"Beggar Thy Neighbor? Or Thyself?" On Inequality and Tax Policy Spillover

Few months ago, my colleague Rizmy wrote a piece on Investor Daily about tax competition and inequality. As you may know, government sometime lowers tax rate in order to attract investments. Lower tax rate means government imposes less tax to the rich, or gets less revenue for redistributive function. Hence, it may induce inequality.

Inequality, it could be argued, could be induced even though domestic policy does not change. Lower tax rate in neighboring country may create inequality via profit shifting in the domestic country (Baker and Murphy, 2019) or by capital inflow surges to the neighboring country which can worsen inequality (Azis and Shin, 2015).

Current version of Stata (Stata 15) can seamlessly create inverse-distance weighting matrix using shapefile, which is basically a map file used in mapping software. Stata 15 also has sp- prefix to enable spatial autoregressive model to be combined nicely with regress, ivregress, or xtregress. Basically, my life would be much easier if I got a hand on this back then.

This post is intended as personal exercise in new Stata feature to answer whether neighbor’s tax policy induce inequality.

Measuring Tax Policy

Our parameter of interest is tax policy, but which one? Headline corporate income tax rate is traditionally used in tax policy research notwithstanding its shortcoming of being too uni-dimensional. Keller and Schanz (2013)'s Tax Attractiveness Index is promising, as it takes into account the existence of important regulation such as transfer pricing and controlled foreign corporation (CFC) rules. Heritage Foundation's "Fiscal Freedom" is less complex than Tax Attractiveness Index, but it takes into account total tax burden to GDP ratio as measurement of the overarching effectiveness of fiscal policy to tax. In here, we use all of them as well as headline personal tax income tax rate as addition, following Duncan and Gerrish (2014).

As you can see above, Asia-Pacific countries on average are becoming less progressive. Fiscal Freedom scores show increasing trend (more freedom = less taxed), while corporate income tax rates are decreasing. Tax Attractiveness Index and personal tax rates, however, are more or less stable.

Measuring Inequality

We use Gini index as computed by Solt (2019)'s Standardized World Income Inequality Database as a basis. Since sp- demands balanced panel data, we complete missing observations using data from UNU-WIDER's World Income Inequality Database. If there are still missing observation, data from World Bank and/or domestic statistics are added.

Plotting Gini to our tax policy measure shows the following relationship:

As is quite expected, inequality is inversely related to the progressiveness of tax policy. The more "free" or "attractive" tax policy, the higher is Gini index (and thus inequality). The reverse happens with tax rate. Whether inequality is more affected by domestic tax policy, or neighbor's, or both, are what we're trying to test here. 

Model

We slightly modify the model in Martinez-Vazquez, et al. (2012) by including neighbor's tax policy measure (headline corporate income tax rate, Tax Attractiveness Index, Fiscal Freedom, and headline personal income tax rate) instead of lagged Gini index. Population growth, percentage of young population age 0-15 to total population, percentage of old population age > 65 to total population, GDP growth, GDP per capita growth, and unemployment are used as control variables.

Results

None of the control variables are significant, except for GDP growth. So the full regression result is omitted for the sake of brevity.

Now, the question whether spillover exists:

Neighbor's tax policy indeed affects domestic inequality though not in the way we expected. Firstly, such in the case of Fiscal Freedom (FISCALFREEDOM) and corporate income tax rate (CITR), it turns out that foreign tax policy is more strongly correlate to inequality than domestic tax policy.

Secondly, it completely reversed what we can infer from the scatterplots above. For example, in the case of Tax Attractiveness Index (TAXATTRACT), the coefficient is negative. This means that the more attractive our neighbor, the less unequal is our domestic economy. In the case of corporate income tax rate (CITR), the lower our neighbor tax rate, the less unequal is our domestic.

This is still consistent with Azis and Shin (2015). If what capital inflow surges increase inequality, and if lower tax rate increase capital inflow in a country, that country's Gini index rises. For example, if Hong Kong lower its tax, capital would flow there and Hong Kong citizen become less equal. Yet this warrants further study, because this one does not test if indeed capital is mobile. As an aside disclaimer: some specification error/bias may exist that I am unaware of.


Conclusion

This post is just an exercise in STATA 15 new feature. The result, in a counterintuitive way: it appears that domestic tax rate does not correlate with GINI, but neighbors tax rates correlate negatively with GINI. That means when the neighbors lower their taxes, it reduce inequality at home insofar as domestic tax rate stays the same. Does this mean because capital flows abroad then inequality is reduced at home? Does this mean that we can lower our tax rate, but still keeping it higher than our neighbors' so as not to increase inequality? Probably. This exercise is not designed to answer that question, nor can it examine the dynamics of GINI (including the persistence of tax rate effect to GINI).

Take this with an ocean worth of salt.

