In contrast to the estimates for the 1890s, elasticities for men estimated for 1973 and 1991, although negative, are fairly small. For workers paid by the hour they are positive in 1991. For women elasticities are positive and around 0.11. Of course, the negative labor supply elasticities estimated for 1973 and 1991 may well be spurious. Estimates derived from modern panel data sets suggest that changes in hours worked are positively related to increases in wages upon instrumenting, even when ordinary least squares indicates that the relationship is negative (e.g. Lundberg 1985). If so, then increasing earnings inequality in recent times may have led to increasing hours inequality. Regardless of whether recent labor supply elasticities are positive or small and slightly negative, the comparison with past labor supply elasticities suggests that at least between the 1890s and 1973 the labor supply curve has become less backwards bending. The next section examines potential explanations.
Table 8 summarizes changes in the supply of daily hours by wage percentile. Between the 1890s and 1973 hours supplied by men in the bottom wage decile fell much more than hours supplied by men in the top wage decile. But, between 1973 and 1991 hours supplied by men in the top wage decile rose while those supplied by men in the bottom decile fell.17 For women hours supplied by those in the top decile rose between 1890 and 1973 and those supplied by women in the bottom decile fell. Between 1973 and 1991 hours supplied by women in the top decile continued to rise while those of women in the bottom decile remained unchanged. A comparison of the size of the relative demand and supply shifts suggests that for both men and women changes in labor supply dominated changes in labor demand.
The Labor Supply Curve
Why is the labor supply curve of daily hours worked now less backwards bending? Within a static labor supply model, the most common explanation is a falling income effect, but when I include measures of home value, the income of other family members, and savings in the 1890s regression, I obtain income elasticities that are close to zero.18 Another explanation is that of Barzel and McDonald (1973) whose model predicts that when assets are below subsistence the labor supply curve is backwards bending and that when assets are above subsistence the labor supply curve is forwards bending. But, inconsistent with the predictions of their model, the elasticity of labor supply in the 1890s is not positive for men with large asset holdings (as measured by home value).