Using a unique dataset from New York City, this article shows that contingent fees seem to have, at least partly, solved the access-to-justice problem in tort litigation. People in poorer zip codes, in fact, make legal claims at a higher rate than those in wealthier areas. This greater propensity to seek and achieve legal redress is partly explained by the fact that poor people are more likely to be injured. In addition, the disparity between rich and poor manifests itself solely in smaller claims. For larger claims, rich and poor make claims at roughly the same rate. African Americans and Hispanics also make claims at a higher rate than whites, and this difference also attenuates when one controls for accident rates or focuses on larger claims.
In recent years, scholars have documented and lamented the fact that most Americans have difficulty gaining access to justice.[1] Most people simply cannot afford a lawyer, even when they have a major legal problem. The issue is especially acute in civil matters, where, unlike criminal proceedings, parties have no constitutional right to free counsel. Even when fundamental issues, such as child custody, are at stake, individuals must generally pay their own lawyers. Legal aid and pro bono lawyers help in some cases, but such services are woefully underfunded and therefore cannot come close to providing legal services to all who need them. Because people so often cannot afford a lawyer, they are frequently unaware of their rights and unable to enforce even the rights they are aware of. As a result, they may be unjustly evicted from their homes, lose custody of their children, or fail to receive compensation to which they are legally entitled.
While the overall access-to-justice problem is undoubtedly large and real, this article highlights a potential bright spot: tort litigation. Using a unique database from New York City, we find that poor people are able to hire contingent-fee lawyers and secure settlements and judgments in personal injury and property damage cases at a rate comparable to that of the middle class and wealthy. In fact, people from poor zip codes make claims, sue, and recover at rates higher than those in richer zip codes. Contingent fees seem to have, at least partly, solved the access-to-justice problem in tort.
We further explore why poor people seem to make claims more often. We analyze two factors: injury rates and claim size. Poor people are more likely to be injured. They are more likely to live in neighborhoods that are less safe, to use older products with fewer safety features, and to work at dangerous jobs. Regression analysis[2] confirms that differences in injury rates account for at least some of the difference in claiming rates between rich and poor. Poor people are also more likely to make small claims. While wealthier people are less likely to think it is worth their time to hire a lawyer to address issues where damages are likely to be only a few thousand dollars, poorer people, for whom even a few hundred dollars may make a large difference, sue more often. The data are consistent with these ideas about who is more likely to sue when the stakes are low. The difference between rich and poor is large for small claims, decreases with recovery amounts, and is small or non-existent for the largest claims.
We also find that, controlling for income, African Americans and Hispanics are more likely to make claims than whites, although again the effect is concentrated in smaller claims and is, in part, explained by higher accident rates. Previous research found that jury verdicts were higher in counties with higher African American populations.[3] One might have predicted that phenomena would also lead to higher claim rates by minorities, and our data provide support for that hypothesis.[4] There is also some evidence that Asian Americans claim at lower rates than whites.
The evidence we find of higher claim rates among the poor and among disadvantaged minorities is also consistent with casual observation of lawyer advertising. Personal injury lawyers often advertise on buses, and the advertisements are often in Spanish, suggesting that they target poorer people and Hispanics.
To our knowledge, this is the first article to rigorously examine tort litigation by income and race of the plaintiff. While contingent fees have been extensively studied,[5] we are not aware of any analysis, except of medical malpractice, that uses regression or similar methods to examine the impact of contingent fees on access to justice. Helen Burstin, David Studdert, and their co-authors have found that the poor make fewer medical malpractice claims.[6] As discussed below, our data suggest the contrary, that even in medical malpractice, the poor make more claims.
Part I describes the data used in this article. Part II analyzes the data, first with maps and then with regressions. Part III discusses some limitations and caveats.
Since 1957, courts in New York City have required contingent fee lawyers to file “closing statements” in cases involving personal injury, property damage, or wrongful death. While these closing statements contain a wealth of data, for this article, the most important information they contain is the plaintiff’s address (including zip code) and the plaintiff’s recovery, whether by settlement or judgment.[7]
These data contain settlement amounts, even though they are ordinarily confidential. Because of the sensitive nature of these data, the New York court system released the data to the authors only with strict data handling and confidentiality requirements. The authors received closing data from 2003–2013 from the First Appellate Division, which has supervisory authority over lawyers in Manhattan and the Bronx.
Closing statements must be filed whenever a lawyer retained on contingent fee resolves a case, whether the case settled or went to judgment, and even if the case settled without ever having been filed in court. That is, even settlements that were reached wholly without judicial intervention, or even without contact with the courts, still trigger the obligation to file a closing statement. In addition, unlike most insurance closed claim data, these statements are filed even if the plaintiff recovered nothing—whether because the claim was abandoned or because the defendant prevailed in court—so they also provide information on claiming rates, not just payouts. Because the data include cases abandoned without a suit or settled before a suit was filed, our unit of analysis is a claim rather than a lawsuit.
In addition to data from closing statements, we use some other standard data sources. The population in the zip code is from the 2010 census.[8] The data for accidental deaths come from NYC Health, which contains zip code level death statistics, averaged over three years.[9] The data on auto accidents are from the New York Police Department’s Motor Vehicle Collisions database. The data on median household income and the demographic data on race by zip code come from the American Community Survey, an annual survey conducted by the U.S Census.
Table 1 provides summary statistics for our key variables.
