy Cases, 83 Am. Bankr. L.J. 795 (2009).

(7) Braucher, Cohen & Lawless, supra note 1, at 405.

(8) Foohey, Lawless, Porter & Thorne, supra note 4, at 36-37 Tables 4-5.

(9) U.S. Const, art. I, [section] 8.

(10) David Stanley & Marjorie Girth, Bankruptcy: Problem, Process, and Reform 74 (1971). At the time of Stanley & Girth's study, the governing law was the Bankruptcy Act of 1898, ch. 541, 30 Stat. 544 (1898). The Bankruptcy Act of 1898 provided for a chapter XIII bankruptcy, a forerunner of today's chapter 13 bankruptcy. See 8 Collier on Bankruptcy [paragraph] 1300.36[3] (Alan N. Resmck & Henry J. Sommer eds, 16th ed. 2017).

(11) Id. at 75.

(12) Gary Neustadter, When Lawyer and Client Meet: Observations of Interviewing and Counseling Behavior in the Consumer Bankruptcy Law Office, 35 Buff. L. Rev. 177 (1986).

(13) Jean Braucher, Lau/yers and Consumer Bankruptcy: One Code, Many Cultures, 67 Am. Bankr. L.J. 501, 581 (1993).

(14) The CBP is a long-running study of consumer bankruptcy filers. The first cohort dates from 1981 with later cohorts drawn in 1991 and 2001, 2007, and 2013-present. Beginning in 2007, advances in information technology allowed national random sampling of bankruptcy filers. Cohorts prior to 2007 were drawn from a subset of judicial districts believed to be nationally representative. Methodologies for the 1981, 1991, 2001, and 2007 CBPs appear in (Catherine Porter, Appendix: Methodology of the 2007 Consumer Bankruptcy Project, in Broke: How Debt Bankrupts the Middle Class 235 (Katherine Porter, ed, 2012). A further, more detailed methodology for the 2007 CBP appears at Robert M. Lawless, Angela K. Littwin, Katherine M. Porter, John A.E. Pottow, Deborah K. Thome & Elizabeth Warren, Did Bankruptcy Reform Fail? An Empirical Study of Consumer Debtors, 82 Am. Bankr. L.J. 349, 391-97 (2008).

(15) Teresa A. Sullivan, Elizabeth Warren & Jay Lawrence Westbrook, Laws, Models, and Real People: Choice of Chapter in Personal Bankruptcy, 13 Law & Soc. Inquiry 661 (1988).

(16) Teresa A. Sullivan, Elizabeth Warren, & Jay Lawrence Westbrook, The Persistence of Local Legal Culture: Twenty Years of Evidence from the Federal Bankruptcy Courts, 17 Harv. J.L. & Pub Pol'y 801 (1994).

(17) Braucher, supra note 13, at 559-60; Chapman, supra note 1, at 387; Teresa A. Sullivan, Elizabeth Warren & Jay Lawrence Westbrook, Who Uses Chapter 13, in Consumer Bankruptcy in Global Perspective 267, 270-71 (J-Niemi-Kiesilainen, et al. eds. 2003); Van Loo, supra note 1, at 232-34.

(18) Braucher, Cohen & Lawless, supra note 1,

(19) Id. at 398.

(20) Id. at 398-403.

(21) Id. at 411.

(22) See S&P Dow Jones Indices LLC, S&P/Case-Shiller U.S. National Home Price Index [CSUSHPINSA], Fed. Reserve Bank of St. Louis, https://.stlouisfed.org//CSUSHPINSA (last visited Jan. 24, 2017).

(23) Delinquency Rate on Single-Family Residential Mortgages, Booked in Domestic Offices, All Commercial Banks [DRSFRMACBS], Board of Governors of the Federal Reserve System (US), https:// fred.stlouisfed.org/series/DRSFRMACBS (last visited Jan. 24, 2017).

(24) A regression tells a researcher the independent effect of each variable while statistically controlling for other variables in the regression. Thus and for example, we can interpret our regressions as the effect being African American has on chapter choice while holding constant the other variables in the regression equation. Robert M. Lawless, Jennifer K. Robbennolt & Thomas S. Ulen, Empirical Methods in Law 273 (2d ed. 2015).

