Does Empathy Pay? Evidence on Empathy and Salaries of Recent College Graduates

Abstract

This paper examines linkages between empathy and salaries of recent college graduates. While it has been suggested that greater empathy enhances performance because empathic individuals work well with others and are good leaders, it is also possible that they will be less productive due to lower motivation to compete for pecuniary rewards or because they are considered to be too accommodating. Also, more empathic individuals may choose occupations that are more socially oriented and less well paid. We find a large, significant negative relationship between empathy and earnings for both men and women. While we cannot pinpoint the cause for this strong correlation or reach conclusions about causation, we find that empathic individuals choose college majors and sectors of employment that pay less. However, gender, major, and sector together do not account for the full negative relationship between empathy and salaries. While it seems likely that preferences lead more empathic people to choose lower paying occupations, we cannot eliminate the possibility that greater empathy has a negative effect on productivity.

Introduction

Empathy has been identified as an important personality characteristic that leads to successful job performance and positive labor market outcomes. Empathic people may work well with others, building and maintaining relationships in diverse organizations, and empathy may be a key component of effective managerial leadership. This position has recently been adapted as part of the conventional wisdom and is found in the business press and postings on the web. Some examples include: Wilson in Fortune (Wilson 2015) cites empathy as one of five essential talents senior executives say they need (the other four are intellectual curiosity, 360-degree thinking, cultural competence, and adaptability); Boyers in Forbes (Boyers 2013) says, “behind every successful business, you are likely to find a leader who has mastered the skill of empathy.” and Greenfield in Bloomberg (Greenfield 2016) states, “The future of American labor lies in jobs that require empathy and critical thinking.” These statements imply that greater empathy may “pay” and lead to higher earnings. However, it could be that more empathy has a negative impact on salaries if more empathic people are less driven to succeed or are so agreeable that others undervalue them. Moreover, those with greater empathy may choose careers that are more socially oriented, such as those in caring fields or in the nonprofit sector that are less financially remunerative. Therefore, it is not self-evident that empathy “pays,” at least with a pecuniary return.

To date, there is not much empirical evidence on the effects of empathy on labor market earnings. In this paper, we examine the linkages between empathy and compensation for recent college graduates one to six years after graduation from four institutions. The virtue of using a sample of college graduates is that the effect of empathy and other personality traits on schooling outcomes will be less of a confounding factor because all subjects have college degrees from well-regarded institutions. However, it should be pointed out that the findings may not generalize to those with less education. We use a well-validated measure of empathy, the Davis Interpersonal Reactivity Index (IRI), and send annual surveys about labor market outcomes to the graduates. We find that subjects in our sample who score higher on the empathy scale have lower earnings one to six years after graduation. We cannot pinpoint the cause of this strong negative correlation or make claims about causation, but some of this effect is due to the fact that women are more empathic than men as well as sorting into college majors and sectors of employment that are less-well paid. However, gender, major, and sector together do not account for the full negative effect of empathy on salaries. While it seems likely that preferences lead more empathic people to choose lower paying occupations that are more oriented toward social welfare or caring for others, we cannot rule out the possibility that greater empathy has a negative effect on productivity.

Literature on Noncognitive Abilities: Relationship between Empathy and Earnings

Cognitive abilities and schooling are important determinants of labor market outcomes as has been widely established in the economics literature. However, psychologists have long stressed the importance of personality in addition to cognitive abilities in affecting educational and occupational success.¹ Recently, economists have begun to examine the effects of “noncognitive abilities” on labor market outcomes. An early paper, Duncan and Dunifon (1998), summarizes the literature on “soft skills” but the authors say, “we are not aware of studies testing explicitly whether workers entering the labor market with problem-solving “soft” skills are more successful in their subsequent labor-market careers.”² However, by 2012, in their introduction to the reprint of the article (Duncan and Dunifon 2012), they summarize the “burgeoning literature based on many longitudinal studies.” The authors point out that there is a lack of consensus regarding the general role of noncognitive abilities on labor market outcomes and on which skills are the most important.