References:

Azis, I. J. and Shin, H. S. (2015) Capital Flows and Income Distribution. In: Managing Elevated Risk. Springer, Singapore

Baker, A. and Murphy, R. (2019) The Political Economy of ‘Tax Spillover’: A New Multilateral Framework. Glob Policy, 10: 178-192. doi:10.1111/1758-5899.12655

Duncan, D. and Gerrish, E. (2014) "Personal Income Tax Mimicry: Evidence from International Panel Data." International Tax and Public Finance, Springer; International Institute of Public Finance, vol. 21(1), pages 119-152

Keller, S. and Schanz, D. (2013) Measuring Tax Attractiveness across Countries. arqus-Working Paper No. 143

Martinez-Vazquez, J., Moreno-Dodson, B. and Vulovic, V. (2012) The Impact of Tax and Expenditure Policies on Income Distribution: Evidence from a Large Panel of Countries. Andrew Young School of Policy Studies Research Paper Series No. 12-30

Novastria, R. O. (2019). "Kompetisi Pajak Picu Kesenjangan". Investor Daily https://investor.id/archive/kompetisi-pajak-picu-kesenjangan (accessed 10 June 2019)

Solt, F (2019) Measuring Income Inequality Across Countries and Over Time: The Standardized World Income Inequality Database. SWIID Version 8.1, May 2019.

After the Leaks

This will be a short post.

I was inspired by I Wayan Agus Eka’s presentation at Seminar Perpajakan Nasional 2019 organized by DJP. Basically, he tested whether illicit financial flow to a country is affected by its financial secrecy, which is proxied by Financial Secrecy Index (FSI).

I was intrigued to know whether a disincentive towards financial secrecy affects illicit financial flow. For example, an expose that consequently put a pressure to secrecy jurisdiction, a la SwissLeaks, LuxLeaks, Offshore Leaks, Panama Papers, Bahamas Leaks, and Paradise Papers.

My theory, if leaks were to happen, countries affected will put some efforts in preventing money to flow from them to secrecy jurisdiction, or otherwise, pressure the secrecy jurisdiction to be more open. As an illustration, SwissLeaks implicated Swiss as destination of shady financial transactions. By making this information public, either government of other country enacts a measure to stop money to flow to Swiss, or people would stop transferring money to Swiss out of a reputational cost/fear from investigation, or that Swiss will make their country more “open” by changing its regulation.

To sum up in hypotheses, I want to know whether leaks correlate to downward pressure to countries that are implicated by the leaks. Such downward pressure may be proxied with lower illicit financial flow.

When I gather the data, however, I ran into some problems.

The latest data of illicit financial flow is 2015. I tried move to FSI as a proxy. Unfortunately, FSI data is also published bienally and the latest data is only available for 2015 as well. More unfortunately, in 2015, only SwissLeaks, LuxLeaks, Offshore Leaks can be taken into account (since they all happened in 2013-2014, unlike Panama Papers which was announced by ICIJ in 2016). Out of these three, only Offshore Leaks can be further used as variable since SwissLeaks and LuxLeaks are tied to Swiss and Luxembourg and neither of them are in Global Financial Integrity data of illicit financial flow. In the end, I get only 45 jurisdictions, year 2013 and 2015 (n = 90).

The model

I first tried to employ difference-in-differences (DID) estimation. But the pre-estimation test suggests violation of parallel trend assumption. Therefore DID cannot be used and I had to do regular regression. Consequently, the result here should be taken as correlational, not causative.

The model below is taken from my own estimation. Basically the first (to my horribly limited knowledge) that estimate such thing. (But I could indeed be mistaken). Anyway, the model:

, where i denotes country, t denotes time, log(IFFIN) is logged illicit financial inflow to country i in time t, OFFLEAKSIN is dichotomous variable equal to 0 if that country is not implicated as destination jurisdiction in the leaks, 1 if it is implicated; TAXRATE denotes statutory/headline corporate tax rate; GDPGROW is GDP growth; log(GDPPCAP) is logged GDP per capita; θ is country fixed effect; and ε is error term.

The predicted sign of OFFLEAKSIN is negative, since we hypothesize that if leaks = 1 it correlates to lower inflow of illicit money. TAXRATE is a control variable whose sign should be negative as well, as there can be theorized that there are tax-motivated illicit financial flow to jurisdiction with low tax rate (Hearson, 2014). GDPGROW and logGDPCAP are control variables that denote inflows that are affected by domestic economy (GFI, 2015). Reuter (2012) also observe that 5 out of 7 countries with high illicit financial flows have high GDP per capita. So I assume the signs must be positive.

As robustness check, I switch illicit financial inflow with Financial Secrecy Index Score. If Financial Secrecy Index Score is lower for countries implicated in the leaks, it could signify a downward pressure. See the above reasoning. The model for robustness check:

, where FSI is Financial Secrecy Index Score and the rests are similar to the previous.

The result

In our model, OFFLEAKSIN is statistically significant with correctly predicted sign. Meaning, there may be a correlation between leaks and downward pressure to countries implicated by them. In this case, illicit money inflow to a country is negatively correlated with whether that country is mentioned in the leaks as destination of illicit flow.

If the result is hard to read, let me put in a graph:

As you can see, pre-leaks illicit financial flow is quite similar amongst all country. However, after the leaks, flow to countries implicated in the leaks is significantly reduced (the blue curve shift to the left).

Our robustness check result is as follows:

In the second model, OFFLEAKSIN is statistically significant with correctly predicted sign. Financial Secrecy Index of a country is negatively correlated with whether that country is mentioned in the leaks as destination of illicit flow. It could be surmised that some form of pressure to be more “open” did indeed exist.

Put in a graph for easier read:

There is a clear shift to the left, but not quite as pronounced as previous graph. (This is to be expected as the second model statistically significant only at 3%, compared to the first model.)

Conclusion

Leaks are good to put pressure to secrecy jurisdiction. Thank God for the whistleblowers and ICIJ.

Reference

Eka, I. W. A. (2019). Does Financial Secrecy Affect Profit Shifting? Presented on Seminar Perpajakan Nasional 2019 (forthcoming)

Global Financial Integrity (GFI). (2015). Illicit Financial Flows: The Most Damaging Economic Condition Facing the Developing World. Global Financial Integrity

Hearson, M. (2014). "Tax-motivated illicit financial flows: A guide for development practitioners." U4 Issue January 2014 no. 2. U4 Anti-Corruption Resource Centre

Reuter. P. (2012). Policy and Research Implications of Illicit Flows, in "Draining development? : Controlling Flows of Illicit Funds from Developing Countries" by P. Reuter (ed.) The World Bank: Washington DC


Data Source:
Illicit Financial Flow is from Global Financial Integrity
Countries implicated in leaked is manually gathered from International Consortium of Investigative Journalists database
Statutory Tax Rate is from KPMG
GDP growth and per capita are from IMF WEO
Financial Secrecy Index Score is from Tax Justice Network

A Time Series Analysis of Indonesian Post-Amnesty Tax Revenues

2016-2017 Indonesian tax amnesty has been an unprecedented success. There were 973,426 taxpayers participating. 61,100 new taxpayers were registering just before and during amnesty, which comprised around 6% of total participating taxpayers. 1,030,014 million asset declaration letters were reported, with 4,884.26 trillion rupiah of declared assets. Out of these, 114.54 trillion rupiah was paid as redemption money, as well as 146.70 trillion rupiah was repatriated back to Indonesia. (Directorate General of Taxes/DGT, 2017). It is, quite simply, the most successful amnesty program in terms of money to date.

However, many have noted that tax amnesty also has some risks. It may instigate public opinion of similar forgiveness in the future (Leonard and Zeckhauser, 1987). It may also reduce government credibility (Stella, 1991) as people saw the weakness and lack of authority from the government to properly collect taxes (Uchitelle, 1989). While amnesty can generate short term revenue, the aforementioned risks make the impact of amnesty to future tax revenues remain uncertain (Marchese, 2014), or outright detrimental (Das Gupta and Mookherje, 1995).

Gauging the Post-Amnesty Effect

As is known, tax amnesty granted several facilities to participating taxpayers. Article 11 para. (5) of Tax Amnesty Law stipulates the annulment of tax arrears, cancellation of administrative penalty, as well as termination of audit for participating taxpayers. While the number of terminated audits were quite numerous (21,177 audit notice were terminated, 12,365 instruction/assignment were also terminated), the amount of settlement of prior notice of tax assessment reached around 74.3% (19.4 trillion rupiah was paid to settle 26.1 trillion rupiah in unpaid taxes including the administrative penalties) (DGT, 2017). Granted, those audits may result in bigger amounts in tax collected were there no tax amnesty. Hence, ex-post testing of tax amnesty's probable counterfactual effect remains difficult.*

On the other hand, the filling compliance ratio in 2017 reached 70.98%, which is an increase from 61% in 2016 (DGT, 2017). Those who report via electronic filing (e-filing) also increased from 59% in 2016 to 70% in 2017 (DGT, 2018). Among tax amnesty participants, 870,000 tax returns were filed in subsequent fiscal year, which correspond to 89.4% of filling compliance ratio –the highest compliance rate of overall taxpayer (DGT, 2017). In terms of money, tax revenues realization grew 4.07% in 2017 compared to previous year (13.75% if non-routine revenues from amnesty is excluded). It further grew to 14.33% in 2018 (15.53% if amnesty is excluded), which is the best tax revenue realization in the last 5 years and the first double-digit growth in the last 7 years (Kemenkeu, 2019). Given this cursory view of the data, it may seem that Indonesian tax amnesty does not significantly impact post-amnesty compliance.

Alm and Beck (1993) had once tried to measure the impact of amnesty to post-amnesty tax revenues. They tested whether the 1985 Colorado tax amnesty affected subsequent years. Similar methodology is employed here.

The Models

Three models are used based on Alm and Beck (1993). First, a simple time trend analysis testing based on OLS in the form of

Where, Yt represents monthly net tax revenue collected by DGT, T is the numeric representation of the month, e represents the error term, and b0 and b are parameters. Additional specifications for a separate intercept or slope change from the amnesty based on periods within and post-amnesty is coded A and its interaction with time is coded A*T. Tests are separated in three ways: full period from January 2014 to March 2019, pre-amnesty periods from January 2014 to June 2016, and post-amnesty periods from July 2016 to March 2019.

Second model employs Autoregressive Integrated Moving Average (ARIMA) by Box and Jenkins (1976) which tests whether the process that generated the tax revenues before amnesty was the same as the process that generated the tax revenues after the amnesty. Additionally, the identified ARIMA processes from before and after the amnesty maybe used to “fit” and “forecast” the tax revenues.

I employ Augmented Dickey Fuller and DFGLS test to find the integration of tax revenues in full period, pre-amnesty period and post-amnesty periods. All three of them are not stationary at level and stationary at first-difference. So first differenced data will be used. Further ACF-PACF and Godfrey tests suggest the existence of autoregression of order 1 and no moving average. So the ARIMA specification will be (1,1,0) instead of (2,1,0) as used in Alm and Beck (1993).

Thirdly, Multivariate ARIMA (MARIMA) intervention analysis is applied. In this case, we introduce three type of intervention assuming that tax amnesty is a discrete “intervention” thus can be represented as an additive effect of the amnesty on revenues (Box and Tiao, 1975). This requires the specification both of a starting point for the intervention and of the shape of the intervention impact. Similarly, ARIMA specification of (1,1,0) will be used here.

The starting point for the intervention (or the amnesty) is simply the time at which the amnesty occurs, which in our case is June 2016. The shape may be modeled by a “step” function with zero values up to the point of the intervention and one for all periods following the intervention, by a “pulse” function where the intervention occurs at one period and the intervention variable has just one nonzero value, or by a “ramp” function in which the step is spread over some period as a ramp response. To sum up, the shapes of intervention may be crudely depicted here:

Results

Turns out amnesty does not significantly affect post-amnesty revenue. A and/or A*T are not statistically significant, as well as partial “chopping” of time in January 2014 – June 2016 and July 2016 – March 2019.

In our robustness check using ARIMA below,

it could be construed that the autoregressive process that generate tax revenues pre- and post-amnesty are quite similar, in terms of sign and coefficient. Indeed, confirming this assumption using Chow breakpoint test and forecast test at July 2016 (when amnesty started) both result in no difference of pre- and post-amnesty revenue (p-values of 0.6776 and 0.2546, respectively).

Lastly, using MARIMA:

either pulse, step, or ramp intervention does not result in statistically significant outcome.

If I were to predict post-amnesty tax revenues from pre-amnesty data, and compare the predicted outcome and actual outcome, this is the result:

You may see that the actual value is actually higher than the predicted value, which confirms the remarkable growth of 2017 and 2018 tax revenues.

Conclusion

It seems that the 2016-2017 tax amnesty in Indonesia does not lower the tax revenues of subsequent years. Yay!

Reference:

Directorate General of Taxes (DGT). (2017). Tax Amnesty Mosaic

DGT (2018). Kepatuhan dan Penerimaan Pajak 2017 Tumbuh Pesat, DJP Optimis Hadapi 2018. http://www.pajak.go.id/kepatuhan-dan-penerimaan-pajak-2017-tumbuh-pesat-djp-optimis-hadapi-2018 (Accessed March 20, 2019)

Kementerian Keuangan. (2019). APBN Kita, January 2019 edition

Alm, J. Beck, W. (1993). Tax Amnesties and Compliance in the Long Run: A Time Series Analysis. National Tax Journal, Vol. 46. Pp. 53-60

Box. G. E. and Jenkins, G. M. (1976). Time Series Analysis, Forecasting, and Control. San Francisco Holden-Day

Box. G. E. and Tiao, G. C. (1975). “Intervention Analysis with Application to Economics and Environmental Problems”. Journal of the American Statistical Association 70 (March 1975) pp. 70-79

Das-Gupta, A. and Mookherjee, D. (1995). Tax Amnesties in India: An Empirical Evaluation. IED Discussion Paper Series, 53. Boston University

Leonhard, D. Zeckhauser, R. J. (1987) “Amnesty, Enforcement, and Tax Policy”. NBER Working Paper No. 2096

Luitel, H. S. dan Sobel, R. (2007). The Revenue Impact of Repeated Tax Amnesties”. Public Budgeting and Finance 27 (Fall 2007): 19-38.

Marchesse, C. (2014). Tax Amnesties. Papers in Comparative Analysis of Institutions, Economics, and Law No. 17.

Stella, P. (1991). An Economic Analysis of Tax Amnesties. Journal of Public Economics 46 (3): 383–400.

Uchitelle, E. (1989). Amnesty Programs in Selected Countries. FRBNY Quarterly Review

Monthly tax revenues data is obtained from internal data. Net total tax revenue is used.

------

*It is possible, given microdata of corporate and self-employed taxpayers which either participated or did not participate in tax amnesty, then conducting a simple ANOVA, ANCOVA, or difference-in-differences analysis in a quasi-experimental research setting.

歩いても歩いても

They say 27 is the age when rockstars often die. We are not rockstars, although life has probably fucked us harder than any drugs imaginable. Most importantly, we aren't died yet. I think we are still blessed in that matter.

Getting older has its bane. These days, I forget easily. There are much to think, but there are less time to commit everything to memory. And in the process I threw away what is essential and kept what is irrelevant instead. Long gone are those times when we stop and ponder and marvel at the banality of our lives. But I am trying.

I know that in the eve of your 27th birthday you wanted to celebrate it in your own way – in solitude, trying to recover some quiet moments. So this time, in your absence, I try to mythologize all the strange roads that we once walked – supplanting the half-remembered nostalgia with an allegory of my own making.

I remember the walk we walked on one night on that alley where we found a kitten, lost and alone in the darkness of the bushes. You told be to throw a cat food that I bring in my pocket. You were worried despite my insistence that it would be fine.

Sometimes I think of us as the kitten. Astray, wailing for its mom. We think it won't survive. But help did come. It survived. The kitten turned out strong despite its apparent helplessness. And so as it survived, we survive.

You survive.

Because you are strong.

Sometimes stubbornly so.

I also remember that time when you once walked for a couple miles, under the scorching sun and the heat waves. Few hours later you were still able to dance it all away, with multitudes of strangers, on the dancefloor in the night market.

I once likened you to the bristlecone pine. A tree that thrives under the cold, harsh gale, yet lives longer than many civilizations. I still think that you are.

A bristlecone pine won't die at 27. 27 is the age when rockstars often die. I think I kinda know why. 27 is the forking path where you can no longer using youthfulness as an excuse. It's a prelude for a full-fledged, well-functioning adulthood. Things are getting more pressing. Societies are more demanding. New expectations are coming. The road that we walk will be harder, perhaps even more punishing. Some people try to resist this, and they crash themselves so hard they destroy themselves in the process.

But I know you are strong, strong enough to walk any road.

That's why I want to walk with you, to watch your figure from behind.

It does not have to be in a faraway land, or places that we dream to walk. I don't care if it's Coburg, or Shimogyo-Ku, or Ubud, or Santorini, or Setiabudi, or Gang Gloria, or Pasar Klewer. I don't care if blisters and cramps come as they may. As long as I walk with you. As long as I walk with you forever. As long as you are happy walking with me.

So, happy 27th birthday, N. May you always be happy on your every journey. I promise I will remember this.

The world gets older, N, and so must we. But if the world gets colder, we must not.

How Much Is the Loss from Tax Avoidance: Revisited

Interesting new paper by Tørsløv, Wier, and Zucman (2018) that came out few days ago intrigued me to revisit the question about missing profits. As you may know, corporation is driven by profits. Taxes cut the profits. So it is economically rational to try to reduce tax. One of the usual modus operandi is to establish subsidiary in low tax jurisdiction and divert the profit to it.

In previous post, I have tried to estimate this loss in macroeconomic setting. Now depending on the condition (whether companies also consider market size or proximity) Indonesia may gain 2% or loss up to 5% of GDP due to tax base spillover. In this post, inspired by an equation mentioned by Tørsløv, et al. I will try to use a much more granular data.

The Setup

Numbers of research concerning profit shifting used commercial database to extract the relevant information from the companies' financial report (mainly from income statement and balance-sheet). These are then entered into some equation. One of the equation mentioned in Tørsløv, et al. is:

where log(πic) denotes natural logarithm of pre-tax profits booked by company i in country c; τp is the tax rate in the parent company; τc is the domestic tax rate; and Firm and Country denote firm- and country-specific fixed effects.

I managed to collect the data for 904 companies for the years of 2012-2017. However, I have to modify the above equation in some ways. First, since I only concern about what happen in Indonesia, I have no use for Country fixed effect. Second, I change the tax differential to include all the subsidiaries in the multinational entities group, instead of the simple parent-subsidiary relationship. The reason is that an Indonesian company (even though it is a subsidiary, for instance) does not necessarily shift the profit to its parent. The profit can be shifted to another subsidiary within its MNE group that is a resident in low tax jurisdiction. The profit can also be "pooled" somewhere and remains unrepatriated to the parent's jurisdiction as in the case of many US companies.

Further, since the logarithmic operation can only be applied to positive numbers, then I must exclude some companies in some years because they booked 0 or loss (negative profit). This makes my panel data unbalanced and the observations are reduced. In the end only 695 companies remain with N = 3,113.

Because of the exclusion too firms may enter and leave the time series. For example, company A booked profit in 2012, 2013, 2015, and 2017 but had zero/negative profit in 2014 and 2016. Due to the log operation, company A is included in the sample for year 2012, 2013, 2015, and 2017, but it is excluded in 2014 and 2016. As a consequence, I cannot apply firm fixed effect and the model essentially turns into random-effect model. This is also confirmed by the results of Hausman and Lagrange Multiplier tests.

That being said, the model turns into:

where log(πi) denotes natural logarithm of pre-tax profits booked by company i in Indonesia; τavg is the unweighted average tax rate of the MNE group following Johansson et al. (2017); τc is the Indonesian corporate tax rate (25%); the rests are errors.

From a theoretical standpoint, choosing only companies that booked profits still has a benefit. You won't have to pay tax if you suffer loss or you have zero profit, so there is not much use to further avoid tax. Moreover, this allows the model to test whether agency theory is at play here: whether the manager tries to reduce tax without sacrificing much of the shareholder values by having zero profit or even booking loss.

The Result

Difference in tax rate negatively correlated with pre-tax profitability (p-value 0.033). If the tax rate of Indonesia is higher than the average tax rate of the MNE group, the pre-tax profit of Indonesian company belonging to said group becomes lower. For every percentage of tax rate difference, the pre-tax profit of the company drops by 4.37%. This means an Indonesian company whose other MNE group members are located in Singapore (tax rate 17%) and Vietnam (tax rate 20%) will have 18.93% less pre-tax profit than a non-MNE Indonesian company.

Extending the result to include all the companies in the sample, this is how much pre-tax profit is loss on average:

Conclusion

Again, we see evidence of tax avoidance in the samples included in this study. Of course this is not a comprehensive view of tax avoidance behavior. Loss-making companies and those who have zero profit are not covered here, although it is very likely that they engage in tax avoidance behaviors (via transfer pricing or excessive interest expense, for instance).

There are also other, more comprehensive methods to detect profit shifting explored in Tørsløv, Wier, and Zucman (2018). I suggest you read their paper if you want to know about the missing profit of nations. Indonesia, sadly, is not covered there.

References

Johansson, A, Ø. B. Skeie, S. Sorbe, and C. Menon (2017), “Tax Planning by Multinational Firms: Firm-Level Evidence from a Cross-Country Database”, OECD Economics Department working paper 1355.

Tørsløv, T, L. Wier, and G. Zucman (2018), “The Missing Profits of Nations”, NBER Working Paper 24701.

How Much Is the Ｌｏｓｓfrom Tax Avoidance?

This is a short post based on the research by Cobham and Janský (2018) which continues from Crevelli, De Mooij, and Keen (2015).

Here's the context: many countries in the world suffer from loss of tax revenues due to tax avoidance and/or tax evasion. One of the most common methods used by multinational corporations is to set up a subsidiary in the so-called tax havens non-cooperative or low-tax jurisdictions to divert their profits.

Having a subsidiary in non-cooperative/low-tax jurisdictions is not necessarily indicative of tax avoidance. Businesses may have legitimate reason such as access to market, diversification, or centralization of one of their business functions. Nevertheless, the concern of tax revenue loss cannot simply be ignored.

But how much is lost?

The study by Crevelli et al. (2015) tries to estimate how much a country loses their tax revenue from tax avoidance. Cobham and Janský (2018) tests and re-estimates this research. I'd like to try estimating the loss of Indonesia.

The Setup
A company avoid taxes in a country by diverting profit to its subsidiaries in other countries. The tax base of the aforementioned country is thus reduced and "spillover" to other countries.

But how does a company choose the jurisdiction of its subsidiaries?

In a micro-data, such as commercial database (that contains financial reports and jurisdiction of subsidiaries/headquarter), this is much easier and clearer to measure. In fact, many of the measures and monitoring tools for base erosion and profit shifting as outlined in OECD's BEPS Action Plan 11 utilize micro-data from commercial database.

Measuring in a macro (country-level) setting thus requires us to make assumptions. Since a company may create a subsidiary for access to market, we must assume that company does not pick a country purely for its low tax. A company may be hesitant to create a subsidiary in a small or faraway country. Or, given a choice of two countries with similar tax rate, a company will choose the country which is closer or has bigger market than the other one. Thus, we weight the effect of tax rate in every country of the world based on two things: distance and market size (that is represented by GDP).

The Model
We first estimate the φ and λ based on the equation:

, where bit denotes the corporate tax base in country i in time t; τit the domestic tax rate (in this case Indonesia); W-it τit a weighted average of the tax rates in countries other than Indonesia; Xit a vector of controls such as trade openness (import + export divided by GDP), share of agriculture to GDP, and log of GDP per capita in constant 2011 dollar; and μt is time specific effect. I employ LASSO (Least Absolute Shrinkage and Selection Operator) in the estimation to penalize model overfitting, as I don't have much data to begin with.

After we get φ and λ we plug them to equation:

for short term effect (L), and

for long term effect (LL), where Whτ-it denotes the average tax rates of tax haven countries as listed by Gravelle (2013).

The Results
The loss of Indonesia's tax revenue as % of GDP in year 2008-2017 is as follows:

Interestingly, this suggests that if companies actually consider GDP as the main factor to create subsidiary in a country, there is a tax base spillover to Indonesia. Meaning that Indonesia actually gains from tax avoidance.

After I examine the data, it turns out that Indonesia's tax rate is lower than the rest of the world if weighted by GDP. This is understandable as small GDP countries with zero tax rates such as British Virgin Island matters less in this scenario than, for instance, Germany (whose tax rate approximately 30% is higher than Indonesia and whose GDP is much bigger than Indonesia as well.)

On the other hand, if a company seeks to establish a subsidiary in low tax jurisdictions near Indonesia (i.e. assuming it also wants to exploit agglomeration effect or to shorten supply chains), then Indonesia loses tax base approximately 3-5% of GDP.

Translating the above result into Rupiah, this is how much Indonesia loses (gains) the tax revenue:

This is a very crude estimation due to severe data limitation. Maybe I will update this if I have the time to collect more and better data. Maybe.

References:

Crivelli E, De Mooij R, Keen M. 2016. Base erosion, profit shifting and developing countries. FinanzArchiv: Public Finance Analysis 72(3): 268–301. https://doi.org/10.1628/001522116X14646834385460

Cobham A, Janský P. 2018. Global Distribution of Revenue Loss from Corporate Tax Avoidance: Re-estimation and Country Results. Journal of International Development. UNU-WIDER.

Gravelle J G. 2013. Tax havens: international tax avoidance and evasion. Washington, DC. http://fas.org/sgp/crs/misc/R40623.pdf. Accessed 25 June 2018

OECD. 2015. Measuring and Monitoring BEPS, Action 11 - 2015 Final Report. Paris: Organisation for Economic Co-operation and Development.

Data is from World Development Indicator of World Bank; World Economics Outlook and International Finance Statistics of IMF; tax rate data is from KPMG Tax Rate Table, with some additional research for small jurisdictions; Badan Pusat Statistik; and APBN of Indonesia.

Inequality and Tax Evasion in Indonesia

Do rich people evade more taxes, compared to the less wealthy? This pose a problem for some scholars (and government). If the richest understated their income in their tax records – and disproportionately so compared to other taxpayers – the study of inequality which often based on tax records may be inaccurate. Inequality may actually be worse than what appears in Gini ratio.

In answering this question, Annette Alstadsæter, Niels Johannesen, and Gabriel Zucman (2017) conduct a detailed study by using data from Scandinavian countries. Scandinavians are arguably one of the places where the data quality is excellent. They are able to collect data from randomized audit results, provided by Scandinavian tax authorities (because the transparency policy there also grants easier access).

Despite the trove of quality data, they still theorize that audit can potentially fail to capture sophisticated tax evasion scheme done by the rich. So it is still possible that the resulting income in the tax assessment is understated. Thus, they combine the audit data with leaked data from offshore financial institutions (Swiss Leaks and Panama Papers). They also supplement it with information from recent tax amnesties (again, provided by Scandinavian tax authorities).

Based on Swiss Leaks, Panama Papers, and amnesties, Alstadsæter, et al. found that the top 0.01% richest households evade about 25% of the taxes they owe by concealing assets and investment income abroad. Adding the result from the tax evasion detected in random audits, the total evasion in the top 0.01% reaches 25-30%, versus 3% on average in the population.

The State of Inequality in Indonesia
In 2017, Indonesian Statistic Agency (Badan Pusat Statistik/BPS) reported that the Indonesia's Gini index is 39.1.  But you've probably heard/read the Oxfam-INFID paper in 2017. Combining the data from Credit Suisse Global Wealth Databook 2016 and the list of richest people in Forbes, the paper shows shows how the 4 richest Indonesians have more wealth ($25bn) than the combined wealth of 100 million Indonesians in the bottom 40% ($24bn). The same report also suggest that Indonesia is the 6th most unequal countries in the world, after Russia, Denmark, India, United States, and Thailand.

By using the similar data from Credit Suisse Global Wealth Databook (year 2017), I found out that the inequality in Indonesia is worse than official Gini index calculated by BPS.

74.8% of Indonesian wealth is owned by the top 10% richest. If we "zoom" it in further, 60% of wealth owned by the top 10% population is owned only by the top 1% richest. In other words, the top 1% richest owned almost half of total Indonesian wealth. 45.4% of total wealth to be precise.

If this distribution is correct, then the Gini ratio should be twice as worse at 83.7, instead of 39.1 as reported by Badan Pusat Statistik.*)

If we were to measure the average wealth of each population decile, the inequality is more pronounced.

The top 10% richest are, on average, 7.31 times more wealthy than the average Indonesian. Meanwhile the top 1% richest are, on average, 44 times more wealthy than average Indonesian. We can go further: top 20% are 287 times richer than the bottom 20%; top 10% and top 1% are 748 and 4,540 times richer than the bottom 10%, respectively. Look at how the average wealth of bottom 10% population did not even show up in the graph above. This is indeed staggering.

I compare the analysis by looking at the distribution of saving accounts in Indonesia. In this regard, saving accounts could be construed as a proxy of wealth (at least, financial wealth), similar to the HSBC data contained in the Swiss leaks used by Alstadsæter, et al. Indonesian Deposit Insurance Agency (Lembaga Penjamin Simpanan/LPS) keeps track of account ownership that are subsequently grouped based on the amount of money in those saving accounts. This is the source of the saving data.

In the LPS data, the state of inequality in Indonesia still looks astoundingly bleak. For instance, in 2017, 98.06% of saving accounts in Indonesia contain less than 100 million Rupiah in amount. But those 98.06% cumulatively only own 14.6% of total deposit in Indonesia.

In contrast, only 0.04% of saving accounts that have more than 5 billion Rupiah in it. But these 0.04% cumulatively own 46.17% of total savings.

To make it easier to imagine: if we scale down Indonesians into 10,000 people, then the richest 4 persons have 3 times more money in their bank accounts than the other 9,800 combined.

I then calculate the average amount in each saving group. Below is the result:

Note how similar this is to the distribution based on Credit Suisse data. The average amount in the lowest group barely show up in the graph at all, despite accounting for 98% of account ownership in Indonesia. The top group (account with > 5 billion Rupiah) on average contains 2,268 times more money than the average Indonesian saving account. This is an obscene picture of inequality.

Tax Evasion

Now we go back to the original question: do rich people evade more taxes? If I had a hand on the data as good as Alstadsæter, et al., combined with a more detailed distribution of wealth in Indonesia, I can get a more precise estimate.

In the case of Indonesian data, however, I face several roadblocks.

First, the distributional data in the World Inequality Database for Indonesia is inadequate. It does not have a detailed distribution for each percentile/decile. That being said, I have to make do with the Credit Suisse data. However, Credit Suisse does not report the range in each decile nor the standard deviations. I have to reconstruct the distribution points based on the assumption that the Indonesian wealth has uniform distribution, so it is efficient to compute its L-statistics to find out the range in each decile.

Second, because of the obvious confidentiality reason, I don't have the full amnesty data, only its statistics. Since the statistics are anonymized and aggregated (meaning I don't know who owns what and how much), I cannot cross them with the names in Swiss Leaks or Panama Papers. Similarly, I also do not have the data from tax audit to cross with the amnesty/leaked data.**)

Despite such limitations, I try my best to follow the method as outlined by Alstadsæter, et al. in their appendix. This is the finding of Alstadsæter, et al. in Scandinavian countries:

Using the available statistics and reconstructed distribution points, this is what I find in Indonesia:

In both cases, we can see that the probability of the rich to use tax amnesty is high, and the richer you are the more likely you use tax amnesty. Nonetheless, in Scandinavia, the probability for 90-95 percentile is quite close to 0%, while the probability for Indonesians in comparable wealth band (91-93 percentile) is twice that number. Admittedly, even the intragroup probability in the 90-100 percentile (top 10% richest) of Indonesia is in stark contrast compared to Scandinavia. You can see that in Indonesia it is skewed exponentially (i.e. increase very quickly) to the very richest compared to the much more steady, linear-like of Scandinavia.

If we consider tax amnesty as a sign of the previously undetected tax-related wrongdoings – be it evading, avoiding, or simply filing the tax return not in correct, complete, and clear manner – then the richest Indonesians disproportionately commit more wrongdoings compared to the general population. The top 1% richest of Indonesia is 16 more likely to use amnesty compared to average Indonesian (and 201 times more likely compared to the bottom 10%). This is after considering the fact that tax amnesty is basically open to all, and that there is no special treatment for the rich.

Unfortunately, this is the furthest I can do with the data. Given better access, it may be possible to gauge the post-amnesty compliance as outlined by Alstadsæter, et al., or by using macrodata similar to James Alm's study in post-amnesty Colorado. It is also interesting to combine audit data, leaked data, and amnesty data to paint a complete picture of tax evasion and inequality in Indonesia. But, alas.

In the end, it's likely that rich Indonesians have the means and opportunities to concoct tax-evading schemes. And being richer granted even more means and opportunities. "In the little game of tax demagogy," Piketty once wrote, "the weakest seldom come out winners."

Reference:

Alstadsæter, A, N Johannesen and G Zucman (2017), “Tax evasion and inequality”, NBER Working Paper 23772. Appendix available at http://gabriel-zucman.eu/leaks/

Alstadsæter, A, N Johannesen and G Zucman (2018), “Who owns the wealth in tax havens? Macro evidence and implications for global inequality”, Journal of Public Economics, forthcoming.

Piketty, T (2016), "Chronicles: On Our Troubled Times", Penguin

Oxfam - INFID. (2017), "Towards a More Equal Indonesia", Oxfam International

Data:
Lembaga Penjamin Simpanan - Distribusi Simpanan Bank Umum Desember 2017
Badan Pusat Statistik
World Inequality Database, available at WID.world
Direktorat Jenderal Pajak
Credit Suisse (2017) Global Wealth Databook 2017. Credit Suisse Research Institute

------------------
*) Different methodologies could result in different Gini ratios for the same country. For instance, Badan Pusat Statistik use consumption data based on household surveys, while Credit Suisse combined household surveys with SUR and upward-adjustment based on available financial wealth data. Different bracketing could also affect Gini calculation. If you rank the population in decile (1-10) or percentile (1-100) or by 40-40-20 (poorest 40%-middle 40%-richest 20%), the resulting Gini ratios can be very different. If the wealth range in each distribution point is different, Gini can also be different; etc.

**) To my knowledge, Indonesia has yet to implement nationwide randomized audit program. The national audit program for individual taxpayers is focused on prominent people, corporate owners, and entrepreneurs. The good: rich Indonesians are given more scrutiny. The downside: there is not much information regarding the tax compliance of middle-to-low income class. Even if I have the audit data, it may not be representative to infer from these focused audit targets as samples of the general Indonesian population.