This paper uses panel data from the Small and Medium Enterprise Survey in Vietnam from 2005 to 2013 to investigate the incidence and size of corruption in Vietnam. The Heckman’s two-step model is employed to take into account censored nature of the data on bribes and sample selection bias. We find strong evidence that the propensity to bribe as well as bribe amounts are highly positively correlated with interaction level with government officials, firms’ ability to pay, and regulatory-type burdens imposed on firms. In addition, firms without official business registration licenses are more likely to avoid paying the informal costs. These results are robust when lagged values of profit are used as instruments for profit.