(25) In 2007, the questionnaire asked bankruptcy debtors if they "did ... do or try to do ... to make ends meet" a series of coping mechanisms, including asking creditors to "work with you on the payments." In the Current CBP, the questionnaire asks what the debtors "did ... do ... to make ends meet" (omitting the language about what they tried to do).

(26) Richard Nisbett & Lee Ross, Human Inference: Strategies and Shortcomings of Social Judgment passim (1980).

(27) Daniel Kahneman & Amos Tversky, On the Psychology of Prediction, 80 Psychol. Rev. 237 (1973).

(28) Id. at 237-38; see also Amos Tversky & Daniel Kahneman, Extensional Versus Intuitive Reasoning: The Conjunction Fallacy in Probability Judgment, 90 Psychol. Rev. 293 (1983) (finding that intuitive heuristics (i.e., stereotypes) will lead people to the mathematically impossible conclusion that there is a higher probability of a conjunctive event (e.g., the probability of A and B) than either of the constituent events (e.g., the probability of A and B separately)).

(29) Adam R. Pearson, John F. Dovidio & Samuel L. Gaertner, The Mature of Contemporary Prejudice, 3 Soc. & Personality Psychol. Compass 1 (2009).

(30) Braucher, Cohen & Lawless, supra note 1, at 413-16.

(31) The methodology for the 2007 CBP is discussed supra note 14. We use the 2007 data for gauging

(32) Debtors' attorneys are not the only professionals who play a role in the consumer bankruptcy system. Chapter 13 bankruptcy trustees, who are not government employees but are subject to oversight from the U.S. Trustee Program at the Department of Justice, administer the financial end of a chapter 13 case, collect the debtor's payments on the chapter 13 plan, and disburse payments to creditors. Chapter 13 trustees have substantial discretion and have standing to raise objections to a debtor's chapter 13 plan with the bankruptcy court.

A few months after we sent the surveys to the attorneys, we sent a survey to all 201 chapter 13 trustees listed on the website for the U.S. Trustee Program. The survey and cover letter were substantially similar to those sent to the attorneys, with the addition that the trustees' survey asked for each group the basis for the trustee's estimate. As with the attorneys, a follow-up letter and another copy of the survey were sent approximately four weeks after the initial mailing.

The trustee response rate was low with only 17-9% of the trustee population providing useable data. After we mailed our survey to the trustees, reports reached us that the trustees had communicated among themselves about the survey. Rather than simply not responding, a much higher percentage of the chapter 13 trustees returned blank surveys (9-0%) or sent refusal e-mails (6.0%), as compared to attorneys (2.2%). In contrast to the attorney comments accompanying the blank surveys that just disclaimed knowledge of chapter 13, the trustee comments (and emails) tended to be critical of the research. As one trustee wrote: "For the same reasons which I believe are being expressed to you by other Trustees, I don't feel comfortable filling out your survey. Its 'estimates' approach in an area requiring statistical rigor is a bit too unscientific for me." Counting blank responses and refusal e-mails, we heard back from 32.8% of trustees, similar to the response rate for attorneys.

Obviously, it is impossible to draw any firm conclusions from trustees' resistance, except to say that some percentage of trustees were uncomfortable with the research. Trustees might have a different perspective on chapter 13 use than attorneys do. In her 1993 article, Braucher interviewed 57 professionals involved in the bankruptcy system in two cities in Texas and two cities in Ohio. One long-time chapter 13 trustee in Cincinnati mentioned that he thought African Americans were overrepresented in chapter 13 cases and another experienced chapter 13 trustee in Dayton also said that he thought African Americans and "poor people" might be overrepresented in chapter 13, though Braucher does not indicate that anyone else mentioned this during her other 55 interviews. Braucher, supra note 13, at 559-60. Nevertheless, it is interesting that in the present data, trustees' estimates for white chapter 13 rates (57%) and African American chapter 13 rates (25%) were close to attorneys' estimates. Overall, trustees showed the same reversal in which a 2:1 disparity with African Americans being more likely to file chapter 13 was seen as a 2:1 disparity with whites being more likely to file chapter 13.

