VAT threshold and small business behavior: evidence from Thai tax returns

Abstract

This paper examines how small firms respond to the VAT registration threshold in the context of high informality. Using the panel of VAT and corporate income tax return data from Thailand, we study bunching response at the threshold, examine the role of informality, and investigate growth effects. We find that Thai firms respond strongly to the notch with the bunching pattern consistent with the incentive at the threshold. We also present suggestive evidence that highlights the VAT informality (defined as the presence of non-VAT firms) as a key incentive underlying the bunching response. Finally, our findings indicate that the threshold provides a brake on small business growth. We demonstrate that firms not registered for VAT have significantly lower growth rate than a propensity score-matched group of firms that voluntarily registered. Such effects are stronger for firms closer to the threshold with part of them likely driven by misreporting. These findings have important implications for the VAT threshold policy in developing countries and are widely applicable since the size-dependent thresholds are commonly applied to small businesses.

Appendix

A. Sensitivity tests and full regression results

See Tables 9, 10, 11, and 12.

B. First stage of the propensity score estimation for the growth effect analysis

In the first stage of the propensity score estimation for the growth effect analysis, we estimate a logit model with a treatment dummy as a dependent variable. This can be written as:

$$\begin{aligned} {\text{treat}}_{it} & = \alpha_{0} + \alpha_{1} {\text{imm}}_{it - 1} + \alpha_{2} {\text{com}}_{it} + \alpha_{3} {\text{inf}}_{it} + \alpha_{4} d_{it - 1} \\ \quad + \alpha_{5} {\text{age}}_{it} + \alpha_{6} {\text{toas}}_{it} + \gamma_{s} + \gamma_{t} + \varepsilon_{it} \\ \end{aligned}$$

(6)

where ${treat}_{it}$ equals 1 for firms that are below the threshold and have never registered for VAT and 0 for firms that are below the threshold and voluntarily register for VAT, ${age}_{it}$ is firm age, ${toas}_{it}$ is total assets (in log), and all other variables are defined as in Eq. (3). We also include industry- and year-fixed effects (${\gamma }_{s}$ and ${\gamma }_{t}$). For each observation, we use this regression to compute the propensity score which is the predicted probability of being in the treatment group. Table 12 illustrates the logit regression estimate.

Table 9 Sensitivity tests for the voluntary VAT registration analysis.(Dep var = voluntary registration indicator)

Table 10 Impact of the VAT regression threshold on revenue growth using propensity score-matched sample (Dep var = Revenue growth)

Table 11 Estimates using alternative matching approaches for the impact of the VAT regression threshold on revenue growth using propensity score-matched sample (Dep var = Revenue growth)

Table 12 Logit regression estimate for the first stage of the propensity score matching analysis (Dep var = Treatment dummy)