VAT experience as an influence on banking crises

1.   Introduction

All countries with Value-Added Tax (VAT), exempting financial services or not, permit a different treatment of financial sector under existing VATs, which sometimes, if not always, exacerbates the risk of the sector becoming excessively large (IMF, 2011). This has recently led to banking crises as the Global Financial Crisis (GFC), and after that economic recessions, as the Great Recession (GR). Due to this, international institutions as the IMF (2011) have advocated for applying more efficient taxes that improve the current treatment of financial services under VAT. Nonetheless and motivated by technical difficulties, there is still no country in which financial services were correctly dealt on VAT as the rest of goods and services. For instance, at New Zealand input VAT credits are allowed, but there is no positive tax rate on these services (Lopez-Laborda and Peña, 2017b). Nonetheless, there are some practical alternatives that allow the full taxation of financial services under VAT (Lopez-Laborda and Peña, 2018), which are recently taken into account in some technical reports from this country¹.

¹ A document entitled "Taxing financial services" has been prepared by the Tax Working Group Secretariat (2018) of New Zealand as request for advice on options for applying Good and Services Tax (GST, national Value Added Tax, VAT, on that country) to financial services. It studies methods of taxation, considering Lopez-Laborda and Peña (2018) proposal as a simpler way to avoid technical difficulties in the full taxation of financial services on VAT.

This paper aims to shed further light on the influence of current VATs and Goods and Services Taxes (GSTs) on the unchaining of banking crises, analysing the transmission channels and providing policy measures. A positive influence of the VAT experience on crises is observed, via inflation and via exemption of financial services on VAT. A dataset of 36 European Union (EU) and Organisation for Economic Co-operation and Development (OECD) countries for the period 1970–2011 to study the importance of VAT experience on banking crises, a relationship not empirically studied before as far as we know. Panel logit probability models with population-averaged and random effects are developed, capturing the most relevant crisis determinants in the literature.

The rest of the paper is divided into five sections. The second section contains the theoretical expectations of the influence of VAT experience on banking crises and develops the literature review of the determinants of these crises. The third section explains data and methodology, using multivariate panel data models. The empirical results are provided in the fourth section. The main result is the corroboration of the significant and robust positive impact of VAT experience on the likelihood of a banking crisis. The fifth section discusses the results and provides conclusions and final remarks.

2.   Expected influence of VAT experience on banking crises

There are two main effects of current VAT systems on banking crises. First, the widening of trade commerce by the establishment of VATs as substitutes of other taxes more distortionary and harmful for trading. Second, the recent GFC has taught economists the importance of systemic risk in these crises. An asymmetric treatment of the several productive sectors of goods and services, as occurs with the financial sector on VAT, encourages banking crises. The exemption of financial services on VAT leads to the unrecoverability of the input VAT on banking purchases, increasing the prices of the real products due to a cascading effect. This last effect, join together with an increase of the price of goods and services due to VAT implementation, could unchain a rise in inflation, which may influence in reducing consumption, credit demand, bank profits and finally, triggering a banking crisis.

The first reason is that VAT experience could increase trade commerce, and this rise leads to a higher exposure to banking crises (Rose and Spiegel, 2010, 2011). Trade commerce could be positively influenced by VAT experience because distortionary taxes that discourages trading are substituted by pro-exporting VAT. If a country turns to be trading-dependent because of the trading increase, an international financial crisis as the GFC would affect in a major sense than countries without external contact (and therefore, less abroad dependents and more self-sufficient). Many authors consider trade openness as determinant for currency crises, although some others find the opposite effect (Cavallo and Frankel, 2008). In fact, according to Haddad et al. (2013), there are other factors as export diversification conditions that could mitigate the effect of trade openness on economic growth volatility. So, the sign of this first effect is not fully clear.

The second reason is the distortion that produces the absence of a full taxation of financial services on VAT with compatibility with the credit-invoice method of standard VAT, mainly due to the complexity and difficulty in implementing a feasible taxation in a transaction-by-transaction basis (Lopez-Laborda & Peña, 2018). As mentioned before, the unrecoverability of the VAT levied to the inputs for banks could lead to banking crises. The next theoretical explanation is based on the assumption that the probability of a banking crisis ($\Phi$) negatively depends on banking profits (${\Pi _b}$).

