Research on the causal effect of corruption on economic performance relies on instrumental variables that can distinguish this relationship from the reverse causal effect of economic development on corruption. Unfortunately, good instruments are hard to find and the few instruments used in this field have been subject to serious criticism. In order to expand the universe of possible instruments, I propose and validate a new strategy for finding instrumental variables: if two exogenous, non-instrumental variables v and w have a conditional relationship with x but an unconditional relationship with y, then the interaction term vw serves as an effective instrument to identify the causal effect of x on y. Using this strategy, I propose and validate three new instruments for corruption and use them to measure the degree to which corruption lowers per capita GDP in contemporary democracies.
How Much Does Corruption Harm Economic Performance? Using One-Sided Conditional Relationships as Instrumental Variables for Causal Identification
Room
409