Table 1: Summary Statistics
|
Five Boroughs |
Bronx & Manhattan | ||||
|
Mean |
S.D. |
Mean |
S.D. | ||
|
Number of Cases per 1,000 residents |
3.55 |
2.08 |
3.90 |
2.49 | |
|
Number of Cases with Recovery >$0 per 1,000 |
2.86 |
1.66 |
3.11 |
1.97 | |
|
Number of Cases with Recovery >$20,000 |
1.27 |
0.67 |
1.44 |
0.74 | |
|
Number of Cases with Recovery >$50,000 |
0.67 |
0.38 |
0.74 |
0.38 | |
|
Number of Cases with Recovery >$100,000 |
0.38 |
0.24 |
0.41 |
0.27 | |
|
Number of Medical Malpractice Cases per 1,000 |
1.16 |
1.57 |
0.98 |
1.42 | |
|
Number of Motor Vehicle Injuries per 1,000 (annual) |
1.37 |
8.84 |
0.92 |
2.43 | |
|
Number of Accidental Deaths |
0.32 |
0.16 |
0.29 |
0.17 | |
|
Median Household Income ACS ($10,000) |
6.27 |
2.72 |
6.80 |
3.76 | |
|
Percentage African American |
22.45 |
25.52 |
20.03 |
20.25 | |
|
Percentage Asian |
12.85 |
12.66 |
9.41 |
9.25 | |
|
Percentage Hispanic |
25.41 |
19.45 |
31.25 |
24.00 | |
|
Observations |
1,872 |
742 | |||
|
Notes: Table reports means and standard deviations on the first district sample from 2003 to 2013 overall and separately for the Bronx and Manhattan. The zip codes with the top 1% of cases per 1,000 residents are trimmed from the sample. All rates “per 1,000” are per 1,000 residents. | |||||
The map on the left of Figure 1 (Panel A) shows the average number of claims per thousand per year between 2003 and 2013. Zip codes with more claims are darker.[10] The claims are divided by quartile, with the bottom quartile having fewer than 0.838 claims per 1,000, while the top quartile has more than three times as many claims, with a cutoff of 2.51 claims per 1,000 residents. The map on the right (Panel B) shows median household income by zip code.[11] Zip codes with lower median household income are darker.
The two maps are almost identical. Places with many claims per capita (dark areas on the left map), such as Harlem and the Bronx, tend to be places with low per capita income (dark areas on the right map). Conversely, places with relatively few suits (light areas on the left graph), such as the Upper East and Upper West Sides of Manhattan, tend to be places with high incomes (light areas on the right map). These maps strongly suggest that there is not an access to justice problem for torts as poorer people bring, on average, more tort claims.
In the upper left part of Figure 2 (Panel A) we map the number of claims per thousand in 2010. The results are similar to Figure 1, suggesting that the pattern of claiming is fairly constant across years, a fact confirmed by other maps not included in this article. In the upper right part of Figure 2 (Panel B) we also map the percentage of African Americans by zip code. The map is very similar to Panel B of Figure 1 which is not surprising given the correlation between race and income, especially in New York. Panel C (at the lower left) shows the location of Hispanics. It is similar to the map for African Americans, although there are differences, especially in Brooklyn. There is a strong positive correlation between claims and the proportion of African Americans and Hispanics in the zip code. Panel D (lower right) shows the distribution of Asian Americans. There appears to be a negative correlation between Asian populations and claims. For example, there are more Asians in Manhattan, and those are areas with relatively few claims per capita.
We explore two hypotheses that might explain the strong negative correlation between income and claims:
(1) The poor sue more often because they are injured more often.
(2) The poor are more likely to bring claims for smaller injuries because the benefit to them is higher, and the cost is lower. The benefit of small claims is higher for poor people than for the rich because the marginal utility of money falls with income. The cost is lower because claims and lawsuits take time, and the poor have lower wages or rely on government assistance, suggesting that the monetary value of their time is, on average, lower.
Table 2 uses regression analysis to explore how the number of claims per thousand varies with median household income and recovery amount. We estimate a simple linear regression of the total number of claims per 1,000 individuals on median household income in the zip code for that year:
where claims per 1,000it is the number of claims per 1,000 residents of zip code i in year t, median incomeit is the median household income for zip code i and year t, are year fixed effects for each year 2004–2013;[12] (the fixed effect for 2003 is omitted to avoid collinearity with the intercept ), and are the robust standard errors.[13]
Table 2: Impact of Median Income on Claims, 2003–2013
|
Panel A: All Boroughs 2003–2013 | |||||
|
(1) |
(2) |
(3) |
(4) |
(5) | |
|
VARIABLES |
Total Cases |
Total Cases with recovery |
Total Cases |
Total Cases >$50,000 |
Total Cases |
|
Median Income ACS ($10,000) |
-0.335*** |
-0.252*** |
-0.073*** |
-0.018*** |
-0.008*** |
|
(0.019) |
(0.015) |
(0.006) |
(0.004) |
(0.002) | |
|
Observations |
1,872 |
1,872 |
1,872 |
1,872 |
1,872 |
|
R-squared |
0.413 |
0.418 |
0.279 |
0.183 |
0.139 |
|
Panel B: Bronx and Manhattan 2003–2013 | |||||
|
Median Income ACS ($10,000) |
-0.435*** |
-0.329*** |
-0.104*** |
-0.032*** |
-0.015*** |
|
(0.023) |
(0.019) |
(0.007) |
(0.004) |
(0.003) | |
|
Observations |
742 |
742 |
742 |
742 |
742 |
|
R-squared |
0.647 |
0.632 |
0.520 |
0.313 |
0.199 |
|
Notes: Table presents coefficients for the regression of median income of a zip code on the number of annual number of cases for each category per 1,000 residents of the zip code. The data include zip codes in the five Boroughs of New York City. All regressions include year fixed effects. Robust standard errors are reported in parentheses. A *, ** or *** indicates statistical significance at the 10%, 5% and 1% levels, respectively. The zip codes with the top 1% of cases per 1,000 residents are trimmed from the sample. | |||||
The first column of Table 2, Panel A confirms the patterns observed in the maps. There is a strong negative association between claims and income. For every additional ten thousand dollars in income, the number of claims per thousand households goes down by 0.335. Given that the average number of claims per thousand is 3.55,[14] these are large differences, amounting to almost ten percent. If one compared the bottom quartile zip codes (with per capita income under $39,000) to top quartile districts (with per capita income of more than $75,000), there would be about forty percent fewer claims per capita in the richer districts.[15]
The columns other than the first show that the effect of income varies with the size of the claim. As recoveries go up, the effect of income gets smaller. For recoveries over $20,000, the effect is small and for recoveries over $100,000, the effect is essentially zero, although still statistically significant. These coefficients are consistent with the hypothesis that poorer people are more likely to turn to the legal system for small claims, but that both rich and poor seek legal redress at roughly the same rate for larger claims.[16]
The coefficients presented in Panel B show that results are similar whether one examines only Manhattan and the Bronx, or all five boroughs of New York City (Panel A). Our figures are most complete for Manhattan and the Bronx because our data are from closing statements from (and thus primarily suits filed in courts in) those two boroughs. Nevertheless, because the data contain large numbers of claims by persons residing in the other three boroughs (presumably because they hired lawyers based in Manhattan or the Bronx or sued there), we report figures for all boroughs as well.
Table 3 analyzes the effect of race on claims.
Table 3: Impact of Race on Claims, 2003–2013
|
Panel A: All Boroughs 2003–2013 | |||||||
|
(1) |
(2) |
(3) |
(4) |
(5) | |||
|
VARIABLES |
Total Cases |
Total Cases |
Total Cases |
Total Cases |
Total Cases | ||
|
Percentage African American |
0.026*** |
0.018*** |
0.002*** |
-0.001*** |
-0.001*** | ||
|
(0.001) |
(0.001) |
(0.001) |
(0.000) |
(0.000) | |||
|
Percentage Asian |
-0.016*** |
-0.014*** |
-0.008*** |
-0.006*** |
-0.004*** | ||
|
(0.003) |
(0.003) |
(0.001) |
(0.001) |
(0.001) | |||
|
Percentage Hispanic |
0.036*** |
0.026*** |
0.008*** |
0.002*** |
0.001* | ||
|
(0.002) |
(0.002) |
(0.001) |
(0.000) |
(0.000) | |||
|
Observations |
1,872 |
1,872 |
1,872 |
1,872 |
1,872 | ||
|
R-squared |
0.499 |
0.483 |
0.304 |
0.212 |
0.170 | ||
|
Panel B: Bronx and Manhattan 2003–2013 | |||||||
|
Percentage African American |
0.040*** |
0.030*** |
0.008*** |
0.001** |
0.000 | ||
|
(0.003) |
(0.002) |
(0.001) |
(0.001) |
(0.000) | |||
|
Percentage Asian |
-0.019*** |
-0.016*** |
-0.006** |
-0.003* |
-0.001 | ||
|
(0.006) |
(0.005) |
(0.003) |
(0.002) |
(0.002) | |||
|
Percentage Hispanic |
0.049*** |
0.037*** |
0.012*** |
0.004*** |
0.002*** | ||
|
(0.002) |
(0.002) |
(0.001) |
(0.001) |
(0.000) | |||
|
Observations |
742 |
742 |
742 |
742 |
742 | ||
|
R-squared |
0.739 |
0.720 |
0.551 |
0.321 |
0.192 | ||
|
Notes: Table presents coefficients for the regression of the racial demographics of a zip code on the number of annual number of cases for each category per 1,000 residents of the zip code. The data include zip codes in the five Boroughs of New York City. All regressions include year fixed effects. Robust standard errors are reported in parentheses. A *, ** or *** indicates statistical significance at the 10%, 5% and 1% levels, respectively. The zip codes with the top 1% of cases per 1,000 residents are trimmed from the sample. | |||||||
The effect of race is broadly consistent with the effect of income, which is not surprising given the correlation between race and income by zip code. African Americans and Hispanics make legal claims at a higher rate than whites, but the effect attenuates as claim size increases. For Bronx and Manhattan, there is absolutely no difference between richer and poorer districts for claims with recovery greater than $100,000. For Asian Americans, the sign of the coefficient on race is always negative, which is consistent with Figures 3 and 6. This suggests that districts with large numbers of people of Asian ancestry had fewer claims.
Table 4 analyzes the effect of accident rates in order to test the hypothesis that claim rates were higher in poorer districts because such neighborhoods are more dangerous and thus more likely to generate claims.