The absence of a theoretical model linking personality or noncognitive abilities to work performance or salaries was partly responsible for the lack of attention of economists to the role of such traits in labor market outcomes. Traditional economic models center on the role of human capital (skills or capacities that contribute to productivity) in determining worker earnings. Noncognitive abilities could increase workers’ productivity but the linkages are often not well specified. A major difficulty in the empirical work in this area is that the precise definition of noncognitive abilities differs and at times is not carefully linked theoretically to employment outcomes. And, it seems clear that different jobs require and reward different noncognitive skills. As a result, research has measured noncognitive skills in various ways, often determined by the availability of measures of such characteristics in the data.

Economists have modeled the role played by noncognitive abilities on salaries and other career outcomes in several ways. In early work on noncognitive skills and earnings, Bowles et al. (2001a) sketch two behavioral models incorporating noncognitive traits. One is a “Schumpeterian model”, where workers with certain capabilities are better able to search out and capture disequilibrium rents in factor payments. The other is a principal-agent model (presented in more detail in Bowles et al. 2001b). “Incentive-enhancing preferences” of employees lead them to increase effort in response to wages so employers are willing to pay employees with these characteristics higher wages. Some examples of incentive-enhancing preferences provided by the authors include: greater orientation toward the future (raises the evaluation of costs of losing future wages), sense of personal efficacy (own work effort has larger influence on the probability of retaining job as opposed to the influence of external factors, often termed “locus of control”), and higher marginal utility of income or shame at losing a job (increases effort by raising the value of holding a job).

An alternative approach is based on comparative advantage where workers choose tasks or jobs based partly on cognitive and noncognitive abilities (Almlund et al. 2011). Productivity in different tasks depends on worker traits (cognitive, personality, physical, etc.), effort, and actions resulting from the former two. Traits are treated as endowments at a given point in time.³ Workers maximize expected utility which is a function of goods consumed, productivity, effort, preferences, and their information set. This leads to a set of response functions for consumption of goods, effort, and actions with respect to each task/job. The determinants of the response functions are: the reward to doing the task (wage and non-pecuniary benefits), price of goods, a vector of traits, situational experiences, unearned income, and the information set. Preferences affect these choices via trade-offs and goals.

In another trade model, Deming (2017) considers “social skills,” where these skills reduce coordination costs so that workers can better sort into doing tasks for which they have comparative advantage. He provides empirical evidence that social skills positively impact wages, and the effect has increased over time with the more rapid growth in demand for employees with both high math skills and social skills.

There is substantial evidence in the empirical literature that noncognitive abilities and personality affect educational achievement, job performance, and earnings. We provide a brief summary here of the literature in economics focusing on effects on earnings; more detailed summaries of the broader literature can be found in Almlund et al. (2011), Duncan and Dunifon (2012), and Humphries and Kosse (2017).

Economists often borrow the “Big Five” personality factors from the psychology literature to identify personality traits. The Big Five are: openness to experience, conscientiousness, extraversion, agreeableness, and neuroticism or emotional stability (acronym is OCEAN). It should be pointed out that different personality traits are likely to be valued differently in different occupations so that the expected sign of the relationship between traits and earnings in an aggregate sense is not always evident.

