I have been invited to present my research at the conference called Public Economics for Development, organized by UNU-WIDER in Maputo, Mozambique, on July 5-6, 2017. My talk regarded a work in progress that is carried out by me and Petr Janský as part of the COFFERS project. First of all, I would like to express my gratitude to UNU-WIDER for organizing such a great event, bringing together researchers from all over the world to talk about what economics can do to end poverty and jump-start sustainable development in poor countries.
Interestingly, this was my first ever visit to a country classified by the World Bank as ‘Low income’. 30 more to go!
Below is a transcript of my talk, here is my presentation in .pdf, here is some more information about the conference, here is the conference program, here are photos from the event.
Miroslav Palanský, Public Economics for Development, July 6, 2017, Maputo, Mozambique
Good afternoon, my name is Miroslav Palanský, I am a researcher at Charles University in Prague in the Czech Republic. Today I am going to present a paper that I have co-written with Petr Janský, which deals with a topic that has become pretty mainstream in the last couple of years, and that is tax evasion of multinational corporations. You can always know a topic is popular when there are lots of picture jokes about it on blogs and social networks. In this one, Apple plays the role of the firm that is being made fun of because of the shape of their logo, but schemes employed by Apple are widely used by most major multinational companies to abuse the tax system and thus reduce government revenue.
So what is corporate profit shifting? In its general sense, what we mean by corporate profit shifting is simply the illegal movement of money, or profits, from a higher-tax jurisdiction to a lower-tax jurisdiction, with the aim, of course, to artificially reduce the tax bill of the multinational corporation that engages in these activities. There are several major categories of practices that lead to profit shifting, such as trade mispricing, debt financing or the sale of assets between member companies of a multinational corporation.
Let’s see, on a most simple example, what a relationship that exploits some or all of these channels might look like. Consider my home country, the Czech Republic, in which a Firm A operates. Also consider a Firm B which operates in Cyprus. Suppose further that Firm B owns Firm A, and thus has some foreign direct investment (FDI), vested in it, and together, Firms A and B form a multinational corporation, or a multinational group of companies. Now suppose that the group artificially shifts profit from Firm A to Firm B. Note that this is not sending dividends, but rather illegally shifting the profit before it was taxed in the Czech Republic. And why would the multinational corporation do that? Because the conditions for the taxation of firms in the Czech Republic and Cyprus are different. More specifically, they are more favorable in Cyprus – not only is the corporate income tax rate lower there, but also other conditions of running a business are more favorable, including higher financial secrecy or not imposing restrictions on minimum holding requirements and so on.
So let me sum up the motivation of our paper in three main points. The first one is that we aim to empirically test whether there even is profit shifting of multinationals. A lot of case studies suggests that there is, so the answer to this question might be anticipated easily. The second question is much more complicated to answer. It asks how much corporate profit is shifted and from which countries. We are talking about quantifying a phenomenon usually described by words such as ‘hiding’ and ‘avoiding’, so one might see the intrinsic difficulty of this task. Today I will present one of the approaches developed recently that uses data on foreign direct investment. Our third motivational question concerns the effects of the profit shifting on the tax revenue of affected countries. We try to answer how much each country loses as a result of profit shifting practices.
Let us start with the first question. As I said, in our paper we use the so-called FDI approach that has been first developed by UNCTAD in their 2015 World Investment Report. Its basic idea states that if there is profit shifting, it will show in the data as deflated reported profits in higher-tax jurisdictions. Basically, shifting profits from a country will reflect in lower profits reported in that country. Therefore, assuming there is profit shifting, if there is a lot of foreign direct investment from lower-tax jurisdictions, or tax havens, the reported profits should be lower.
So coming back to our example, we are focusing on the profit channel, and we say that as a result of this channel’s existence, the reported profits in the Czech Republic will decrease and the reported profits in Cyprus will increase.
To facilitate the method, we need two basic data sources. The first one is data on bilateral stocks of FDI. We take ours from the IMF’s Coordinated Direct Investment Survey, which contains precisely what we are looking for here – on its basis, we can construct a bilateral FDI matrix with investing countries in the rows and receiving countries in the columns. We complement this data with UNCTAD’s FDI Statistics that report the stock of inward FDI in each country. We use this mainly as a robustness check. The second source we need is data on the rate of return of the above mentioned FDI stocks. We want to show that these reported returns are lower when the share of inward FDI from tax havens is higher.
We thus define a variable that we call the offshore indicator as the share of inward FDI from “risky” countries. What this means is countries in 2 groups. The first one is composed of your classic tax havens, countries such as the Cayman Islands or Cyprus. We consider all investment from these countries as risky. The second group are the special purpose entity enabling countries. Special purpose entities are a type of firms whose main purpose is to route funds through a country. The central banks of some countries report data on how much FDI flows through these SPEs and we use this data to classify this reported share as risky investment.
