Structural parameter identification using causal inference: The income elasticity of demand for culture

This paper proposes a strategy to estimate the income arc-elasticity of the demand for culture and provides evidence using the Spanish household budget survey. The idea is to interpret income-expenditure pairs as potential outcomes. Based on this interpretation, we use employment variation of household members as a source of within subject variation. The procedure uses the semi-parametric difference-in-difference and matching methods to estimate two treatment effects: the treatment effect of unemployment on cultural expenditure and the treatment effect of unemployment on income. The income arc-elasticity is then computed as a function of those treatment effects. The main results are as follows. First, when a household member enters an unemployment spell, households reduce cultural expenditure and their income falls significantly. The derived income elasticity estimates are economically sensible and comparable to those based on more traditional methods. Second, a placebo analysis suggests that tests used for inference have the correct empirical size. Third, a conditional on positive analysis indicates that households' participation in cultural markets does not change significantly as a result of unemployment. Within the subset of participating households income elasticity is higher. Fourth, unemployment generates an income and cultural expenditure decrease which is larger when the unemployed household member is a man than when the unemployment person is a women. Similarly, when a household member with tertiary education enters an unemployment spell, household income falls more and cultural expenditure less than when unemployment affects a household member with less than tertiary education. Fifth, the effect of a household member entering unemployment is fairly symmetrical, although of the opposite direction, with respect to the effect of a household member leaving the unemployment state as long as that household member is the only one unemployed. When we compare the effect of a second household member entering unemployment with the effect of the second household leaving unemployment, the effects are not symmetric. This is evidence of hysteresis in the relationship between household expenditure and unemployment. Sixth, business cycles mediate in the relationship between household expenditure on culture and unemployment. When the economy is in an expansion, unemployment does not have a significant effect on household cultural spending, while it does have a large and significant effect in the negative growth period.