Table 4: Impact of Income and Accident Rates on Claims,
2003–2013
|
Panel A: All Boroughs: Including Auto Accidents 2012 & 2013 | |||||
|
(1) |
(2) |
(3) |
(4) |
(5) | |
|
VARIABLES |
Total Cases |
Total Cases with recovery >0 |
Total Cases with recovery >$20,000 |
Total Cases with recovery >$50,000 |
Total Cases with recovery >$100,000 |
|
Median Income ACS ($10,000) |
-0.156*** |
-0.114*** |
-0.038*** |
-0.004 |
-0.003 |
|
(0.030) |
(0.024) |
(0.012) |
(0.006) |
(0.004) | |
|
Number of Motor Vehicle Injuries per 1,000 (annual) |
-0.009* |
-0.007* |
-0.003* |
-0.001 |
-0.000 |
|
(0.005) |
(0.004) |
(0.002) |
(0.001) |
(0.001) | |
|
Observations |
374 |
374 |
374 |
374 |
374 |
|
R-squared |
0.078 |
0.067 |
0.036 |
0.003 |
0.003 |
|
Panel B: All Boroughs: Including Accidental Death, 2003–2013 | |||||
|
Median Income ACS ($10,000) |
-0.337*** |
-0.250*** |
-0.077*** |
-0.020*** |
-0.009*** |
|
(0.017) |
(0.014) |
(0.006) |
(0.003) |
(0.002) | |
|
Number of Accidental Deaths per 1,000 (3 year average) |
0.127 |
0.302 |
-0.129 |
-0.071 |
-0.023 |
|
(0.278) |
(0.225) |
(0.095) |
(0.055) |
(0.036) | |
|
Observations |
1,872 |
1,872 |
1,872 |
1,872 |
1,872 |
|
R-squared |
0.197 |
0.176 |
0.094 |
0.019 |
0.009 |
|
Panel C: Bronx and Manhattan: Including Auto Accidents 2012 & 2013 | |||||
|
Median Income ACS ($10,000) |
-0.177*** |
-0.128*** |
-0.046*** |
-0.006 |
-0.006* |
|
(0.031) |
(0.025) |
(0.012) |
(0.006) |
(0.003) | |
|
Number of Motor Vehicle Injuries per 1,000 (annual) |
-0.102*** |
-0.084*** |
-0.038*** |
-0.016** |
-0.005 |
|
(0.035) |
(0.029) |
(0.014) |
(0.007) |
(0.004) | |
|
Observations |
148 |
148 |
148 |
148 |
148 |
|
R-squared |
0.278 |
0.243 |
0.165 |
0.052 |
0.042 |
|
Panel D: Bronx and Manhattan: Including Accidental Death, 2003–2013 | |||||
|
Median Income ACS ($10,000) |
-0.449*** |
-0.338*** |
-0.107*** |
-0.033*** |
-0.017*** |
|
(0.020) |
(0.016) |
(0.007) |
(0.004) |
(0.003) | |
|
Number of Accidental Deaths per 1,000 (3 year average) |
-0.300 |
-0.102 |
-0.021 |
-0.005 |
-0.041 |
|
(0.437) |
(0.356) |
(0.146) |
(0.084) |
(0.062) | |
|
Observations |
742 |
742 |
742 |
742 |
742 |
|
R-squared |
0.448 |
0.412 |
0.298 |
0.109 |
0.050 |
|
Notes: All regressions include year fixed effects. Robust standard errors are reported in parentheses. A *, ** or *** indicates statistical significance at the 10%, 5% and 1% levels, respectively. The zip codes with the top 1% of cases per 1,000 residents are trimmed from the sample. Notes to Table 2 also apply. | |||||
Panels A and C control for the number of motor vehicle injuries per thousand. Motor vehicle injuries are not a particularly good proxy for overall injuries in New York City because so few people drive. Nevertheless, even those who do not drive are at risk of being hit as a pedestrian. In addition, the variable used here includes bus and truck accidents, which are at least as relevant to New York as elsewhere. Oddly, there is usually a negative relationship between injuries and claims, although it is not usually statistically significant. Including motor vehicle injuries has a dramatic effect on the coefficients on median income. Most drop by fifty percent or more as compared with the coefficients in Table 2, which is consistent with the hypothesis that the higher claim rate among poor people can be explained by differences in injury rates.
Panels B and D control for injuries using data on accidental deaths. As with motor vehicle injuries, the results are a bit strange, as claims fall as accidental deaths increase. The effect of income on claims remains negative and about the same magnitude as in Table 2, which is contrary to our hypothesis. We suspect accidental deaths are so infrequent that the data are unreliable.
Table 5: Impact of Income, Race, and Accident Rates on Claims per 1,000 Residents, 2003–2013, All Boroughs
|
Panel A: Median Income and Demographic Variables Only | |||||
|
(1) |
(2) |
(3) |
(4) |
(5) | |
|
VARIABLES |
Total Cases |
Total Cases with recovery >$0 |
Total Cases with recovery >$20,000 |
Total Cases with recovery >$50,000 |
Total Cases with recovery >$100,000 |
|
Median Income ACS ($10,000) |
-0.166*** |
-0.140*** |
-0.053*** |
-0.025*** |
-0.016*** |
|
(0.021) |
(0.018) |
(0.009) |
(0.006) |
(0.004) | |
|
Percentage African American |
0.018*** |
0.012*** |
-0.000 |
-0.003*** |
-0.002*** |
|
(0.002) |
(0.002) |
(0.001) |
(0.001) |
(0.000) | |
|
Percentage Asian |
-0.025*** |
-0.022*** |
-0.010*** |
-0.007*** |
-0.005*** |
|
(0.003) |
(0.003) |
(0.001) |
(0.001) |
(0.001) | |
|
Percentage Hispanic |
0.020*** |
0.013*** |
0.003*** |
-0.001 |
-0.001** |
|
(0.003) |
(0.003) |
(0.001) |
(0.001) |
(0.000) | |
|
Observations |
1,872 |
1,872 |
1,872 |
1,872 |
1,872 |
|
R-squared |
0.302 |
0.261 |
0.137 |
0.064 |
0.055 |
|
Panel B: Including Auto Accidents 2012 & 2013 | |||||
|
Median Income ACS ($10,000) |
-0.078** |
-0.060* |
-0.029* |
-0.003 |
-0.004 |
|
(0.039) |
(0.033) |
(0.017) |
(0.010) |
(0.005) | |
|
Percentage African American |
0.016*** |
0.010*** |
0.002 |
0.001 |
-0.000 |
|
(0.004) |
(0.003) |
(0.002) |
(0.001) |
(0.001) | |
|
Percentage Asian |
-0.005 |
-0.004 |
-0.004 |
-0.001 |
-0.001 |
|
(0.007) |
(0.006) |
(0.003) |
(0.002) |
(0.001) | |
|
Percentage Hispanic |
0.005 |
