Christina and David Romer have a new paper (pdf), focusing on the interwar era:
This paper uses the interwar period in the United States as a laboratory for investigating the incentive effects of changes in marginal income tax rates. Marginal rates changed frequently and drastically in the 1920s and 1930s, and the changes varied greatly across income groups at the top of the income distribution. We examine the effect of these changes on taxable income using time-series/cross-section analysis of data on income and taxes by small slices of the income distribution. We find that the elasticity of taxable income to changes in the log after-tax share (one minus the marginal rate) is positive but small (approximately 0.2) and precisely estimated (a t-statistic over 6). The estimate is highly robust. We also examine the time-series response of available indicators of investment and entrepreneurial activity to changes in marginal rates. We find suggestive evidence of an impact on business formation, but no evidence of an important impact on other indicators.
For the pointer I thank Ilya Novak. I have not yet read the paper.