(33) Specifically, a debtor can get a second discharge within six years of a previous chapter 13 discharge if the debtor made 100% repayment in the previous case or made 70% repayment through what the court deems the debtor's "best effort." 11 U.S.C. [section] 727(a)(9). Of course, as with many legal matters, there are further complications. The eight and six-year prohibitions are on successive discharges. If the court dis missed a previous bankruptcy case without issuing a discharge, then the prohibition on successive discharges would not apply. Separate statutory sections regulate successive filings (as opposed to successive discharges), 11 U.S.C. [section][section] 109(g), 362(b)(19)-(20), (h), and courts also have claimed a general judicial power to dismiss successive filings for a lack of "good faith." None of these legal technicalities change the primary point in the text: persons who filed bankruptcy in the past eight years have vastly diminished incentives to choose chapter 7, as evidenced by the real-world data showing their propensity to select chapter 13.

(34) A debtor ineligible for a discharge might file a chapter 7 to get the benefit of the automatic stay against creditor actions. The automatic stay can be severely curtailed after repeated filings. For example, the stay does not go into effect if a chapter 7 debtor filed two or more cases within the previous year that were dismissed. 11 U.S.C. [section] 362(c)(4). Other benefits from filing a chapter 7 include the right to redeem personal property from a lien securing a dischargeable consumer debt (typically an auto loan), 11 U.S.C. [section] 722, or the ability to bring an action that avoids judicial liens that impair an exemption or nonpossessory security interests against certain household goods, 11 U.S.C. [section] 522(f).

(35) If one expands the subsample to the thirty-nine attorneys (29% of respondents) who were on average accurate for prior bankrupts (average estimate for prior bankrupts = 82.6%), results look quite similar. These attorneys overestimate white filing rates by 12%, underestimate African American filing rates by 28.3%, and are off by only +0.8% to-11.8% (average =-5.2%) for every other group.

(36) Because our racial categories were not exhaustive (i.e., there was no "other" category), it is mathematically possible but exceedingly unlikely that the true values for all four racial categories would all be above or below the mean.

(37) We alternatively checked for survey responses where the estimates on the racial categories summed between the narrower range of 95-105%. Using the narrower range produces the same result reported in the text, namely that the pattern that emerges from the "numerate" data resembles the pattern for the dataset as a whole. We report the results for the more broadly defined range of possible errors as a more conservative approach toward the data.

(38) Adam J. Berinsky, Gregory A. Huber, & Gabriel S. Lenz, Evaluating Online Labor Markets for Experimental Research: Amazon.com's Mechanical Turk, 20 Pol. Analysis 351 (2012).

(39) See supra notes 34-35 and accompanying text.

(40) We refer to MTurk workers as "relatively" naive. Nine percent of our respondents said they personally had filed for bankruptcy with an additional 53% saying they knew an immediate family member, extended family member, friend, or acquaintance who filed for bankruptcy. However, neither going through bankruptcy oneself (r = .02) nor knowing someone who did (r = .01) were correlated with accuracy in judgments.

(41) See, e.g., Paul Meehl, Clinical Versus Statistical Prediction passim (1954); Philip E. Tetlock, Expert Political Judgment: How Good Is It? How Can We Know? passim (2005).

(42) Ray Boshara, William R. Emmons & Bryan J. Noeth, Ctr. for Household Fin. Stability, The Demographics of Wealth-How Age, Education and Race Separate Thrivers from Strugglers in Today's Economy, Fed. Reserve Bank of St. Louis (2015), https://www.stlouisfed.org/~/media/Files/PDF5/HFS//-l" 2015-Race-Ethnicity-and-Wealth.pdf; Jeff Guo, The Staggering Difference Between Rich Asian Americans and Poor Asian Americans, Wash. Post Wonkblog (Dec. 20, 2016), https://www.washingtonpost. com/news/wonk/wp/2016/12/20/why-asian-americans-arent-as-rich-as-they-seem.

