Looking at variation over time as well as across racial groups, we find additional evidence against the compassion theory. Where the compassion theory would seem to suggest that the gender gap on social welfare issues should exist across time and political context, we find that issue-position gender gaps often emerged later than partisan gaps. There were no apparent or consistent differences between men’s and women’s positions on, for example, government provision of health care prior to 1988 (or 1992), of social welfare services prior to 1989, of spending on AIDS and HIV care except in 1994 and 1996, and of jobs prior to the 1970s.

Even on perhaps the prime “compassion” issue, federal spending on the poor, the gender gap was not consistently present from year to year. Males differed significantly from females in attitudes towards spending on the poor only during presidential election years between 1984 and 2006 (with the exception of 1988) and a brief period from 1991 to 1994, where gaps were somewhat significant in nonelection years. A similar link between elections and male attitudes toward whether the government was “too strong” or “not strong enough” is also noteworthy, given that a white gender gap on this question is only found in presidential election years prior to 1980 (after which time it is found consistently).

Notably, these issues stand in contrast to noncompassion issues, for which there is a consistent gender gap, often across racial lines and, again, often in the “reverse” direction. Recall that, as shown in Table 1, more men than women prefer increased spending on many discretionary domestic federal budget items. These gender differences hold across racial lines.

As most frequently proposed in the gender gap literature, the state dependency theory is a straightforward set of claims involving women’s self-interest. The premise is that women, more often than men, are economically vulnerable and actually or potentially dependent on state welfare provision. Thus, they will favor such programs out of sheer self-interest. In turn, they will favor the Democratic Party and its candidates because they, too, favor more spending on social welfare programs.

The self-interest hypothesis has been disputed on many grounds. However, since it continues to be seriously entertained in widely read scholarly journals, it seems worth further debunking with the following straightforward test: If women’s greater reliance on social welfare spending is driving the gender gap, then the gender gap should be primarily a lower-class or lower- and middle-class phenomenon. As seen in Figure 2, the data show that the opposite is true.

Far from being the province of the lower echelons, gender gaps among white Americans are larger and more consistent among the respondents whose income places them into the upper third of American families, precisely the stratum in which women are least likely to need (or have ever needed) state welfare programs for economic support. More women than men in the highest income bracket voted for the Democratic presidential candidate in every election from 1968 to 2004, a gap that is consistently statistically significant from 1988 to 2004 and close to statistical significance in 1968, 1980, and 1984. Among respondents whose income falls into the lowest third of household income, there is no discernible gender gap in any year except 1996.

Another version of the state dependency argument suggests that women’s greater dependence on or closeness to the government leads to greater support for government spending and government programs in general, as opposed to market-based solutions to social problems. Table 1 shows no evidence of a consistent female preference for government spending in general. Men, more often than women, preferred increased spending on many government programs, including parks and recreation, scientific research, space exploration, transportation infrastructure including highways and mass transit, and the military and defense.