Openness to experience refers to the degree to which people are curious, creative, artistic, and receptive to new ideas. Mueller and Plug (2006), Heineck (2011), and O’Connell and Sheikh (2011) find a positive link between openness of experience and salaries, but most other studies do not find a significant relationship. Conscientiousness is the tendency to be dependable, organized, efficient, and hardworking. One would expect this trait to lead to success in employment and hence higher wages, and the evidence generally supports this to be the most important of the Big Five in affecting educational and labor market outcomes. A positive relationship is found in O’Connell and Sheikh (2011), Fletcher (2013), Prevoo and ter Weel (2015), Gensowski (2018), and Maczulskij and Viinikainen (2018), while conscientiousness is found to affect earnings only for women by Mueller and Plug (2006) and Heineck (2011). Extraversion is an orientation to the outside world, characterized as being sociable, energetic, adventurous, and assertive. The empirical work finds some linkages between extraversion and salaries; there is a positive effect in O’Connell and Sheikh (2011) and Maczulskij and Viinikainen (2018), who find a positive impact of “activity” which they identify as a facet of extraversion. Mueller and Plug (2006) find a negative effect for women that disappears with additional controls, and the effect is positive for men only in Fletcher (2013) and Gensowski (2018). Agreeableness is the tendency to be cooperative, helpful, and compassionate. This trait is often found to be negatively associated with earnings (Heineck 2011; Gensowski 2018); some find the negative effect only for men (O’Connell and Sheikh 2011) and others for women (Nyhus and Pons 2005, 2012; Heineck and Anger 2010). The interpretation of the negative effect of agreeableness might be that being agreeable could hurt career prospects if being too cooperative hurts own performance or leads to the perception of superiors that the individual is submissive rather than a leader. Of the Big Five, agreeableness may be the trait most closely related to empathy and the negative relationship with wages is consistent with our results below. The last trait, neuroticism refers to experiencing distressful emotions such as anxiety, fear, and nervousness. Mueller and Plug (2006), Fletcher (2013), and Maczulskij and Viinikainen (2018) find a negative relationship between neuroticism and earnings while Heineck (2011) finds a negative effect for women only. Studies sometimes focus on the other side of neuroticism, emotional stability, which consists of self-regulation, calmness, and consistency. Nyhus and Pons (2005) and O’Connell and Sheikh (2011) find emotional stability to be positively related to earnings for both men and women.

Another personality characteristic that has been often studied is locus of control, the degree to which individuals believe that outcomes depend on their own abilities and actions rather than on outside factors. The logic is that the greater the worker’s belief that he or she has control over own life outcomes, the more likely the worker will put in effort to be successful, an activity which is expected to have positive effects on wages. Heineck and Anger (2010) find that having an external locus of control is associated with lower wages for men and women; while Groves (2005) and Semykina and Linz (2007) find lower earnings for women. Heckman et al. (2006) use a composite measure of locus of control and self-esteem and find this positively related to earnings for men and women at all educational levels except for male college graduates. Similarly, Girtz (2014, 2015) finds a relationship between self-esteem and wages, with about a third of the effect occurring via higher levels of education.

We contribute to this literature by examining the link between empathy and labor market earnings, a relationship not previously investigated. In a principal-agent context (Bowles et al. 2001a, 2001b), empathy could be an “incentive-enhancing” trait in that workers with greater empathy might exhibit greater feelings of reciprocity (more prosocial attitude) in response to higher wages; if higher wages prompt greater effort, employers will respond with increased pay. However, more empathic people might also be less motivated by pecuniary incentives so their effort response to higher wages would be less, implying lower pay.

Applying the comparative advantage model of Almlund et al. (2011), empathy can be considered to be a trait that affects productivity and it can have different effects on productivity in different jobs, either positive or negative. Greater empathy might increase a worker’s productivity in occupations that require people to work well with others in teams or on joint projects, and it could increase managerial effectiveness in supervising employees. On the other hand, it is possible that being more empathic could reduce productivity in some jobs, for example, empathic workers may be less motivated to work hard in the absence of perceived social benefits of the output or may be less productive in a more independent, competitive environment, for example in commission work. We would expect workers to sort into jobs for which greater empathy increases their productivity implying a positive effect on earnings, controlling for type of job, but we cannot rule out the possibility that more empathic people have lower productivity and lower wages. At the same time, preferences associated with being more empathic might lead workers to choose occupations that provide sources of satisfaction (utility) other than financial earnings. Consistent with extensive evidence in the psychology literature showing that more empathic people act more prosocially,⁴ greater empathy may be linked to preferences to work in jobs that serve the disadvanted or provide care for others. In the U.S. economy, these jobs are usually relatively low paying so that more empathic workers employed in these jobs would earn less than similarly qualified peers in less “caring” jobs.

In summary, there is not an unambiguous prediction about the sign of the relationship between empathy and earnings. Because preferences would likely reduce wages as people choose into lower paid professions, if a positive association between earnings and empathy is found this suggests that it is due to greater productivity and comparative advantage sorting. However, if a negative association is found, it could either be due to lower productivity (that is not offset by sorting into occupations with comparative advantage) or to preferences for socially oriented work that is lower paid. The absence of a significant linkage between empathy and earnings indicates that either the positive effects of productivity and comparative advantage is offset by negative sorting due to preferences or that empathy does not affect either productivity or sorting by preferences.