(43) Monica Biernat, Toward a Broader View of Social Stereotyping, 58 Am. Psychol. 1019 (2003); Norbert Schwartz & Herbert Bless, Constructing Reality and Its Alternatives: An Inclusion/Exclusion Model of Assimilation and Contrast Effects in Social Judgment, in The Construction of Social Judgments 217 (L.L. Martin & A. Tesser eds., 1991).

(44) See, e.g., John M. Darley & Paget H. Gross, A Hypothesis-Confirming Bias in Labeling Effects, 44 J. Personality & Soc. Psychol. 20 (1983); Harold H. Kelley, The Warm-Cold Variable in First Impressions of Persons, 18 J. Personality 431 (1950); Raymond Nickerson, Confirmation Bias: A Ubiquitous Phenomenon in Many Guises, 2 Rev. Gen. Psychol. 175 (1998); Mark Snyder, On the Self-Perpetuating Nature of Social Stereotypes, in Cognitive Processes in Stereotyping and Intergroup Behavior 183 (David L. Hamilton ed., 1981); Mark Snyder & William B. Swann, Jr., Behavioral Confirmation in Social Interaction: From Social Perception to Social Reality, 14 J. Experimental Soc. Psychol. 148 (1978).
Table 1: Logistic Regression on Probability of Filing Chapter 13

                                     (1)        (2)        (3)

African-American household          3.51 *     3.54 *     3.47 *
Legal & financial circumstances                 yes        yes
Attempts to renegotiate debt                               yes
Demographic information
Local legal culture
Model Statistics
N                                    428        428        428
Wald chi-squared                    29.9 *    160.9 *    163.7 *
McFadden pseudo R-squared            .05        .29        .30
Correctly classified                69.2%      79.4%      79.7%

                                     (4)        (5)

African-American household          3.25 *     2.39 *
Legal & financial circumstances      yes        yes
Attempts to renegotiate debt         yes        yes
Demographic information              yes        yes
Local legal culture                             yes
Model Statistics
N                                    428        428
Wald chi-squared                   172.4 *    191.6 *
McFadden pseudo R-squared            .32        .35
Correctly classified                80.1%      80.4%

Legal & financial variables: representation by an attorney, prior
bankruptcy, foreclosure as reason for filing bankruptcy, home
ownership, monthly income (In), total assets (In), priority debt
(In), secured debt/total debt

Attempts to handle debt: asked creditors to "work with you" on the
debt; obtained a loan modification, refinanced the home, or took
out a home equity loan; sold the house (including a short sale) or
gave it to the bank.

Demographic information: bachelor's degree or higher, number of
dependents, live with a spouse or partner, female head of household

Local legal culture: overall chapter 13 rate in district

Notes: The table reports odds ratios. An asterisk denotes
statistical significance at the 5% level. The word "yes" indicates
whether a particular vector of control variables was in the
equation, not necessarily whether they were statistically
significant. Due to extreme outliers, the financial variables were
Winsorized at 3.5 standard deviations. The complete results, with
odds ratio, are presented in Appendix 1.

Table 2. Differences Between Attorney Estimates of Chapter 13
Selection and Actual Rates of Chapter 13

                            (A)                  (B)
                     Attorney Estimate       Real World         (C)
                            Mean          Chapter 13 Rates     Error

Whites                     46.3%                28.6%         +17.8%
Homeowners                 54.5%                47.1%          +7.3%
Hispanics                  14.0%                21.7%          -7.7%
Women filing alone         23.1%                31.4%          -8.3%
65 and over                14.7%                25.3%         -10.6%
Men filing alone           21.5%                33.3%         -11.8%
Southerners                34.1%                48.6%         -14.5%
Asian Americans             9.9%                24.4%         -14.5%
Pro se debtors              8.6%                25.6%         -17.0%
African Americans          22.0%                54.6%         -32.6%
Prior bankrupts            35.2%                82.6%         -47.4%

Notes: For different groups, consumer bankruptcy attorneys were
asked to "estimate the percentage of bankruptcies in that group
that were filed as chapter 13 cases." Attorneys were given a
prompt of 28% for the "Nation as a whole," which was the
percentage of bankruptcy cases that were chapter 13 cases at the
time of the survey. The answers are based on a national random
sample of 400 consumer bankruptcy attorneys of which 144 returned
the survey for a response rate of 36%.

