Location-Based Tax Incentives for Non-Farm Rural Enterprises in Armenia

Abstract

The paper analyses the effect of location-based tax incentives on rural non-farm small business performance in Armenia. It uses administrative panel data for the population of the registered taxpayers in the areas concerned, and the vast tax exemptions are granted for businesses operating in the selected set of border communities. Using an ‘event study’-style difference-in-differences fixed effects estimator we initially find a negative effect of tax exemptions on the reported gross income of the affected enterprises and a positive effect on employment, mainly driven by a smaller subset of non-trade firms. To uncover these inconsistent results, we consider a possible channel for decreased reported revenues – changes in audit probabilities for the exempt businesses. Indeed, introducing controls for predicted audit probability eliminates the negative effects on gross revenues, while the positive effect on the number of employees persists, with even increased magnitudes. Gross income under-reporting incentives could emerge in the supply chain where the suppliers might be inclined not to report their sales thus inducing the exempted businesses to under-report their incomes as well. Given the short pre-treatment data, the long-term development common trend of treated and control communities is verified using satellite data on night-time illumination.

Keywords:

• Place-based tax exemptions
• tax audit
• rural non-farm business
• small and micro enterprise performance

Acknowledgements

The authors are grateful to Zareh Asatryan for his valuable comments on the earlier version of the manuscript and indebted to the State Revenue Committee of the Republic of Armenia for providing opportunity to conduct research on the administrative dataset of taxpayers in the regions concerned. The authors did not receive support from any organization for the submitted work.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The datasets generated and analysed during the current study are not publicly available due to data privacy requirements imposed by the State Revenue Committee but are available in anonymised format from the corresponding author on reasonable request.

Notes

1 Here and after enterprises comprise physical and legal persons engaged in business activities, and we will refer to them as taxpayers, enterprises and businesses interchangeably.

2 HO-156-N-Tax exemption of activities realized in the border region, RA 2015.

3 In comparison, generally in Europe and Central Asia region these indicators were 20 and 5% accordingly.

4 Actually analyzed data differs as we remove inactive taxpayers as well as those, which end up in the top 1% revenue earners to ensure no outliers drive the effects observed. In addition, we treat the years when zero gross income is reported as an inactive period for the taxpayer and assign missing value. If zeros are taken into account, the total number of observations is around 18,000. More information on this is available in the Data section below.

5 The research on developed countries is also available, we do not explore that here due to focus of our study and space considerations. Interested reader can refer, for example, to Givord et al. (2013) who examine the effect of EZ on economically depressed or moderately depressed urban areas in France or to Ambroziak & Hartwell, 2018 who investigate the effects of Polish SEZ on economic growth.

6 Marzes are major administrative units, Armenia consists of ten marzes and the capital city of Yerevan.

7 The list of these communities was determined by another legal act much earlier, in 2002 with a Government decree establishing list of border communities which have limited capacity to utilize the agricultural land due to consequences of military conflict with Azerbaijan (N 163, February 21, 2002).

8 The taxpayers in Armenia have an option to opt for simplified tax regime with turnover tax only levied upon gross income in cases when the income is less than certain threshold value. If the gross income is higher than the threshold, the conventional value added tax (VAT) and corporate profit taxes are applied.

9 We use the stata routine developed by Clarke and Tapia-Schythe (2021) –eventdd-.

10 We consider alternative definitions for post-treatment periods.

11 If zero income years are included the total number of observations is around 18,000.

12 Therefore, the labeling was not done based on the tax-exempted locations, but rather, based on the tax declarations, that is whether the borderline trade or production was actually declared. In particular, the electronic tax return system utilized by all business taxpayers in Armenia envisages special data input fields for such taxpayers

13 We are thankful to the anonymous referee for highlighting this important aspect of possible consequences of policy implementation.

14 Supplementary Appendix Table A3 contains the report of the marginal effects and the model applied to obtain this result.

15 This would at least require identification of community of the taxpayer so that the difference in differences framework could be applied to estimate the extensive margins on registrations.

16 The new government indicated improved tax revenue collection as one of its primary objectives together with its fight against corruption. So the 2018 results can bear some of the consequences of these mobilization efforts (UNDP, 2021)

17 Audit lists are predominantly constructed using a risk-based approach, which utilizes historical data on taxpayers. SRC mentioned that even when the system signals a treated community taxpayer for audit, it might be well removed from the list because of absence of any tax obligation on the part of such taxpayers. The only exception is the audit regarded unregistered employees as the income taxes for employees are still payable.

18 The primary audit mechanism for detecting underreporting of revenues in retail trade is the “mystery shopping” audits by tax agents. The reports in media by SRC officials confirm that the majority of underreporting is happening in rural areas and in small and micro taxpayers (for example, an interview by deputy head of State revenue committee Mr. Mashadyan in armtimes.com journal published on December 30, 2019 available in Armenian language at: https://armtimes.com/hy/article/177189?fbclid=IwAR28s5wtaSy-xh5Ys06OVl5Feh_-quKTFY-jIj10q7IjbiXz_PmYwtfP1-U

19 To be explicit, the ATT effect can be a consequence of decreased attention to treated taxpayers or increased attention to control taxpayers by the tax auditors. We are more inclined to go with the first interpretation due to reasons outlined in endnote 17.

20 Because this is a two-stage approach, to mitigate the measurement-error-induced bias problem for the second stage the standard errors are obtained via bootstrapping.

21 Year 2013 is not available in pre-treatment period because we use the lags of paid taxes and gross income in audit prediction model in (3).

22 In particular, we have obtained time averages of all the variables and subtracted from each year’s value of the same variable. The conventional fixed effects estimator with built-in stata routine applies somewhat different differencing.