0.004 |
0.000 |
-0.000 |
-0.000 |
|
(0.005) |
(0.004) |
(0.002) |
(0.001) |
(0.001) | |
|
Number of Motor Vehicle Injuries per 1,000 (annual) |
-0.009** |
-0.007** |
-0.003*** |
-0.001 |
-0.001 |
|
(0.003) |
(0.003) |
(0.001) |
(0.001) |
(0.001) | |
|
Observations |
374 |
374 |
374 |
374 |
374 |
|
R-squared |
0.128 |
0.103 |
0.050 |
0.008 |
0.009 |
|
Panel C: Including Accidental Death, 2003–2013 | |||||
|
Median Income ACS ($10,000) |
-0.149*** |
-0.122*** |
-0.055*** |
-0.027*** |
-0.017*** |
|
(0.024) |
(0.020) |
(0.010) |
(0.007) |
(0.004) | |
|
Percentage African American |
0.019*** |
0.012*** |
-0.001 |
-0.003*** |
-0.002*** |
|
(0.002) |
(0.002) |
(0.001) |
(0.001) |
(0.000) | |
|
Percentage Asian |
-0.025*** |
-0.022*** |
-0.010*** |
-0.007*** |
-0.005*** |
|
(0.004) |
(0.003) |
(0.001) |
(0.001) |
(0.001) | |
|
Percentage Hispanic |
0.021*** |
0.014*** |
0.003*** |
-0.001 |
-0.001** |
|
(0.003) |
(0.003) |
(0.001) |
(0.001) |
(0.000) | |
|
Number of Accidental Deaths per 1,000 (3 year average) |
0.592* |
0.626** |
-0.062 |
-0.076 |
-0.034 |
|
(0.307) |
(0.266) |
(0.128) |
(0.088) |
(0.053) | |
|
Observations |
1,872 |
1,872 |
1,872 |
1,872 |
1,872 |
|
R-squared |
0.304 |
0.264 |
0.137 |
0.065 |
0.056 |
|
Notes: Table presents coefficients for the regression of median income of a zip code on the number of annual number of cases for each category per 1,000 residents of the zip code. The data include zip codes in the five Boroughs of New York City. All regressions include year fixed effects. Robust standard errors are reported in parentheses. A *, ** or *** indicates statistical significance at the 10%, 5% and 1% levels, respectively. The zip codes with the top 1% of cases per 1,000 residents are trimmed from the sample. | |||||
Table 6: Impact of Income, Race, and Accident Rates on Claims per 1,000 Residents, 2003–2013, Manhattan and Bronx
|
Panel A: Median Income and demographic variables only | ||||||
|
(1) |
(2) |
(3) |
(4) |
(5) | ||
|
VARIABLES |
Total Cases 2003–2013 |
Total Cases with recovery >$0 |
Total Cases with recovery >$20,000 |
Total Cases with recovery >$50,000 |
Total Cases with recovery >$100,000 | |
|
Median Income ACS ($10,000) |
-0.127*** |
-0.096*** |
-0.055*** |
-0.026*** |
-0.021*** | |
|
(0.031) |
(0.026) |
(0.014) |
(0.009) |
(0.007) | ||
|
Percentage African American |
0.031*** |
0.023*** |
0.004** |
-0.001 |
-0.002* | |
|
(0.005) |
(0.004) |
(0.002) |
(0.001) |
(0.001) | ||
|
Percentage Asian |
-0.026*** |
-0.022*** |
-0.009*** |
-0.005** |
-0.002 | |
|
(0.008) |
(0.006) |
(0.003) |
(0.002) |
(0.002) | ||
|
Percentage Hispanic |
0.034*** |
0.026*** |
0.005*** |
0.001 |
-0.001 | |
|
(0.005) |
(0.004) |
(0.002) |
(0.001) |
(0.001) | ||
|
Observations |
742 |
742 |
742 |
742 |
742 | |
|
R-squared |
0.527 |
0.486 |
0.329 |
0.122 |
0.056 | |
|
Panel B: Including Auto Accidents 2012 & 2013 | ||||||
|
Median Income ACS ($10,000) |
-0.002 |
0.008 |
0.002 |
0.017 |
0.003 | |
|
(0.049) |
(0.044) |
(0.024) |
(0.015) |
(0.007) | ||
|
Percentage African American |
0.024** |
0.017** |
0.006 |
0.003* |
0.001 | |
|
(0.009) |
(0.008) |
(0.004) |
(0.002) |
(0.001) | ||
|
Percentage Asian |
0.001 |
-0.004 |
-0.003 |
0.001 |
0.000 | |
|
(0.014) |
(0.013) |
(0.007) |
(0.003) |
(0.002) | ||
|
Percentage Hispanic |
0.022** |
0.017** |
0.006 |
0.003 |
0.001 | |
|
(0.009) |
(0.008) |
(0.004) |
(0.002) |
(0.001) | ||
|
Number of Motor Vehicle Injuries per 1,000 (annual) |
-0.098*** |
-0.079*** |
-0.036*** |
-0.016*** |
-0.005 | |
|
(0.024) |
(0.019) |
(0.010) |
(0.005) |
(0.003) | ||
|
Observations |
148 |
148 |
148 |
148 |
148 | |
|
R-squared |
0.338 |
0.301 |
0.199 |
0.086 |
0.060 | |
|
Panel C: Including Accidental Death, 2003–2013 | ||||||
|
Median Income ACS ($10,000) |
-0.116*** |
-0.084*** |
-0.052*** |
-0.024** |
-0.022*** | |
|
(0.036) |
(0.030) |
(0.016) |
(0.010) |
(0.008) | ||
|
Percentage African American |
0.031*** |
0.023*** |
0.004** |
-0.001 |
-0.001* | |
|
(0.005) |
(0.004) |
(0.002) |
(0.001) |
(0.001) | ||
|
Percentage Asian |
-0.027*** |
-0.022*** |
-0.009*** |
-0.005** |
-0.002 | |
|
(0.008) |
(0.006) |
(0.003) |
(0.002) |
(0.002) | ||
|
Percentage Hispanic |
0.035*** |
0.026*** |
0.005*** |
0.001 |
-0.001 | |
|
(0.005) |
(0.004) |
(0.002) |
(0.001) |
(0.001) | ||
|
Number of Accidental Deaths per 1,000 (3 year average) |
0.359 |
0.409 |
0.117 |
0.041 |
-0.035 | |
|
(0.430) |
(0.357) |
(0.191) |
(0.119) |
(0.093) | ||
|
Observations |
742 |
742 |
742 |
742 |
742 | |
|
R-squared |
0.527 |
0.487 |
0.330 |
0.122 |
0.057 | |
|
Notes: Table presents coefficients for the regression of median income of a zip code on the number of annual number of cases for each category per 1,000 residents of the zip code. The data include zip codes in the Bronx and Manhattan. All regressions include year fixed effects. Robust standard errors are reported in parentheses. A *, ** or *** indicates statistical significance at the 10%, 5% and 1% levels, respectively. The zip codes with the top 1% of cases per 1,000 residents are trimmed from the sample. | ||||||