Column (A) shows the respondents' mean estimate for the
percentage of bankruptcies from each group that are filed as
chapter 13 cases. Column (B) shows the percentage of cases from
each group that were actually chapter 13 cases from real-world
data captured by the 2007 Consumer Bankruptcy Project, a
nationally representative sample of 2,437 bank-ruptcy filers.
Column C computes the error of the attorney estimates.

Table 3. Error Rates for Attorney Estimates in Judicial Districts with
Above- and Below-mean Chapter 13 Rates

                     Attorneys in Below-    Attorneys in Above-
                        mean Judicial          mean Judicial
                          Districts              Districts

Whites                      +22.8%                 +10.9%
Homeowners                 + 11.5%                 +1.2%
Hispanics                   -6.9%                  -8.8%
Women filing alone          -12.9%                 -1.5%
65 and over                 -13.5%                 -6.4%
Men filing alone            -13.8%                 -8.9%
Southerners                 -23.8%                 -3.5%
Asian Americans            -14.10%                 -15.1%
Pro se debtors              -16.6%                 -17.5%
African Americans           -38.1%                 -25.0%
Prior bankrupts             -49.7%                 -45.5%

                     Overall Error Rate,
                        All Districts
                           Combined

Whites                      +17.8%
Homeowners                  +7.3%
Hispanics                   -7.7%
Women filing alone          -8.3%
65 and over                 -10.6%
Men filing alone            -11.8%
Southerners                 -14.5%
Asian Americans             -14.5%
Pro se debtors              -17.0%
African Americans           -32.6%
Prior bankrupts             -47.4%

Notes: For different groups, consumer bankruptcy attorneys were
asked to "estimate the percentage of bankruptcies in that group
that were filed as chapter 13s." Attorneys were given a prompt of
28% for the "Nation as a whole," which was the percentage of
bankruptcy cases that were chapter 13s at the time of the survey
(n = 144).

This table compares the errors in the attorney estimates for
attorneys located in federal judicial districts above and below
the national mean for chapter 13 filings.

Table 4. Comparison Between All Respondents and Only "Numerate"
Respondents

                        Error in Estimated Probability
                               All Respondents

Whites                              + 17.8
Homeowners                           +7.3

Hispanics                            -7.7
Women filing alone                   -8.3
65 and over                         -10.6
Men filing alone                    -11.8
Southerners                         -14.5
Asian-Americans                     -14.5
Pro se debtors                      -17.0
African Americans                   -32.6
Prior bankrupts                     -47-4

                        Error in Estimated Probability
                               "Numerate" only

Whites                               +25
Homeowners                          + 14.2

Hispanics                            -6.4
Women filing alone                   -5.2
65 and over                          -9.1
Men filing alone                     -9.5
Southerners                          -9.3
Asian-Americans                     -14.3
Pro se debtors                      -16.4
African Americans                   -31.0
Prior bankrupts                     -50.8

Notes: For different groups, consumer bankruptcy attorneys (n =
144) were asked to "estimate the percentage of bankruptcies in
that group that were filed as chapter 13s." To check results
against the possibility that some respondents misunderstood the
question as asking what percentage of chapter 13 bankruptcy
filers have the characteristics listed above, we exclude the
responses for persons who returned a survey where the race
categories were all either above or below 28% (the chapter 13
rate for the nation as a whole that was given to respondents at
the beginning of the survey). For shorthand reference, we refer
to the remaining responses as "numerate." The pattern for the so-
called "numerate" responses resembles the pattern in the overall
data.