Defining and Measuring Empathy

Empathy has been defined in different ways, both in academic work and in people’s everyday experiences. Empathy has both cognitive and affective components: Theory of the Mind (ToM) refers to being cognizant or knowing what another person is thinking or feeling; Sharing includes thinking or feeling the same way as the other, and compassion (empathic concern) means feeling for as well as with the other. The complexity of defining empathy is revealed by the categorization of Batson (2011) who identifies eight concepts of empathy:

• 1. Knowing another person’s internal state, including his or her thoughts and feelings

• 2. Adopting the posture or matching the neural responses of an observed other (mirroring)

• 3. Coming to feel as another person feels

• 4. Intuiting or projecting oneself into another’s situation

• 5. Imagining how another is thinking or feeling

• 6. Imagining how one would think and feel in the other’s place

• 7. Feeling distress at witnessing another person’s suffering

• 8. Feeling for another person who is suffering (Empathic Concern)

Because empathy is an internal state, measuring it usually relies on self-report.⁵ We utilize the Davis (1994) Interpersonal Reactivity Index (IRI), which is a well validated instrument for measuring empathy in the psychology literature.⁶ The IRI consists of 24 statements (such as “I try to look at everybody’s side of a disagreement before I make a decision.”) asking subjects to choose on a five point scale from A: Does Not Describe Me Well to E: Describes Me Very Well. The 24 statements are provided in Table 1, in some cases there is reverse scoring so that higher numbers always reflect higher empathy. The IRI has the virtue of representing the various categories of empathy identified by Batson with four subscales:

• Empathic Concern: experience feelings of sympathy and compassion for unfortunate others.

• Perspective Taking: spontaneously adopt the psychological of view of others in everyday life.

• Fantasy: imaginatively transpose oneself into fictional situations.

• Personal Distress: experience distress and discomfort in response to extreme distress in others.

Table 1 Davis Interpersonal Reactivity Index (IRI)

Data and Methodology

The data on empathy and labor market outcomes are taken from a longitudinal study conducted on college graduates from four colleges: Haverford College, Mills College, Wellesley College, and Santa Clara University. In 2011, 2012, and 2013, we ran laboratory experiments with seniors at the four schools to measure willingness to compete, confidence, personality characteristics, and other behaviors.⁷ From 2012 to 2018, the subjects were sent post-graduation surveys which were filled out on Google Forms. The surveys cover a broad set of questions, including occupation, salary, years at job, marital status, etc. Participants were paid $10 in Amazon.com or Starbucks gift cards in 2012–13 for filling out the survey, a sum that was increased to $20 in 2014–18 to encourage a higher response rate. The Davis IRI questions were added to the surveys in 2016 and thereafter. Therefore, most subjects completed the IRI questions more than once. In our analysis, we utilize the average of the person’s scores on the IRI if they answered more than once.⁸ Using average scores may smooth out the effects of mood on how the person responds reducing measurement error. We measure compensation as annual salary plus bonus reported by respondents. If the respondent does not report salary we use the hourly wage, multiplied by usual hours per week, to calculate the weekly salary. Then we assume 50 weeks worked per year to calculate the annual salary.

Table 2 summarizes the number of participants in each school, the number of surveys completed, and the percentage of subjects filling out one or more of the annual surveys. The response rates are quite high, 82% for the total sample and ranging from 69% to 87% for the individual schools. We do not claim that the participants in the experiment are representative of the student bodies or university graduates in general. However, within this group of people we examine the relationship of empathy to earnings.

Table 2 Experiment and Survey Samples

Results - Empathy & Salaries

Table 3 provides regressions of log salaries plus bonuses on the total IRI empathy score and each of the empathy subscales. To control for human capital (ability, work experience, and additional education beyond a college degree), we include in the regressions years since graduation, year of the postgraduate survey, a part-time dummy, GPA, and a dummy for having earned a post graduate degree. Empathy measured by the total IRI score is significantly negatively correlated with earnings, with a one point increase in the empathy score associated with 17% lower salaries plus bonuses. The three empathy subscales, empathic concern, perspective taking, and fantasy all have significant negative coefficients, with empathic concern having the largest negative effect. Personal distress has a small negative but insignificant coefficient. These results indicate a negative relationship between empathy and salaries, similar to findings for agreeableness in studies using the Big Five personality traits discussed above.

