Overall, American college students scored lower on EC and PT between the 1979 and 2009 (see Figures 1 and 2). There is a significant negative correlation between year of data collection and EC (b = –.38, p = .002, k = 66) and PT (b = –.27, p = .03, k = 64) when weighted by sample size.3 There were no significant changes in either the FS subscale (b = –.19, p = .26, k = 37) or PD (b = .09, p = .55, k = 46). Thus, more recent generations of college students are reporting less EC and PT, which are the most central components of empathy.
We calculated effect sizes by calculating the difference between the average scores for the earliest and latest year in our sample, taking into consideration the average standard deviation for each IRI subscale. For the EC subscale, the regression equation (EC mean = –0.0140 × year + 31.771) yields a score of 4.06 for 1979 and 3.64 for 2009. Considering the average EC standard deviation of 0.6508, there was a drop of 0.65 standard deviations over time, which is a medium to large effect size (medium = 0.50 and large = 0.80) by Cohen's (1977) guidelines. For the PT subscale, the regression equation (PT mean = –0.0099 × year + 23.349) yields a score of 3.66 for 1979 and 3.36 for 2009. The average standard deviation reported in the individual PT samples (from the data we collected) is 0.6786. Thus, PT scores decreased 0.44 standard deviations from 1979 to 2009. This is a small to medium effect size (small = 0.20 and medium = 0.50) by Cohen's guidelines. (The effect sizes for FS and PD are d = 0.17 and d = –0.21, respectively.) Converting the changes in EC and PT to percentile scores is also informative. If the average student in 1979 scored at the 50th percentile of the distribution of EC or PT, the average student in 2009 scored at the 26th percentile of EC and the 33rd percentile of PT (a**uming a normal curve). In other words, between two thirds and three quarters of recent college students are below the 1979 PT and EC means, respectively. This represents a 48% decrease in EC (26 out of 100 in 2009 vs. 50 out of 100 in 1979) and a 34% decrease in PT (33 out of 100 in 2009 vs. 50 out of 100 in 1979). Gender and Ethnic Background Women tend to score higher than men on each of the four subscales of the IRI (Davis, 1980), so there may be interesting gender differences in how empathy is changing over time. Unfortunately, means were presented separately by gender in only 14 of the 72 samples, making it impossible to examine differences in men's versus women's scores over time. When means were presented separately by gender, we calculated the average score for participants, weighting by the number of males and females.
Of the 72 samples, 69 included the number of males and female participants, thus allowing us to examine whether our effect changes when controlling for the percentage of the sample that was male (range = 0% to 100%, M = 36.9%). We first looked for effects of gender proportion on empathy by regressing the percentage of male participants onto each IRI subscale (weighting for number of participants) and found no effects of percentage male on EC (b = –.17, p = .19, k = 64), PT (b = –.14, p = .28, k = 62), or FS (b = –.17, p = .32, k = 36), although samples with higher percentages of male participants had lower PD scores overall (b = –.34, p = .02, k = 44). We next examined the effect of year of data collection on each IRI subscale when controlling for percentage male. Our results remain similar as in the original an*lysis: EC (b = –.43, p < .001, k = 64, d = 0.75), PT (b = –.31, p = .02, k = 62, d = 0.51), FS (b = –.21, p = .23, k = 36, d = 0.19), PD (b = .009, p = .96, k = 44, d = –0.02).
Unfortunately, only 36 of the 72 samples reported the ethnicity of the participants (M = 69.0% Caucasian), which makes it not viable to reliably examine the effect of year of data collection on each IRI subscale. However, just for the sake of reporting the data, we did find that the samples with a higher percentage of Caucasian participants had lower EC (b = –.44, p = .009, k = 34), PT (b = –.36, p = .04, k = 33), and PD scores (b = –.51, p = .02, k = 21). There was no relationship between the percentage of Caucasian participants and FS scores (b = –.27, p = .29, k = 17). Publication Status. We next an*lyzed the data excluding the nine unpublished samples (including two from our own lab) and found nearly identical results. There were similar declines in EC (b = –.40, p = .002, k = 58, d = 0.71) and PT (b = –.27, p = .04, k = 55, d = 0.47) and again no changes in FS (b = –.29, p = .13, k = 30, d = 0.27) or PD over time (b = .09, p = .61, k = 39, d = –0.20). Economic Variables It is possible that declines in empathy can be explained by parallel changes in economic variables over time. There is evidence, for example, that being primed with money-related concepts leads to less helping behavior (Vohs et al., 2006). Thus, in times of greater economic prosperity people may be less empathic. We therefore tested whether observed changes in empathy over time would remain after controlling for measures of general economic health. We collected national statistics on annual average unemployment rate (civilian labor force, age 16 or older) and annual average inflation from 1979 through 2009 (U.S. Bureau of Labor Statistics, 2009a; 2009b). Overall, unemployment significantly declined from 1979 to 2009 (b = –.30, p = .01), whereas inflation has also been declining (b = –.49, p < .001), suggesting increasing economic prosperity from the late 1970s until the late 2000s, despite the economic downturn in 2008–2009. When examining the effect of unemployment and inflation on each of the IRI subscales, we find no significant relationships (ps > .10).
Importantly though, even when controlling for both of these variables simultaneously, EC (b = –.75, p Time Period As there were only seven samples collected before 1990, we also ran the regression an*lysis for samples collected from 1990 to 2009. The results were similar. There remained a negative correlation between year of data collection and EC (b = –.50, p < .001, k = 61, d = 0.95) and PT (b = –.24, p = .075, k = 57, d = 0.40). FS (b = –.24, p = .18, k = 32, d = 0.26) and PD remained nonsignificant (b = –.06, p = .73, k = 41, d = 0.14). We next split our data set into two time periods to examine whether the decreases in empathy were specific to more recent time periods. When the an*lysis was restricted to the years 1979 to 1999, we no longer found changes in any of the IRI subscales: EC (b = .16, p = .43, k = 26, d = –0.12), PT (b = –.11, p = .59, k = 26, d = 0.13), FS (b = .22, p = .37, k = 18, d = –0.10), and PD (b = .09, p = .71, k = 21, d = –0.09). When examining changes in the IRI between 2000 and 2009, however, we found that the declines were most pronounced for EC (b = –.44, p = .004, k = 40, d = 0.83) and PT (b = –.31, p = .06, k = 38, d = 0.55). The changes in FS (b = –.16, p = .52, k = 19, d = 0.21) and PD (b = –.28, p = .19, k = 24, d = 0.82) were again nonsignificant. Taken together, this an*lysis suggests that empathy has been decreasing in college students primarily since 2000.