Table 5. Differences Between Attorneys' Estimates of Chapter 13
Rates, MTurk Workers' Estimates, and Actual Rates of
Chapter 13 Filing

                                        (A)                 (B)
                                 Attorney Estimate    Actual CBP Data

Whites                                 46.3%               28.6%
Homeowners                            54.5% *-             47.1%
Hispanics                             14.0% *-             21.7%
Women filing alone                     23.1%               31.4%
65 and over                           14-7% *-             25.3%
Men filing alone                      21.5% *-             33.3%
Southerners                            34.1%               48.6%
Asian-Americans                        9.9% *-             24.4%
Pro se debtors                         8.6% *-             25.6%
African Americans                     22.0%                54.6%
Prior bankrupts                       35.2% *+             82.6%
Avg. accuracy (absolute value         24.1% *-
of guess - actual)
Avg. distance from base rate          20.8% *              13.2%
(absolute value of guess -28%)

                                       (C)
                                 MTurk Estimates

Whites                                43.0%
Homeowners                            48.6%
Hispanics                             23.2%
Women filing alone                    25.1%
65 and over                           30.5%
Men filing alone                      32.8%
Southerners                           30.2%
Asian-Americans                       31.7%
Pro se debtors                        25.1%
African Americans                     25.4%
Prior bankrupts                       24.7%
Avg. accuracy (absolute value         22.7%
of guess - actual)
Avg. distance from base rate          17.4%
(absolute value of guess -28%)

Notes: For different groups, consumer bankruptcy attorneys and
users of Amazons Mechanical Turk were asked to "estimate the
percentage of bankruptcies in that group that were filed as
chapter 13s." Both groups were given a prompt of 28% for the
"Nation as a whole," which was the percentage of bankruptcy cases
that were chapter 13s at the time of the attorney survey. The
attorneys' answers are based on a national random sample of 400
consumer bankruptcy attorneys of which 144 returned the survey
for a response rate of 36%.

Column (A) shows the attorneys' mean estimate for the percentage
of bankruptcies from each group that are filed as chapter 13s.
Column (B) shows the percentage of cases from each group that
were actually chapter 13s from real-world data captured by the
2007 Consumer Bankruptcy Project, a nationally representative
sample of 2,437 bankruptcy filers. Column (C) shows the MTurk
workers' estimates. A *+ in column A indicates that the average
attorney estimate was significantly different from the average
MTurk worker estimate, with the attorney mean being closer to the
actual data from the CBP. A *-in column A indicates that the
average attorney estimate was significantly different from the
average MTurk worker estimate, with the attorney mean being
farther from the actual data from the CBP.

Figure 1: Variations in chapter 13 usage by federal
judicial district, 2013 2014

Percentages for the top and bottom 10 judicial districts

1. 79.6%-Louisiana, Western           42.11.7%-New York, Eastern
2. 78.9%-Alabama, Middle              43.10.9%-Wyoming
3. 78.7%-Georgia, Southern            44.10.8%-Oklahoma, Northern
4. 75.0%-Tennessee, Western           45.10.7%-Iowa, Southern
5. 69.6%-Alabama, Southern            46. 9.2%-Idaho
6. 68.2%-North Carolina, Eastern      47. 9.1%-North Dakota
7. 66.8%-Georgia, Middle              48. 9.0%-Oklahoma, Eastern
8. 61.0%-Texas, Southern              49. 8.9%-South Dakota
9. 58.5%-South Carolina               50. 7.5%-New Mexico
10. 58.0%-Mississippi, Northern       51. 5.7%-Iowa, Northern

Notes: For each judicial district, Figure 1 shows the percentage of
bankruptcies in that district that are chapter 13 cases as a
percentage of all bankruptcy petitions in the district. To minimize
the effect of temporary fluctuations in filing rates, data from two
years (2013-2014) are used. The data were provided by Epiq Systems,
Inc
COPYRIGHT 2017 National Conference of Bankruptcy Judges
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2017 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Cohen, Dov; Lawless, Robert M.; Shin, Faith
Publication:American Bankruptcy Law Journal
Date:Sep 22, 2017
Words:15567
Previous Article:Bankruptcy's Quiet Revolution.
Next Article:The Bankruptcy of Refusing to Hire Persons Who Have Filed Bankruptcy.
Topics:

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters