Dynamics of evasion around tax thresholds: Evidence from Indian firms

How do firms respond to size-based thresholds when dynamic responses matter? Using the context of an Indian revenue-based tax registration threshold that selectively affected some firms, coupled with administrative tax data, we examine how firm growth responds to a threshold. We find that firms respond by slowing down growth in reported revenue below the threshold by around 14 percentage points or roughly 42 % of average growth. A lack of corresponding change in reported costs suggests an evasion response rather than a real response by firms. We show that an increase in audit probability on crossing the threshold, rather than an increase in tax liability or compliance costs is the likely reason for firm slowdown. We rationalize our empirical findings by modifying the standard Allingham-Sandmo model of evasion to a dynamic setting. We use this framework to calculate deadweight loss due to a threshold when dynamic responses are relevant and find that the welfare cost can be substantial in the long run.