Table 3 Log Salaries & Bonuses Regressions - All Empathy Scales

These findings appear to contradict expectations that greater empathy will lead to greater career success and higher salaries. However, they are also consistent with the following: women are more empathic but earn less than men for reasons not linked to empathy; more empathic people choose majors that lead to jobs that are less well-paid (e.g. humanities as opposed to business); and more empathic people choose into sectors of employment that pay less (nonprofit as opposed to for-profit). Therefore, we cannot ascertain whether the negative correlation between empathy and salaries is due to lower productivity of the more empathic people or preferences that lead them to choose into occupations that pay less. It may also be that studying the material in a major or working in a sector of employment affects empathy in that the activities one undertakes and the persons with whom one interacts influence empathic thoughts or feelings. We can use the data to estimate a correlation between sector and major and empathy, but we cannot determine the direction of causation. However, we can see how controlling for sector and major impacts the relationship between empathy and compensation. We turn to investigating these possible interrelationships.

Table 4 provides salaries plus bonuses and the empathy scales for men and women. Consistent with most studies, we find that women in our sample have significantly higher scores then men on the total IRI and the empathy subscales and that women earn 22% less than men, with average salaries plus bonuses of $59,019 for women and $72,278 for men (significantly different at the 0.01 significance level). Therefore, gender may explain part of the negative correlation between empathy and earnings.

Table 4 Mean Salaries & Bonuses and Empathy by Sex

We examine linkages between empathy, major, and earnings in Table 5. We group majors into five categories: humanities, sciences, social sciences, engineering, and business. Column (1) provides mean salaries and bonuses for each major, ranked by amount. As expected, students who major in humanities earn the least ($44,337) while those majoring in business earn the most ($80,929). Salaries of each major grouping differ significantly at the 0.01 level from the full sample. For the most part, total IRI scores follow the same pattern, with humanities and social sciences majors displaying the highest empathy and engineering and business majors displaying lower empathy. However, sciences majors have the lowest total IRI score but earn less than engineering or business majors. Each of the major’s empathy scores are significantly different from the rest of the sample. The ranking of majors for the subscales varies somewhat, but is roughly similar to the total IRI where humanities and social sciences display the highest scores and the other three majors display lower scores with the exception of personal distress where the rankings reverse with humanities and social sciences the lowest. The personal distress scale consists of thoughts and feelings that are less positive in terms of being other-regarding (e.g. “When I see someone who badly needs help in an emergency, I go to pieces.”) and appears to measure a form of empathy that is less connected to feeling with or for others but instead focuses more on one’s own uncomfortable feelings. This type of empathy can lead to avoidance rather than prosocial behavior, i.e. leaving the scene to avoid discomfort rather than helping the victim. Overall, the evidence in Table 5 confirms linkages between choice of major, empathy scores, and earnings.

Table 5 Mean Salaries & Bonuses and Empathy by Major

Table 6 presents salaries plus bonuses and empathy scores for people working in different sectors of the economy, categorized as for-profit, nonprofit, and government, again ranked by earnings. Mean salaries are the highest in for-profit sector ($70,092) and lowest in the nonprofit sector ($44,263), and the differences in salaries are significant at the 0.01 significance level for all three categories. The total IRI score for those working in the highest paid sector, for-profit, is significantly lower than those of the other two sectors, which are similar. This also holds for both empathic concern and perspective taking. Empathy scores are not significantly different for fantasy, and the rankings again reverse for personal distress with employees in for-profit sectors having the highest score. These results confirm that those working in the highest paid sector have the lowest empathy scores.

Table 6 Mean Salaries & Bonuses and Empathy by Sector of Employment

The evidence of Tables 4, 5, and 6 shows that there are significant differences in empathy and salaries & bonuses between sexes, majors, and sectors of employment and that these differences in empathy are negatively related to earnings for each category. This suggests that people with higher empathy may be sorting into majors and sectors of employment. We investigate this by estimating multinomial logits on major and sector in Tables 7 and 8, where the omitted major is humanities and the omitted sector is for-profit. We control for female in these equations since females are also more likely to be in low paying majors and low paying sectors.

Table 7 Multinomial Logits – Sorting into Major (Omitted category is humanities)

The results in Table 7 support the hypothesis that empathy is associated with sorting into majors. For the total IRI, the coefficients on engineering and sciences are negative and significant. Further, the joint hypothesis that empathy sorts by major has a p-value of 0.088. Empathic concern and fantasy appear to drive this result. Surprisingly, distress is also a significant predictor of sector choice but predominantly because it increases the probability that a respondent is a business major.

The evidence of sorting into sector in Table 8 is strongest for empathic concern, which has a strong positive relationship between empathy and choosing employment in the nonprofit and government sectors with a p-value of 0.045. For perspective taking, there is also a positive relationship with a p-value 0.097. Neither fantasy nor personal distress have significant coefficients and p-values indicate lack of significance, contributing to the result of the total IRI also not having significant effects.

Table 8 Multinomial Logits – Sorting into Sector of Employment. (Omitted category is for-profit sector)

We next examine whether sex, major, or sector of employment explain the negative association between empathy and earnings. We estimate log salary & bonuses regressions with dummies for sex, major, and sector of employment, beginning with just including sex, then successively adding majors and sectors. The full model regressions are provided in Appendix Table 11. The dummies for sex, major, and sector of employment all have significant coefficients in the regressions indicating that these variables are important determinants of earnings. In Table 9, we summarize these results by providing the coefficients on the empathy variables of the original regressions with basic controls from Table 3 in column 1 (basic controls are years since graduation, year of the postgraduate survey, a part-time dummy, GPA, and a dummy for having earned a post graduate degree). In column 2 we add a dummy for female. Including the female dummy reduces the effect of empathy on salaries & bonuses from – 0.173 to – 0.128, a reduction of about a quarter. However, empathy continues to have a large significant effect on earnings beyond differences caused by gender. Examining the effects by empathy subscale, empathic concern (row 2), perspective (row 3), and fantasy (row 4) all have negative significant coefficients with the largest effect coming from empathic concern. In the third column, we add dummies for major (with the omitted category humanities) continuing to include the female dummy. The coefficient on total IRI declines only slightly and remains significant at the 0.01 level. Both empathic concern and fantasy continue to reveal significant effects, but the effect of perspective declines and is no longer significant. Finally, in column 4 we add sectors of employment to the other controls and the results do not change much from column (3). Therefore, even after including sex, major, and sector of employment we continue to find a negative relationship between empathy and earnings. While there are differences in empathy by sector and major, these differences do not seem to explain the differences in salary by empathy once we control for gender. Further, the magnitude of empathy on salary is not small with a one standard deviation increase in the IRI scale (0.4) corresponding to a 4.5% decrease in earnings, controlling for sex, major and sector.

Because of well-documented differences between men and women in occupations and salaries, we estimate the salaries plus bonuses equations for men and women separately including the full set of controls. Table 10, column 1 presents the coefficients on the total IRI empathy score for men and women. The results reveal that the negative empathy effect on salaries plus bonuses is significant for both men and women, and, while the coefficient is larger for men, significance testing reveals that the effects are not statistically different.

Table 9 Coefficients on Empathy Measures in Log Salary plus Bonus Equations with Increasing Numbers of Controls

Table 10 The effect of total IRI scale on compensation by sex, major and sector

With the persistence of the negative salary effect of empathy, we investigate whether the effect is consistent across sector and major, or whether there are only certain majors and sectors where productivity differences or further sorting may be present. We estimate the salary plus bonuses equations within each major and within each sector. Column 2 in Table 10 displays the coefficients on the total IRI for the different majors. The coefficients reveal that the effect is significantly different from zero only for social science majors. This suggests that for the other majors, once preferences lead people to choose their major, there is not a negative productivity effect of empathy on earnings, while for social sciences majors, the negative relationship persists. This could be due to lower productivity of the more empathic or to further sorting into service oriented jobs that hire social sciences majors. However, the lack of significance of the negative coefficients on the total IRI for sciences and business students may be due to the smaller number of observations for these majors, and significance testing cannot reject the hypothesis that these negative coefficients equal that of the social science majors.

Similarly, in column 3 of Table 10 where the regressions are run by sector, the coefficient on empathy is significant only for the for-profit sector, although one cannot reject the hypothesis that the coefficient is equal across sectors, and point estimates of the coefficients for the government and for-profit sectors are almost identical.

These results are suggestive that the negative empathy differential holds only for a subset of majors (business, sciences, and social sciences) and a subset of sectors (for-profit and government). It may be that the kind of jobs in these sectors and held by these majors punish empathy; empathic workers are less productive because of lack of appropriate skills or inability to thrive in work environments that characterize these jobs. Alternatively, the jobs held by humanities and engineering majors and nonprofit workers do not punish and may even reward empathy.

While the evidence supports sorting into majors and sectors based on preferences, it is not consistent with sorting based on comparative advantage. For example, empathic individuals are least likely to be engineers, whose jobs do not seem to punish empathy, and government employees have high empathy scores, even though the negative point estimate for the empathy coefficient for government workers is above 10 %. Therefore, it does not seem that empathic workers are sorting into jobs where they are more likely to have comparative advantage and be rewarded.

It may be that there is additional sorting among the jobs within these sectors and held by these majors that is the cause of the negative coefficient on empathy; empathic individuals are further choosing jobs whose services, work environments, or clients meet their preferences and that are relatively low paying. Certainly the jobs within the for-profit and government sectors are quite varied and the students majoring in social sciences, business and science have varied skills and interests. Unfortunately, with the current data available, we cannot distinguish between preference sorting and a causation, low productivity explanation of the negative linkages between empathy and earnings.⁹

Conclusion

In this study, we examine the linkages between empathy and earnings of college graduates from four institutions utilizing six years of survey data. While it might be thought that empathy would enhance productivity and wages (e.g. via better personal skills that foster working with others and enhance leadership ability), it is also possible for greater empathy to have a negative impact on productivity (e.g. through less commitment to the profit-making goals of an employer). In addition, greater empathy would likely have different effects on productivity in different occupations and industries so that empathy might be rewarded or punished depending on the job and sector. Comparative advantage implies that people would sort into occupations that reward their skills so that the more empathic should sort into employment that values empathy, increasing their productivity and earnings. At the same time, more empathic people may have preferences for jobs that provide services and caring for others and where workplaces are less competitive. It is likely that in the U.S. such occupations pay lower salaries.

Our empirical results show that in our sample of college graduates, on balance empathy has a negative relationship with compensation and this is true for both men and women. The negative effects persist even with the inclusion of controls for gender, human capital variables, major, and sector. While we acknowledge that we can only document a strong negative correlation not causation, we do find evidence of sorting into major and sector consistent with preferences that are likely to be held by more empathic people: humanities and social science majors score higher on the empathy scale while business and engineering students score lower, and workers in nonprofit and government sectors score higher while those in for-profit sector have lower empathy scores. This evidence is supported by multinomial logits. However, we cannot rule out that empathy levels of people in certain majors or employed in some sectors may be affected by experiences and people with whom they interact so we cannot claim causation.

Examining the variables within major and within sector shows that the negative link between empathy and earnings continues to be significant only for social science majors and individuals working in the for-profit sector. However, the coefficients on empathy for business majors and sciences majors are not significantly different from that for social sciences and the coefficient for government sector does not differ from for-profit sector. Therefore, there continues to be evidence within some majors and sectors of a negative link between compensation and empathy, while there is no such negative relationship for humanities or engineering majors or those working in the nonprofit sector. It may be that more detailed information on jobs within major and sector would show further sorting by preferences but the negative link between empathy and earnings could be due to lower productivity of empathic workers. However, a sorting story in which high empathy individuals sort into low paying jobs that satisfy preferences for a certain type of service or work environment does not rule out the possibility that within a job, more empathic individuals are more productive. Our data does not preclude this but much more detailed information on jobs would be needed to test such a theory.

Appendix

Table 11 Full Models of Log Salary & Bonus Equations with Different Empathy Scales