So for illustration, imagine a country X with inward FDI of this size. One part of it is FDI coming from tax havens, another part is FDI coming from SPE-enabling countries, and the last part is other, non-risky FDI. The first two parts then constitute our offshore indicator.
As I said before, we assume a negative relationship between the offshore indicator and the rate of return on FDI. The whole FDI approach stands and falls with the negativity of this relationship. A first look at the data shows that even at a global scale, the negativity is present. This data thus points to a confirmation of the anticipated positive answer to the question whether profit shifting is actually a real thing.
Let’s now move to the second central question of our research, that is how much profit is actually shifted and from which countries. To do so, we use the results of a regression model that estimates the relationship I talked about until now. We perform this regression using income group- and regional-fixed effects as well as year-fixed effects and obtain significantly negative estimates for most of the examined countries. Based on these results of the regression analysis, we derive what we call the probability gap, which is the difference in the actual rate of return on FDI in each country and the rate of return that we would expect based on our model. The reasoning of the FDI approach is that this profitability gap is a result of profit shifting.
Looking at the results of the regression, we see two columns. The first is based on simple rate of return as the dependent variable while the second reports results from a similar model that, however, uses only the equity component of the rate of return, which is also included in the IMF’s data set. The final profitability gap for each country is then obtained as a sum of the coefficient for the offshore indicator and its interaction with the relevant dummies. So for the Czech Republic, for example, which is a high income OECD country in Europe, we get the coefficient as the sum of these three numbers, and for Mozambique, for example, which is a low income country in Sub-Saharan Africa, it is the sum of these three, which, coincidentally, means just the base estimate.
Finally, the missing profit in each country can be calculated as the product of the profitability gap and the amount of risky FDI in that country. For those who are not lost yet, the amount of risky FDI in US dollars is the product of the offshore indicator and the total FDI. Calculating these numbers for each country answers our second question, that is how much corporate profit is shifted and from which countries.
The third and final question that we want to answer in this paper is how much these countries lose on tax revenue as a result of this profit shifting. This is equivalent to asking how much more would countries collect in corporate tax if the missing profits were not missing but rather stayed in the country of their origin and were taxed there. The tax revenue loss can be easily calculated as the product of the missing profit in US dollars terms and the corporate tax rate. We use data on effective tax rates from KPMG and the World Bank.
So what do the results look like? The first thing we notice is the limited coverage of countries, mainly among the developing countries. This is due to missing data on FDI, mainly rate of return on FDI, in a large number of countries. We hope that this is a limitation that can be alleviated in the future by including data from additional sources as well as gradual improvement of existing sources.
This table shows the countries that are, based on our results, the biggest losers of the profit shifting business. You may notice that among these there are some of the poorest countries in our sample. Mozambique finished second, which falls well with the venue of this conference, but otherwise is rather sad, considering the high numbers of tax revenue lost.
Let’s zoom in a little bit to Europe which has good coverage. In dollar terms, the biggest losers are some of the richest countries in Europe, but if we take shares of these losses on GDP, the picture changes markedly and the burden is relatively larger on the poorer countries mainly in Eastern Europe.
To conclude, our results confirm some of the previous findings and provide further corroboration on the sizeable estimates of corporate tax abuse. We also confirm the notion that poorer countries are the ones that are hurt the most in relative terms, while having a worse position to fight back.
Regarding the FDI approach, the method has its weaknesses and they are considerable. I am also confident that I have not listed all of them and that you will be able to come up with some additional ones and I would be grateful for a feedback of this kind. I want to divide the problems of the methodology into two parts – those that we think we might be able to solve and those that we know we cannot. In the ones that we might be able to solve, we have as the first one developing these estimates into shares of corporate tax collected. This is now made possible by the new update of the GRD which includes data for 2015, so this will be an easy one thanks to the great work done at WIDER. Limited coverage is the next one. Additional data sources are emerging and the current ones are being improved, so finding and exploiting new sources is one area of potential improvement. Another one is that the method still partly relies on a dichotomous definition of tax havens that has been heavily criticized in the literature lately. We might be able to incorporate instead a variable that places all jurisdictions on a continuous spectrum rather than assigning them with a binary variable. One of these sources is the Financial Secrecy Index which we will analyze in this context in the coming months. We also want to focus on individual cases of countries to get to better understand the results. We are looking at tax treaties data and whether these might play a significant role in the estimated model.
Of those that we consider not solvable, I listed that this method only includes those profit shifting practices that require an FDI link. Our estimates can thus be considered a lower bound of what is actually happening, since many popular profit-shifting practices are carried out with no FDI link whatsoever. We also recognize the inherent weaknesses of the method, such as the issue of what fixed effects to include in the model. As a result, right now we can only say with certainty that our estimates are wrong. We should thus regard them rather as estimates of the order of magnitude. Finally, we are constrained by the decision of which tax rates to use, whether effective or nominal, because attracting more investment in a country is often connected with tax cut deals which bring the effective tax rates down.
That is all from me today, thank you for your attention and I welcome any comments or feedback you might have.