Are Pain and Suffering Awards (Un-)Predictable? Evidence from Germany

To analyse the differential impact of personal-, injury- and procedural-specific determinants of DPS awards we extract information from the German verdict collection Hacks, Ring and Böhm (2006, 2007). This dataset exclusively covers DPS proceedings from courts of the first and second instance. It only focuses on DPS and leaves out any other related compensations, notably changes in earning capacities. Originally, our sample includes around 2.839 proceedings on DPS between 1980 and 2006 (about 15 percent of all cases are disputed after the year of the change in German tort law in 2002; see Table 1). It contains individual information on the victims’ gender and age, on the number of injuries and on the amount of compensations. In addition, it includes information on the court type as well as the instance and location where the decision took place, on the rate of contributory negligence (measured in percentages) and on details of the injury (i.e., duration, intensity and severity of the injury). Further, the dataset comprises sentences from the District, Regional and Province Courts, thereby fully covering the first two instances of civil proceedings in Germany (sentences from the level of the Supreme Court are not covered in the dataset). We also have information whether a plaintiff was represented by a lawyer in the first instance District Court.

Table 1 provides a descriptive overview over the data. As can be seen from the table, our sample contains 2.244 observations where full information on victim-, injury- and procedural-specific characteristics of DPS is available. The average DPS, measured in 2005 EUR, amounts to about EUR 22.303, with a minimum of EUR 69 and a maximum of EUR 657.701. The median compensation is around EUR 5.460 (not reported in the table). Generally, the whole distribution of DPS awards is skewed to the left (see Panel A of Figure 1). Panel B of Figure 1 further shows that more than 60 percent of all DPS below the median is lying within a range of EUR 0 to 2.500. In the econometric analysis below, we account for this pattern taking the logarithm of compensations when defining the dependent variable (the resulting variable is close to a log-normal distribution).

Figure 1

Around 42 percent of the victims in the dataset are females. Regarding the age of the victims, our sample includes exact information on this variable only for 883 individuals. For the remaining observations, however, we are able to assign each individual to one of four age cohorts: “child” (victims below 14 years), “young” (between 15 and 18 years), “adult” (between 19 and 65 years) and “retired” (older than 65). As can be seen from Table 1, adults are the most represented group in our sample (around 65 percent of all observations). For those 883 individuals where we have exact information on age, we have a mean age of around 30 years, with minimum and maximum values of 0,5 and 85 years.

Figure 2 shows the distribution of (log) compensations over gender and age cohorts. Overall, we observe higher payments for males than for females (the unconditional gender difference is around EUR 4.422 which is significant at the five percent level). Further, we can see a weak hump-shaped pattern in payments over age classes, with children and retired people receiving on average the lowest compensations while young ones and adults obtaining the highest DPS. Children and adults exhibit the largest variations in compensations.

Figure 2

With respect to the injury characteristics we distinguish between the severity of the injury and the number of injuries. Accordingly, 58 percent experienced more than one injury.

We are able to describe the severity of personal injuries and the impairment of the quality of life by a nine-point scale adopted from the US National Association of Insurance Commissioners (NAIC). This scale ranges from emotional injuries only to death (see Table A.1 in the Appendix for further details). Table 1 shows that the majority of the individual injuries can be classified as minor temporary damages (around 40 percent), followed by temporary insignificant damages with a share of about 30 percent; about 10 percent belong to minor permanent or major temporary injuries. The remaining five categories are less of importance; for permanent major and permanent grave injuries we have only one and two observations in our sample (see notes in Table 1). For this reason, we summarize these injuries with permanent significant ones, obtaining a sample size in this new class of injuries of 98 observations (henceforth, we refer to this class as “permanent serious”).

The distribution of DPS over the severity of pain is illustrated in Figure 3. The lowest compensations can be observed for temporary insignificant damages (in absolute numbers EUR 4.860, on average), while the highest ones are paid for permanent serious ones (EUR 96.800). Within the classes of temporary and permanent damages we find that compensations increase with the severity of pain. This pattern is broadly in line with the German tort law, suggesting higher payments for more intensive pain.

Figure 3

Most of the verdicts were disputed at South and West German courts (about 32 and 41 percent; see Table 1). Further, about two thirds of the proceedings can be assigned to the first instance (District or Regional Courts), while one third reflects the second instance (Regional and Province Courts). Around 83 percent of all 544 cases disputed at the first level District Court draw upon a lawyer. The average compensation at this level is around EUR 1.422, and the difference for cases with/without legal representation is about EUR 284 (not reported in Table 1), which is significant at the 5 percent level.

Figure 4 plots the distribution of DPS awards over court types and years before/after the change in the German tort law in 2002 (see footnote 7). Three aspects are worth noting. First, payments are insignificantly higher in years after 2002 than before. This suggests that the change in tort law does not have any systematic impact on DPS. Second, and in line with the legal background described above, we can see a strong difference in compensations within the first instance (i.e., the District and the Regional Courts), which, in absolute numbers and unconditional on other factors, amounts to about EUR 27.873, on average. Third, this gap becomes much smaller when comparing payments between the first and the second instance. The corresponding absolute differences are, on average, EUR 9.146 between the first instance District Court and the second instance Regional Court, and EUR 1.541 for the one between the first instance Regional Court and the second instance Province Court. In other words, if a victim decides to appeal a sentence from the first instance, we might expect considerable additional payments. However, it also appears that variation in payments is lower in the first (second) instance District (Regional) Courts than in the other courts. In addition, we also observe a number of outlying observations which might be influential, and, therefore, should be addressed appropriately in the subsequent empirical analysis.

Figure 4

Generally, Figures 2 to 4 suggest substantial differences in DPS over individual-, pain- and procedural-specific characteristics. However, the underlying descriptive statistics do not allow drawing any conclusions on the relative importance of these determinants and whether they are systematically associated with compensations. These questions are addressed in the next section, providing an econometric analysis on determinants of DPS awards.

To study the role of individual-, injury- and procedural-specific characteristics of and their relative importance on DPS awards we estimate a standard regression model, which reads as: (1)yi=α+XiVβV+XiIβI+XiPβP+t+εi.y_i = \alpha + X_i^V \beta _V + X_i^I \beta _I + X_i^P \beta _P + t + \varepsilon _i .

The variable y is the log of compensation awarded to victim i, X denotes a vector of explanatory variables, where superscripts V, I and P indicate the group of variables belonging to victim-, injury- and procedural-specific variables, respectively. The variable t refers to a time trend informing about the average change in DPS over the whole time period between 1980 and 2006.

To rule out that DPS awards are not systematically influenced by the change in German tort law (see footnote 7) we additionally included a dummy variable for sentences after the year 2002. The corresponding parameter estimate turned out to be insignificant, suggesting that there is no structural break around this time period. For this reason, we suppressed the year 2002 dummy in all of the regressions.

Our variables of interest might be assigned to three broad groups: Vector X^V includes the victim-specific characteristics gender and age. Since the victims’ exact age is not entirely available in our sample, we estimate three versions of (1): In Model A we leave out any age information, in Model B we include the above mentioned age categories, and in Model C we use a victim’s exact age if available. Then, our sample includes 2,126 observations in Model A, 1,207 in Model B and 855 in Model C.

The variable X^I covers injury-specific characteristics, i.e., (i) six indicator variables for the intensity and severity of pain as measured by the above-mentioned nine-point NAIC-scale (the reference category is temporary insignificant injuries), (ii) one on whether the plaintiff is exposed to more than one injury, and (iii) one on whether she is at least partially guilty. Further, to investigate whether there are age and gender specific differences in the impact of injury-specific determinants on DPS awards, we estimate a more saturated Model D where we interact each of the variables in (i), (ii) and (iii) separately with gender and age (i.e., we include eight interaction terms for gender and eight ones for age; in sum, we have 16 additional variables as compared to Model C).

Finally, X^P accounts for procedural-specific variables. In particular, we incorporate a set of indicator variables indicating at which region, instance and court type the case was disputed. To capture that victims are free to choose a lawyer as legal representation at the first level District Court (only), we add a dummy variable indicating whether a lawyer was hired at this stage.

Equation (1) is estimated with ordinary least squares. To avoid that our estimation results are driven by outlying observations (as suggested by Figures 2 to 4), we exclude the upper and lower end percentile of the remainder error term, losing 44 (Model A), 26 (Model B) and 22 (Models C and D) observations, respectively.

It turns out that both versions, the original one and the outlier-corrected one, come to very similar qualitative conclusions with regard to all variables in the model. Therefore, we only present the results of the outlier-corrected models here. The ones of the original model are available from the authors upon request.

Table 2 reports our estimation results. First of all, we can see that the four models do not vary considerably with regard to the sign and the magnitude of the estimated parameters as well as their significance. Therefore, and to draw further conclusions on the age-related variables, we concentrate on Models B to D when discussing our estimation results.

Panel A of Table 2 informs about the impact of individual-specific characteristics on DPS. First, the victim’s gender turns out insignificant throughout. Second, age is insignificant in Model B (“child” is the reference category), but exerts a significantly negative impact on compensations when using the victim’s exact age (Model C). The marginal effect is rather low, however, indicating that a victim may obtain about 0,4 percent [≈ 100(exp(−0,004) − 1)] lower compensations than people one year younger, all else equal. The (main) effect of age changes to insignificant when incorporating interaction terms between age and the injury-specific characteristics (Model D). The interaction effects themselves are insignificant throughout, which is also the case for the interactions between gender and the injury-specific characteristics.