Tables 5 and 6 show the effect of income and race together. Given the strong correlation by zip code of race and income, we are not confident that these regressions can disentangle the effects of income and race. Table 5 analyzes all of New York City, and Table 6 analyzes just the Bronx and Manhattan. All six panels of Tables 5 and 6 show that, when race is included, the negative coefficients on income become closer to zero, except for the largest recoveries, as compared with the coefficients in Tables 2 and 3. Panels B and C of Tables 5 and 6 add controls for the two measures of accident rates used in Table 4—motor vehicle accident rates and accidental deaths. The effect of income on claims remains consistently negative. The coefficient on accidental death is generally positive, as one would expect, but it is not statistically significant. The analysis of income and race together suggests that both are significant factors in explaining the number of claims. It is not the case that one is just a proxy for or driven by the other.
The data upon which this Article is based do not categorize the cases by type. That is, there is no indication in the closing statements whether the case involved a motor vehicle accident, premises liability (e.g., “slip and fall”), medical malpractice, or some other tort. The data do, however, indicate who the insurer is. Some insurers are “multiline,” meaning they provide several types of insurance, e.g. both for auto insurance and medical malpractice insurance. Other insurers are “monoline.” They insure only one type of insurance, e.g. medical malpractice insurance. The latter type of insurance is helpful for this research because it can be used to identify which cases involve medical malpractice claims. Table 7 shows an analysis of just the medical malpractice claims. It suggests that poorer people bring more medical malpractice claims. For every $10,000 increase in median household income, the number of medical malpractice claims goes down by 0.069 per thousand residents. Given that the average number of medical malpractice claims per thousand is around one,[17] this means that, for every $10,000 increase in income, the number of medical malpractice claims goes down by about 7%, and thus that the difference between the lowest and highest quartile would be about thirty percent. This is a smaller difference than for claims overall,[18] but still relatively large.
Table 7: Medical Malpractice Cases 2003–2013
|
(1) |
(2) | |
|
VARIABLES |
All Boroughs |
Bronx and Manhattan |
|
Median Income ACS ($10,000) |
-0.069*** |
-0.069*** |
|
(0.013) |
(0.013) | |
|
Observations |
1,872 |
742 |
|
R-squared |
0.083 |
0.094 |
|
Notes: Table reports means and standard deviations on the first district sample from 2003 to 2013 overall and separately for the Bronx and Manhattan. The zip codes with the top 1% of cases per 1,000 residents are trimmed from the sample. | ||
These results for medical malpractice differ from those in prior studies, which generally found that the number of medical malpractice claims go up with income.[19] There are a number of possible explanations for the different results. Perhaps medical providers who purchase monoline insurance, which are the only ones we can identify in our dataset, tend to insure medical providers who disproportionately treat poorer people. Or perhaps New York City, where our data come from, is different from the locations studied in other articles. Or perhaps there are problems with either our study or the other studies, or our data or the data in other studies. This would be a fruitful topic for further research.
The analysis described so far has been essentially cross-sectional. That is, it attempts to explain differences in claim rates between zip codes. The regressions included year fixed effects, which control for changes over time. Another possible approach is panel-data analysis. That is, one could use both year fixed effects and zip-code fixed effects. One would therefore be trying to explain changes over time that vary from general trends. For example, if a neighborhood gentrified and average income went up faster than in other parts of the city, one might expect claims to go down in that area. This approach is not very promising because there is, in fact, very little change over time as can be seen by comparing Figure 1, Panel A (which is the average claim rate over the entire period 2002–2013) and Figure 2, Panel A (which is the claim rate in 2010, near the end of the period). The two are nearly identical, indicating little change over time.[20] Nevertheless, for completeness, we attempted a panel-data analysis using the equation above on page nine, but with zip-code fixed effects in addition to year fixed effects. As we expected, because of the limited temporal variation, the results are mixed.[21] In the panel-data analysis, the coefficient on median income is sometimes positive rather than negative, indicating that more claims are filed as a neighborhood gets richer. Those results are not usually statistically significant, but they sometimes are. In addition, the signs on the coefficient are unstable; sometimes the results are negative and statistically significant, but the reverse is also true depending on the specification. As noted at the outset, the panel-data results are not very reliable because of the limited amount of variation over time. Nevertheless, for completeness, we mention them.