According to Lopez-Laborda and Peña (2018), financial value added could be obtained by the addition (sum of profits and wages) and subtraction (difference between sells and purchases). The exemption of financial services on VAT leads to an additional "purchase": the unrecoverable input VAT, so separately considering value added of the main banking services, i.e. loans (l) and deposits (d):

where $w$ are the wages, $IR$ and $IP$ are the interest receipts and payments, respectively, $\varepsilon$are the public bonds hold by Banks, which reflect the "pure" interest, $IC$ are the input costs and $\tau$ is the rate of general VAT, which is related with VAT experience. The total amounts are reflected by the "b" subscript in the next equation that reflects total banking profits:

For analysing the effect of VAT experience on banking crises, the derivative of bank profits respect to the VAT rate is calculated, obtaining:

Which theoretically confirms the expectations that VAT experience positively influences on banking crises through the channel of VAT exemption of financial services. This paper aims to fulfil a hole in the banking crises literature regarding the influence of tax factors as VAT on banking crises.

Recently, several studies in the literature have aimed to explain banking crises (Peña, 2016), while the financial, domestic and credit determinants of banking crises have been well established, the literature has typically avoided the study of tax factors related with the financial institutions as influences on these crises.² Gavin & Haussmann (1996) provide theoretically a seminal study of the main factors that trigger a banking crisis, related to liberalization, banking competition, regulation of deposit interest rates, information, macroeconomic volatility, capital assets ratio, bank liquidity, monetary policy and lending booms, but there is no focus on determinants related to the tax system applied to financial entities.

² Boudriga & Ghardallou (2012) provide a good review of the literature of banking crisis determinants.

Most of the studies use binary variables reflecting the presence of a banking crises, as Nocetti (2006) shows stating the importance of the use of probit or logit models to identify economic weaknesses and anticipate crises. After the seminal paper, Demirguc-Kunt & Detragiache (1998) captured new determinants as the rate of growth of real Gross Domestic Product (GDP), the external terms of trade and the real short-term interest rate, short-term real interest rates, financial liberalization, included by the ratio of credit to the private sector to GDP (Peña, 2017a), and the rate of depreciation of the exchange rate.

Other studies mainly focus on financial or domestic and commercial variables, such as Hardy and Pazarbasioglu (1999), who include consumption and investment variables as explanatory variables, observing that consumption booms in the years previous to a crisis can be a good predictor of banking crises. These authors state that "banking crises are associated with a sharp decline in the real effective exchange rate, but an appreciation in this rate often precedes a crisis". Rose & Spiegel(2010, 2011) include exports/GDP ratio as a factor of a financial crisis. Finally, Boudriga & Ghardallou (2012) find that deterioration in competitiveness is associated with an increased risk of problems in the banking sector. In contrast to this paper, none of all these authors theoretically analyses the possible interacted effects of domestic and commercial variables with tax factors on the triggering of a crisis.

Regarding financial variables, Beck, Demirgüç-Kunt & Levine (2006), find that bank concentration increases banking crisis probability, and Büyükkarabacak & Valev (2010) differentiate private credit from household and enterprise credit. Financial variables can also have international channels of transmission. For instance, Ye & Han (2010) show that financial contagion decreases with geographical distance from the United States, which was the centre of the 2008 subprime mortgage crisis, and that the closer the distance to that centre, the greater the effect. More recently, Cesa-Bianchi et al. (2019) obtain that, even taking into account domestic loans, credit growth in the other countries has a relevant and positive effect on the likely of banking crises in the specific country. Pedro et al. (2018) specify the financial contagion among countries of the same region and from G7 countries to others. Nevertheless, none of these specifications combines the presence of the largest economic powers with G7 countries in a diversified manner.

Finally, Demirguc-Kunt & Detragiache (1998) include government surplus as a percentage of GDP, used to capture the financial needs of the public administration, but without entering into the impact of the source of revenues for the public administration. They provide two reasons: first, a high surplus involves postponing measures to strengthen bank balance sheets, and second, a failure to control the budget deficit is an impediment to successful financial liberalization, which creates problems for banks. The last relevant variable is the ratio of bank cash and reserves to bank assets. Nonetheless, these authors do not go further in the analysis of the public sector regarding the connections with other variables, as Peña (2017b) does by studying the negative interacted effect of income inequality and public surplus on banking crises. Another recent study carried out by Sosa-Padilla (2018), finds three connections between banking crises and sovereign defaults, observing they usually happen at the same time, the high amount of public debt the banks holds, and the effect of these defaults on higher bank credit and GDP crushes.

