Battle of the ballet household decisions on arts consumption

Abstract

Women and men differ in their tastes for the performing arts. Gender differences have been shown to persist after accounting for socioeconomic factors. This paper uses this difference to shed light on how decisions on arts consumption are made in households. Based on relatively recent theoretical developments in the literature on household decision-making, we use three different so-called distribution factors to show for the first time that the relative bargaining power of spouses affects their arts consumption. Using a sample from the US Current Population Survey, which includes data on the frequency of visits to cultural activities, we regress attendance on a range of socioeconomic variables using a count data model. The distribution factors consistently affect attendance by men at events such as the opera, ballet and other dance performances, which are more frequently attended by women than by men. We conclude that when men have more bargaining power, they tend to attend such events less frequently.

Appendix

Table 8 shows results of a regression similar to that reported in Table 5. The difference is in the link function chosen. Instead of a zero-inflated negative binomial, here a negative binomial model was used. This model is reported here as a robustness check. Notably, the effect of the distribution factor on attendance at female-dominated events remains significant and very similar in magnitude, still implying the same $4\%$ reduction in the number of visits per year for a $10\%$ shift in income in favor of the husband.

Table 8 Determinants of cultural consumption—Negative binomial

Table 9, which comes in two parts, shows results for regressions on attendance by either member of the couple for each of the nine activities available in the data. This allows for comparisons of individual attendance across activities. Unfortunately, the effects of the distribution factor are much harder to detect here, both because visits by both members of the household are considered and because there are very few nonzero observations for some of the activities. This results in a lack of variation in the dependent variable on which to base the estimation. Especially Latin and salsa performances, the opera and ballet are attended so rarely overall that inference becomes difficult. For the same reason, inclusion of state fixed effects becomes hard here, and they are omitted.

Table 9 Annual attendance by either partner—Zero-inflated negative binomial

The income share variable is significant at the 1% level only once, for jazz music, and though the sign is consistent with the previous result that relatively wealthier men depress attendance in all categories, there is no particular reason why this effect should be strong for jazz and not for the others. Again, the relatively small numbers of nonzero observations may not be sufficient to see through the noise here.

Other effects are strong enough though, to be detected in the majority of categories. These are the usual suspects, log income and education. Log income very consistently reduces the probability of never going in the inflate part of the models. The fact that this variable has this prominent role in determining who does and does not go at all limits the interpretability of its coefficient in the main part of the model, where it is found significant only in the case of classical music. Theoretically, this means that higher income increases the probability of going at least once per year while decreasing the expected number of attendances among those who have the means.

Education, both of the man and the woman, has a similarly consistent role. The associated coefficients are mostly significant, greater than one whenever they are significant and often very large. One exception to this rule is the Latin and Salsa category, which is unusual in other ways also. This may be partly due to the fact that though it is performance art, it is less high-brow in character than most of the other activities. But again, the oddity may also be the result of the small number of nonzero observations (or of the omission of a variable indicating Hispanic origin, which may be important for this category but not for the others).

The effects of education are especially strong for very high-brow activities such as classical music, the opera and dance performances, where the latter two are especially influenced by female and male education, respectively.

Another trend that is relatively consistent comes from the dummy for data from the year 2012, the second wave considered here. Coefficients are mostly below one, indicating a negative trend in the number of visits across activities.