For the sake of brevity, we do not report the parameter estimates of these 16 interaction terms in Table 2, but the corresponding estimation results are available from the authors upon request.

Panel B of the table reports the estimated parameters for injury-specific effects. In line with the German tort law and most of previous research, we generally observe significantly increased payments with the intensity and severity of pain (see, e.g., Amaral-Garcia 2011, for similar evidence from Spain; Bovbjerg, Sloan and Blumstein 1989, Viscusi 1988 and Geistfeld 1995, on common law countries). Starting with contributory negligence, we find a significant negative coefficient, which is in line with the legal framework described above. The parameter estimate is around −0,01, translating into a marginal reduction in compensations of about one percent per one percentage point increase in contributory negligence. Second, the parameter estimate of more injuries is significantly positive. A marginal effect of around 180 percent (Model D) suggests that people with more injuries are assigned with more than twice of compensations than persons with only single violations. We further find that individuals suffering from emotional injuries obtain about 50 percent more compensation than victims experiencing temporary insignificant injuries (forming the reference category). Relying on Model D, we can see that victims with minor (major) temporary injuries may expect about 54 (484) percent higher compensations than people in the reference category (i.e., ones with temporary insignificant injuries). Payments for minor (serious) permanent injuries are about 132 (879) percent higher than payments for insignificant temporary injuries. In case of death, we estimate a significant positive parameter estimate with a marginal effect of 254 percent.

As the German law of tort and damages does not know wrongful death cases, DPS are used to compensate for the pain and suffering after accident and before death. Hence, death does not lead to the highest pain and suffering damages in Germany.

Regarding the procedural-specific variables reported in Panel C of Table 2, we firstly can see that it does not really matter at which location the cases are disputed. The corresponding regional dummies are insignificant throughout.

As a robustness check, we also include interactions with the courts’ geographical location and a period dummy controlling for DPS awards before the German reunification in 1990 (results are available from the authors upon request). While the direct location dummies remain still insignificant, the period dummy indicates significantly higher DPS awards (in Model A only) after 1990. Furthermore, the interaction terms reveal significantly lower payments in Eastern and Northern Germany (Models C and D) before the reunification compared to the payments in Western Germany.

Panel C of the table further informs about the importance of legal representation at courts. As outlined before, victims are not obliged to be legally represented on the level at the District Court (first instance), whereas all other courts (instances) ask mandatorily for a legal representation. Hence, it seems interesting to see if the existence of a lawyer within a District Court proceeding shows a systematic impact on the outcome conditional on a certain configuration of personal-, injury- and (other) procedural-specific characteristics. The corresponding interaction term enters significantly positive with a marginal effect between 50 to 60 percent. Given average compensations of about EUR 1.422 at this court level, this difference translates into additional payments between EUR 710 and 850, on average (notice that this difference is substantially higher than the unconditional difference reported in Section Descriptive Statistics). Interestingly, the average lawyer fees for the average compensation of EUR 1.420 on this court level are lying around EUR 315 to 525

For detailed information on calculating lawyer fees in DPS cases, see http://www.ratgeber-schmerzensgeld.de.

Overall, we notice that about two thirds of the total variation in DPS awards is explained by the explanatory variables included in our empirical model. Most of the explanatory variables in Table 2 take the expected sign following the legal framework of the German tort law. From this, we firmly may conclude that DPS are reasonable and, from this perspective, widely transparent. Moreover, Figure 5 shows that predicted DPS awards are uncorrelated with the residuals from the regressions in equation (1) (Model D), suggesting that unobserved characteristics of victims, injuries and the jurisprudence do not systematically affect our predicted values of DPS awards.

Figure 5

In Table 3, we further decompose the variation in compensations into two major components: one explained by the independent variables (i.e., the “model”) and one by the remaining determinants (the “residual”).

Since the interaction terms between age/gender and the injury-specific variables enter insignificantly in our regressions (Model D in Table 2), the analysis of variance is related to Model C from above.

The first two columns of Table 3 reflect the variance in absolute and in relative terms. The latter is the variance due to each effect in percent of the total variance. Similarly, the last column of the table reports the relative importance of each variable group, calculated as the sum over the variables’ shares (column 2) within a group of characteristics. From this, we can see which determinants actually account for the lion’s share in the variation of compensations. Obviously, these are not variables belonging to personal- or procedural-specific characteristics (with a share of 0,2 and 11,4 percent), but the ones of determinants on the severity and intensity of pain, contributing to about one third in total variation of DPS awards. Again, this finding seems consistent with the legal basis of the German tort law, which, together with our results from the regression analysis, indicates that German DPS awards are mostly predictable rather than purely random.