The importance of income inequality on banking crises has already been theoretically explained before Peña (2017b) by Claessens & Perotti (2007), Rajan (2010) and Kumhof & Rancière (2011) who propose that an increment in inequality led to a credit boom and finally to a financial crisis in the USA at the beginning of the 21^st century, as it did in the 1920s. Krugman (2007) focuses on the extension of bad loans by the private sector, whereas Rajan (2010) highlights the role of the government through its agents. For the determinants of the financial crisis of 2008, Wisman (2013) considers that income inequality, jointly with wage stagnation, generated three dynamics that made the economy vulnerable to crisis.

Icaviello (2008) finds that the long run rise in household debt is explained by higher income inequality, while Roy and Kemme (2012) use the difference of the average productivity and the real wage rate as a measure of income inequality. They also find that the rise of income inequality also contributes to global financial crises.

Finally, as far as we know, the topic of the impact of the VAT experience in a country, measured as the number of years in which a country has settled VAT, on the likelihood of a crisis, is a no dealt issue.

3.   Data sample and methodology

In order to check the above-mentioned relationship, binary econometrical models are estimated, including the most relevant determinants from the seen at the previous section.

Our dependent variable of all the econometric models, crisis, is available in the World Bank according to Laeven & Valencia (2013), and represents a dummy variable that is one if the country is experiencing a systemic banking crisis and zero if not.³ Equations are estimated using a population-averaged and random effects panel logit probability model, as Büyükkarabacak & Valev did (2010)⁴.

³ According to the World Bank, "a banking crisis is defined as systemic if two conditions are met: a. Significant signs of financial distress in the banking system (as indicated bysignificant bank runs, losses in the banking system, and/or bank liquidations), b. Significant banking policy intervention measures in response to significant losses in the banking system"

⁴ As these authors state, for a detailed description of the population-averaged model, see Zeger et al. (1988), Neuhuas et al. (1991), and Wooldridge (2002).

where ${Y_{it}}$ represents the dependent variable crisis, ⁵ ${X_{it}}$the explanatory variables, ${\beta ^*}$ the shift in the logit of the proportion with $Y = 1$ for a raise in $X$ of a unit and ${\alpha ^*}$ is the constant.

⁵ The crisis variable is limited to the Laeven & Valencia (2013) sample, so its threshold is data until 2011. Studying alternative explained variables regarding the presence of economic crises, which could be binary variables based on the growth of the GDP may improve the number of observations, but this try is not close to the targets of the paper.

The employed model is the logit probability model. Other ones as probit probability models have been tried to use, but some conditions for those models as normality were not fulfilled.

Panel data is used, specifically, an unbalanced panel⁶ from the year 1970 to 2011 for 36 countries, all the EU (27) and OECD countries (including New Zealand) with the exceptions of Switzerland, Cyprus, Romania and Malta. Table 1 gives some basic information about data.

⁶ Unbalanced data usually lead to less consistent estimations, nonetheless, robustness checks applied to the results could partially solve this problem.

The explanatory variables used in this paper are the main determinants of banking crises that do not present multicollinearity problems, in addition to our target variables. Some variables are lagged in order to avoid simultaneously problems (Büyükkarabacak & Valev, 2010). The main characteristics of the variables are summed up in Table 2.

The main determinants of financial crises used in this paper are: gdppc, which is the annual percentage variation rate of Gross Domestic Product (GDP) per capita using constant local currency and agammaegated by constant 2010 US dollars. The population is calculated at midyear. The exch variable is the official exchange rate (national currency unity per US$, average for a period). A real exchange rate appreciation affects the competitiveness of the economy and leads to a deterioration in the profitability of the corporate sector leading to financial crisis.