Our analysis has a number of limitations. First, our data come only from New York City, which is atypical in a number of ways. Perhaps outside New York City, or outside large cities more generally, the contingent fee system is not as effective in enabling rich and poor to make tort claims at similar levels. Second, our analysis is only at the zip code level. We do not have individual income data. It is possible that the poor make claims with much lower frequency, but that higher income individuals living in poorer zip codes make claims at much higher rates. That is, our analysis may be subject to the ecological fallacy.[22] Third, our accident data are not very good. Motor vehicle accidents are not a good proxy for accidents overall, especially in New York City, and analysis of accidental deaths is unreliable because accidental deaths are relatively rare. Fourth, although poorer people and minorities seem to claim as much as, if not more than, wealthier whites, perhaps even rich whites are not accessing tort justice at rates commensurate with their needs. This seems unlikely, but it is possible. Fifth, our data cover only claims brought on contingent fee. If wealthy people are more likely to hire tort lawyers on an hourly (or fixed) fee, that would undermine our results. Nevertheless, anecdotal evidence suggests that even the richest plaintiffs hire contingent-fee lawyers for their tort claims. Sixth, wealthier people are more likely covered by health, disability, and other forms of insurance and may make fewer claims for this reason.[23] On the other hand, poor people often have Medicaid or Medicare, which may dampen their incentive to sue for similar reasons. Finally, although we have data on settlement and judgment amounts, we do not have ex ante measures of claim size, value, or quality. It, therefore, is possible that poorer people, African Americans, and Hispanics make higher claims at greater rates than others, but that they recover less, making it appear that claiming rates are more equal for larger claims than is actually the case.
Although contingent fees are certainly not without their problems, evidence from New York City suggests that they have largely eliminated inequalities in access to justice between rich and poor and between whites and racial minorities. In fact, the poor make claims and recover at higher rates than the rich, and African Americans and Hispanics make claims at higher rates than whites. The higher claiming and recovery rates of the poor are partly explained by their higher injury rates, thus providing some support for the first hypothesis. The second hypothesis is more strongly confirmed by the data. The higher claiming rates among the poor and disadvantaged minorities is concentrated among smaller claims, and there is little disparity among larger claims.
Eric Helland[24]* and Daniel Klerman[25]**
See, e.g., Deborah L. Rhode, Access to Justice (2004); Beyond Elite Law: Access to Civil Justice in America (Samuel Estreicher & Joy Radice eds., 2016). ↑
Regression analysis is a statistical technique that summarizes the relationship between an outcome of interest (the dependent variable) and one or more possible explanatory factors (the dependent variables). In our analysis, the key dependent variable is the number of claims per capita, and independent variables include the average income, racial composition, and accident rates in the plaintiff’s zip code. ↑
Eric Helland & Alexander Tabarrok, Race, Poverty, and American Tort Awards: Evidence from Three Data Sets, 32 J. Legal Stud. 27, 41 (2003). ↑
See also Theodore Eisenberg & Martin T. Wells, Trial Outcomes and Demographics: Is There a Bronx Effect?, 80 Tex. L. Rev. 1839 (2002) (finding mixed evidence of demographic variables impact on jury awards in federal cases); Mary R. Rose & Neil Vidmar, Commentary, The Bronx “Bronx Jury”: A Profile of Civil Jury Awards in New York Counties, 80 Tex. L. Rev. 1889 (2002) (providing a critique of existing evidence of a “Bronx Effect,” i.e. higher awards in the more African American Bronx); Issa Kohler-Hausmann, Community Characteristics and Tort Law: The Importance of County Demographic Composition and Inequality to Tort Trial Outcomes, 8 J. Empirical Legal Stud. 413 (2011) (finding that demographic variables at the county level do not impact litigation outcomes but do impact the size of awards). ↑
See, e.g., Herbert M. Kritzer, Risks, Reputations, and Rewards: Contingency Fee Legal Practice in the United States (2004); Eric Helland & Alexander Tabarrok, Contingency Fees, Settlement Delay, and Low-Quality Litigation: Empirical Evidence from Two Datasets, 19 J.L. Econ. & Org. 517 (2003); David L. Noll, The Effect of Contingent Fees and Statutory Fee-Shifting, in Beyond Elite Law: Access to Civil Justice in America 170 (Samuel Estreicher & Joy Radice eds., 2016). ↑
Helen R. Burstin, William G. Johnson, Stuart R. Lipzitz & Troyen A. Brennan, Do the Poor Sue More? A Case-Control Study of Malpractice Claims and Socioeconomic Status, 270 JAMA 1697, 1699 (1993); David M. Studdert et al., Negligent Care and Malpractice Claiming Behavior in Utah and Colorado, 38 Med. Care 250, 250 (2000). ↑
For a more detailed description of the data, see Eric Helland, Daniel Klerman, Brendan Dowling & Alexander Kappner, Contingent Fee Litigation in New York City, 70 Vand. L. Rev. 1971, 1972–81 (2017). ↑
We have excluded two zip codes where the 2010 population is extremely low: 10020 (population 1) and 10464 (population 109). Because our dependent variable is claims per population, and because some claims are reported from these areas, these zip codes become extreme outliers. Judging from Google Maps satellite view, which shows many homes, the reported population for 10464 is almost certainly wrong. 10020 includes Radio City and may, in fact, contain no residents, although the fact that a fair number of closing statements report 10020 as the plaintiff’s zip code is rather strange. ↑
NYC Vital Statistics, NYC Health Scis. Libr., https://datacatalog.med.nyu.edu/dataset/10098 [https://perma.cc/DJ6X-5EZK]. ↑
These figures were first published in Helland et al., supra note 7, at 1991–92, which contained a short (two paragraph) discussion of “the demographics of tort claims.” This article expands that discussion through the use of other data sets and regression analysis. ↑
