Expected Taxes and Household Consumption Behavior
In this dissertation I ask two basic questions: First, how predictable are personal income tax changes in the U.S. and second, does household consumption respond to news about future tax changes, or does it mostly respond at the time when the tax rates actually change? These are interesting questions because they have broad implications for macroeconomics and public economics. The rational-expectations life-cycle theory of consumption is the workhorse in modern macroeconomics. While there are various specifications of this theory, two predictions are common across them. First, consumption should not respond to predictable income changes and second, consumption should respond to news about future after-tax lifetime income. There is a large literature that tests the first implication of the rational-expectations life-cycle theory and generally rejects the model by finding significant consumption responses to predictable income changes -- that is, it finds that consumption is in fact excessively sensitive to predictable income changes. Very few studies focus on the theory's second main prediction, that household consumption responds to news about after-tax income changes, even if current after-tax income has not changed yet. To the best of my knowledge this dissertation is indeed the first study to use micro-level data to estimate the consumption response to news.
I use fiscal policy to study these two questions because it offers two main advantages over other empirical frameworks commonly used by macroeconomists to test the consumption theory and to analyze the effect of news on macroeconomic aggregates. First, exploiting the fact that there is a lag between the decision to change taxes and the implementation of the tax changes allows me to separate the behavioral response to news from the response to the actual policy changes. Therefore, the response to tax news is not confounded by the response to the actual tax change. Second, actual tax changes are directly observable without measurement issues, which is different from other news shocks that have been recently studied, in particular news about future total factor productivity. Therefore, my measure of news about future taxes can be directly compared with the actual evolution of the tax rates.
Regarding public economics, this dissertation addresses another question that is of interest to public policy makers. During the current Great Recession, in which conventional monetary policy is not effective due to the zero lower bound on nominal interest rates, policy makers have shifted attention to fiscal interventions. In order to assess the effectiveness of fiscal policy we have to know the total effect of a tax reform on the economy, i.e. the tax multiplier. Unfortunately, almost all studies that provide estimates of tax multipliers focus on the response of the economy to actual tax changes. These estimates might miss a fraction of the total effect of a tax reform if tax changes are predictable and if the behavior of economic agents is forward-looking. Ignoring anticipation effects can therefore bias the tax multiplier downward.
The identification of news about future tax rates is key for answering these questions. In this dissertation I exploit the fact that there exist two classes of fixed-income securities in the U.S. that are very similar except for the tax treatment of their income streams. Interest on municipal bonds is tax-exempt, while interest on Treasury bonds is subject to federal income taxes; thus, relative price changes between municipal and Treasury bonds reflect changes in expected future tax rates, holding fixed other risk factors. I go beyond identification of the timing of news to directly measure the entire path of expected tax rates. The fact that different bonds have different maturities quantifies the degree of tax foresight, since yield spreads of bonds with different maturities reflect information about future taxes over different horizons. Hence, the tax news shocks derived from the bond prices measure not only when households receive information, but also what information they receive.
Identifying the entire path of expected tax rates in turn is important for testing the basic rational-expectations life-cycle model of consumption, as the theory predicts that consumption responds one-for-one to changes in expected after-tax lifetime income. The term structure of municipal yield spreads identifies the expected persistence of a tax shock, which is a crucial factor that determines the optimal consumption response according to the theory. For instance, if a tax change is expected to be only transitory, then the theory predicts that consumption does not respond much. On the other hand, if a tax reform is expected to have a large persistent component, then consumption should respond much stronger.
Combining these market-based tax expectations with consumption data from the Consumer Expenditure Survey I find that consumption of high-income households increases by close to 1% in response to news of a 1% increase in expected after-tax lifetime income, consistent with the basic rational-expectations life-cycle theory. On the other hand, households who have lower income, less education, or are more credit constrained respond less to news. However, the same households also respond one-for-one with large news shocks, consistent with rational inattention.