The lerner variable is the Lerner index, a measurement of banking competition that compares output pricing and marginal costs (that is, the mark-up) in the banking market: an increase in the Lerner index indicates a deterioration in the competitive conduct of financial intermediaries. A low banking competition provokes banking concentration and this generates systemic risk, which is involved in crises. The variable infl, which is inflation, is measured by the consumer price index and reflects the annual variation of the cost of a shopping basket for a customer. A lower inflation usually indicates a more stable economy, and then, less exposed to banking shocks. The logarithm of the sum of the bilateral distance of a country to the USA, Japan and France⁷ is measured by dist, in order to capture the contagion effect (Pedro et al., 2018), i.e. geographical contagion of a crisis. A nearer distance with these three economic potencies could mean a higher level of contagion of financial crises. The budget surplus is measured by surplus, public cash surplus or deficit is revenue (including grants) minus expense, minus net acquisition of nonfinancial assets. In general terms, a higher surplus involves postponing measures to strengthen bank balance sheets and is an impediment to successful financial liberalization, which creates problems for banks. Both effects provoke banking crises. Table 3 shows the years when crisis takes the value one.

⁷ According to IMF (2019), USA was the first economic power in Gross Domestic Product (GDP) at current prices, China, from Asia, was the second largest power of the world, and from the same continent Japan was the second largest Asian economy. The next economic potency of another continent was Germany, and France was the third biggest economy of the European continent. The criteria followed in this paper has been to choose the first, the second and the third potency from the three continents of the sample: America, Europe and Asia-Oceania. The continents are ordered by the same order the first representative for each continent appears in the classification: first in the classification to appear is America, so the first potency of this continent, secondly is Asia, so the second potency of this continent, Japan, is employed, and finally France is the third potency of the third continent to appear in the ranking. By doing so, the selected economic potencies of contagion are combines of G7 countries (Pedro et al., 2018) diversified among the biggest economies (Ye & Han, 2010) for each continent.

Finally, the novelty is our target variable, exp, which reflects the VAT experience of a country lagged two years, which is the accumulation of years since a country adopted VAT. A positive impact on the likelihood of a banking crisis is expected, according to the second section. In addition, the interaction between the target variable and exe, and the former with infl, on the one side, and the interaction among these three variables is studied, reflecting the influence of the experience of VAT via exemption of the financial services on VAT (exe), via inflation (infl), and via both of them, expecting a positive impact on banking crises as explained at Section 2. The exemption variable is binary, taking the value one whether there is exemption of financial services on VAT and zero otherwise, calculated as one minus the presence of financial VAT in a country.

VAT experience is lagged for two periods, to minimize simultaneity problems (Büyükkarabacak & Valev, 2010). Data were obtained from the World Bank Database, except exe, which the author obtained from the Lopez-Laborda and Peña(2017a, 2017b) tables and database, exp, obtained from different sources, and dist, obtained from the GeoDist Database (Mayer & Zignago, 2011). In Table 4 we can see the name of the variable with the respective source, and expected sign of the coefficients of our variables with its references, based on the arguments in Section 2. Table 5 shows the correlation matrix.

4.   Empirical results

The main findings of the paper are summarized in Tables 6 and 7, which show the effects of the variables on the likelihood of a banking crisis. The models are estimated following the population-averaged (Table 6) and the random-effects (Table 7) panel logit probability model.

Models a and b from Tables 6 and 7 include population-averaged and random effects logit panel data estimations, respectively, where Models (I) include all the main explanatory variables previously known in the literature, without including the variable VAT experience. Models (II-V) include consecutively the exp variable and the interacted variables exp*exe, exp*infl and exp*exe*infl, reflecting the target variable and the respective channels of transmission of VAT experience to banking crisis. This means that the longer the country levies under VAT, the higher the likelihood of a banking crisis, as expected via increases of prices, or exemption of financials services on VAT, or both of them, could leading to lower credit demand and lower banking profits.

The following impacts are statistically significant on encouraging a banking crisis: infl, so a high economic volatility is detrimental for banking stability, and the target variable with the three channels of transmission. The reason is the statistically significance of the positive coefficients of their respective variables. On the other hand, the following variables discourages the unchain of banking crises gdppc, which means that an economy with stronger economic growth is less exposed to a banking crisis, lerner, therefore a higher competitive banking sector leads to crises, which was not expected by the previous section (Table 4), but could be explained by considering a higher concentrated financial sector is more efficient and stronger and more robust for affording a banking crisis. A higher distance to USA, France and Japan reduces the probability of a banking crisis and the surplus variable is also beneficial for avoiding banking crises, as the statistical and economic significance of the negative coefficients of the variable show. The rest of the variables obtain a statistical significance lower than 10%.

