Using responses from an incentivized tax forecasting task, we estimate the prevalence of previously discussed heuristics for simplifying tax forecasts (Liebman and Zeckhauser, 2004). We find strong evidence for "ironing" (linearizing the tax schedule using one's average tax rate), no evidence for "spotlighting" (linearizing the tax schedule using one's marginal tax rate), and we identify the qualitative features of the remaining misperceptions that are not captured by existing models. We then embed these misperceptions in a standard model of income taxation and study a social planner's decision to "nudge" taxpayers. We find that a social planner would not choose to correct the misperceptions that we estimate because they are helpful in achieving redistributive goals.
↧