Explore Census Data, U.S. Census, https://data.census.gov/ [https://perma.cc/8L6E-TD5Y]. The figures use data from the American Community Survey three-year averages, accessed with a program called Social Explorer. Social Explorer, https://www.socialexplorer.com/ [https://perma.cc/4H3T-88MM]. The figures were created using data from the 2010 ACS using the program spmap. ↑
The “fixed effects” are binary (0/1) variables for each year between 2004 and 2013. They measure the effect of factors that vary by time and would otherwise be hard to measure. ↑
The results are robust to clustering the standard errors on zip code. However, several of our regressions involving Manhattan and the Bronx have as few as 70 zip codes, which may give rise to a small cluster problem. For this reason, we report the generally larger robust standard errors. ↑
See the first row of Table 1. ↑
The average income in the bottom quartile is $33,700 and the average income in the top quartile is $99,980. So, the difference is $66,280. Given the coefficient of 0.335, that means the predicted difference between the average household in the top and bottom quartiles is 2.2 (0.335 x 66,280)/10,000 claims per thousand, which is almost two-thirds of the mean (3.55). If the distribution is symmetric, that would mean that the claim rate is about 4.65 (3.55+ (2.2 divided by 2)) in the top quartile and 2.45 (3.55 – (2.2 divided by 2)) in the bottom quarter. 2.45 is 59% of 4.65. So, claims in the bottom quartile would be 41% lower than in the top quartile. ↑
One way of thinking about the magnitudes of these coefficients is to divide them by the rates in Table 1 (Summary Statistics). For all cases with positive recoveries, the coefficient of -0.252 in Table 2 would be divided by 2.86, the number of cases with positive recoveries in Table 1, yielding -8.9%, meaning that for every increase of ten thousand dollars in income in a zip code, the number of claims with positive recoveries goes down almost 9%. For cases with recovery over $100,000, the corresponding figure would be -2% = -0.008 / 0.38, meaning that for cases with recoveries over $100,000, an increase in ten thousand dollars in income in a zip code means the number of recoveries goes down by only 2%. ↑
As indicated in table 2, the rate is 1.16 for all boroughs and 0.98 for the Bronx and Manhattan. ↑
See supra note 15. ↑
Burstin, supra note 6, at 1699; Studdert, supra note 6, at 257. ↑
The absence of significant change of time is confirmed by more thorough analysis, available on request. ↑
Results available upon request. ↑
The ecological fallacy refers to mistakes in the interpretation of data that result from incorrectly making inferences about individuals from group data. ↑
The effect of insurance is complex and requires more research. To the extent that they have already received compensation from their insurer for their injuries, injured persons may be less motivated to make a claim against the injurer. Nevertheless, under the common law collateral source rule, the injured party could still make a claim and could receive recovery from the injurer, even for amounts for which the insurer had already paid compensation. This would mean that the injured party’s financial incentive to sue would remain the same, even if s/he were insured. On the other hand, insurance contracts usually have subrogation clauses, which mean that the insurer would get the relevant portion of any plaintiff recovery. This reduces the plaintiff’s incentive to sue but encourages the insurer to assert a claim on the plaintiff’s behalf. Nevertheless, if the insurer asserts the claim, it probably would not show up in the dataset used for this article because the insurer would probably not hire a contingent fee lawyer. The analysis is further complicated by the fact that New York abolished the collateral source rule in 1986. N.Y. C.P.L.R. 4545 (McKinney 2003). This unambiguously reduced the plaintiff’s incentive to make a claim, although case law suggests that insurers continued to have the right to subrogation even for amounts that the injured party was barred from recovering due to the abolition of the collateral source rule. See, e.g., Kelly v. Seager, 558 N.Y.S.2d 403 (N.Y. App. Div. 1990). There is also an open question as to whether the statute in force between 1986 and 2009 actually changed the common law collateral source rule when the insurance contract had a subrogation clause, because the statute made an exception for “such collateral sources entitled by law to liens against any recovery of the plaintiff,” and insurers with subrogation rights could be constructed as holding “liens against any recovery of the plaintiff.” N.Y. C.P.L.R. 4545(c) (McKinney 2009). That statutory provision was changed in 2009, so the statute now only exempts “payments as to which there is a statutory right of reimbursement.” N.Y. C.P.L.R. 4545(a) (McKinney 2023). Because insurance subrogation rights are ordinarily contractual, they would no longer be exempted by the statute. Under prior case law, the insurer might still be able to sue independently for subrogation, but no cases address whether insurers retain that right after 2009. Medicaid and Medicare liens, however, would be statutory liens, so a plaintiff covered by these programs could sue for her full damages, including medical expenses paid by these federal programs, although the federal government could file a lien so that the injured party would not actually receive reimbursement for those expenses. The analysis is further complicated by the fact that New York statutes relating to collateral source payments are preempted by ERISA when insurance is provided by a self-funded employer plan. See Iron Workers Locs. 40, 361 & 417 Health Fund v. Dinnigan, 911 F. Supp. 2d 243, 250–51 (S.D.N.Y. 2012). ↑
* William F. Podlich Professor of Economics, Claremont McKenna College and Senior Economist, RAND. ↑
** Edward G. Lewis Professor of Law and History, University of Southern California Law School. The authors thank John Donohue, Lee Epstein, Jeffrey Fagan, Michael Frakes, James Greiner, Jordan Neyland, Chris Robertson, Seth Seabury, Tara Sklar, George Vojta, and participants at Academia Sinica Center for Institutional and Behavior Studies workshop, Sichuan University law lecture, Brooklyn Law School Faculty Workshop, Notre Dame Law School Faculty Workshop, USC Center for Law & Social Science (CLASS) workshop, 2018 Civil Procedure Workshop, American Law and Economics Association 2018 Annual meeting, and the 2018 University of Arizona Law School QuantLaw conference for helpful comments and suggestions. ↑