Models from Table 7 aim to check the robustness of the estimations by applying random effects logit models, instead of population averaged, keeping other factors constant. Results from Table 7 are similar to those from Table 6, with the exception of the lerner variable, with a statistical significance on some models of Table 7 but none from Table 6. It is relevant to highlight that possible endogenous issues on the estimation of the models can be diluted by having in mind that usually VATs are implemented for tax collection, trading and efficient purposes, but nor for avoiding banking crises, so the endogenous issue in this case would be mitigated taking into account the exogenous features of VAT related to banking crises.

5.   Discussion, policy measures and concluding remarks

A significant and robust positive influence of VAT experience on the banking crisis risk can be observed. This result confirms our theoretical expectations that time experience with current VATs and GSTs could trigger banking crises. This positive impact of VAT experience on crises is theoretically obtained by studying how bank profits decrease with the VAT exemption on financial services through higher input costs. This unrecoverable VAT on inputs is passed through to the consumers via prices, increasing inflation as well as by the implementation of VAT, and leading to a higher probability of banking crises. This two transmission channels, via exemption and inflation, and the joint interaction of both channels, are also empirically corroborated on the positive coefficient of all these variables. Another theoretical expectation, with a lack of empirical corroboration in this paper, is the link between inflation and banking crises via lower credit demand, and so, lower bank profits.

The implications for researchers and practitioners are, on the one side, showing some weaknesses of the current VAT regimes, which could trigger banking crises, and on the other side, suggesting the need of avoiding the current VAT exemptions of financial services, because it could partially solve the shown negative effects on crises derived from the inefficiencies of not allowing banks to recover their input VATs. Having in mind the difficulty of taxing financial services on indirect taxes in a transaction-by-transaction basis, the solution could be to apply approximately accurate, but feasible, ways of taxation of this kind of services on VAT/GST. An example is the mobile ratio method proposed by Lopez-Laborda and Peña (2018), where the same proportion of financial margin, calculated by an addition basis, is applied for each transaction of the same entity at the same quarter, directly taxing this amount and the explicit fees and commissions. Another possible solution would be applying Financial Activity Taxes (FATs) or direct taxes on value added as applied at Norway.

Further research could include to analyse not only banking crises but also economic crises, for instance considering as explained a binary variable taking one if GDP is falling two consecutive years, contributing a new and more updated database for economic instead of banking crises. As explanatory variables for the contagion effect could also be used the first potencies of the three continents of the sample. Furthermore, it would be worth to research the empirical check of the impact on banking or economic crises for the proposed alternatives to the current status of VAT, i.e. fully taxing financial services, for example, by applying the mobile ratio method on VAT, increasing the participation of direct taxes on tax collection or applying initiatives as the Norwegian experience, where financial services have been fully, or at least higher, taxed by increasing the income and business taxes.

This paper provides further evidence on banking crises determinants: the positive and significant relationship between VATs/GSTs and the likelihood of a banking crisis. The impact of current VAT systems on banking crunches has been analysed and policy measures have been provided in order to decrease the risk of a banking crisis. Using a sample of 36 countries from the EU and OECD during 42 years for 1970-2011 and binary panel logit econometrical models, a robust significance of the positive influence of VAT experience on banking crises is found, corroborating our theoretical expectations, based on the influence of the trade exposure of countries with VAT and the inefficiencies of not fully taxing financial services on VAT. A reform of the current VAT and GST systems reducing the impact of VAT regimes on the likelihood on banking crises, for instance fully taxing the value added of financial services on these taxes as IMF (2011) suggests, would be beneficial for economic efficiency and for the avoidance of crises.

Acknowledgments

The author expresses his gratitude for the funding received from the Regional Government of Aragon, the European Regional Development Fund and the European Social Fund (Public Economics Research Group).

Conflict of interest

The author declares no conflicts of interest in this paper.