Fiscal devaluations: evidence using bilateral trade balance data

Abstract

We study the effects of fiscal devaluations on the trade balances of European Union countries over the 2000–2014 period using bilateral trade balance data. This enables us to control for the coincidence of tax policy measures in different countries, which is an aspect left unconsidered in previous econometric studies. A fiscal devaluation consisting of a budget-neutral tax shift in the amount of 1% of gross domestic product (GDP) from employers’ social security contributions to value added tax leads to a short-term improvement of bilateral trade balance ranging between 0.3 and 0.6% of GDP. An extrapolation of our baseline estimate to the overall trade balance yields an impact of 4.3% of GDP for the whole sample, which is slightly higher than presented in previous empirical research. Applying extrapolation to the trade deficit countries in the euro area shows that these countries’ balance of trade with the rest of the euro area improves by only 0.75% of GDP. Thus, the magnitude of the fiscal devaluation impact on the trade balance varies significantly across countries, depending on their trade openness, among other potentially relevant factors.

Appendices

Appendix 1: variables and data sources

Bilateral goods trade balance (TB): exports minus imports of goods in trade between a home and foreign (trade partner) country, from the home country’s perspective (the home country is the reporting country), in  % of home country’s GDP; source: WITS (COMTRADE) database.

Value added tax revenues (VAT): cyclically adjusted (see “Appendix 2”) revenues from VAT, in  % of GDP. Actual revenues from Eurostat: General government, Value added type taxes (D.2111).

Value added tax rate (VAT rate): standard statutory VAT rate, in %, source: European Commission, Tax and Customs Union. When VAT has been changed within a year, the rate is calculated as the weighted average where the weights correspond to time periods during which each rate has been applied in that year.

Employers’ social security contributions (SCR): cyclically adjusted (see “Appendix 2”) revenues from SCR, in % of GDP. Actual revenues from Eurostat: General government, Compulsory employers’ actual social contributions (D.611C).

Employers’ social security rate (wedge) (SCR rate): average rate of SCR for a single person at the average level of earnings, in % of total labor costs; sources: European Commission, Tax and Benefits Database, and OECD.

Nominal GDP (GDP): Gross domestic product at market prices (current prices); source: Eurostat.

GDP growth (GDPG): Chain-linked volumes, GDP change from the previous period, in %; source: Eurostat.

General government balance (GGB): general government balance, in % of GDP; source: Eurostat: General government, net lending (+)/net borrowing (−).

Government revenues from sources other than VAT and SCR (GREV): Total general government revenue minus actual VAT and SCR revenues, in % of GDP; source: Eurostat.

Dependency ratio (DEP): share of population aged 65 and above in population aged 15–64, in %; source: Eurostat: Population on 1 January by broad age group and sex.

Unemployment rate (UR): registered unemployment rate, in %; source: IMF International Financial Statistics.

Public debt (PDEBT): General government debt, in % of GDP; source: Eurostat: General government, Government consolidated gross debt.

Appendix 2: cyclical adjustment

In order to cyclically adjust the tax revenue series for EU member states, we apply the empirical approach of the European Central Bank (ECB) used for the estimation of the cyclically adjusted budget balance set out in Bouthevillain et al. (2001). This approach is based on the estimation of the cyclical component of individual revenue and expenditure items with respect to relevant macroeconomic bases and it consists of four steps:

• (1) The first step is to identify relevant aggregate demand components or other macroeconomic variables that serve as the natural aggregate base for budgetary items in need of cyclical adjustment. We follow Bouthevillain et al. (2001) who recommend using household consumption for the value added tax (VAT) and the total wage bill for employers’ social security contributions (SCR).

• (2) The second step consists of estimating the elasticity of the budgetary item with respect to the aggregate base. For that purpose we use two specifications. In case that the budgetary item and aggregate base are not cointegrated, we use:

  $$ \Delta lnB_{t}^{j} = \alpha + \delta t + \beta_{1} \Delta lnV_{t}^{j} + A + \vartheta , $$

  (12)

  where $B_{t}^{j}$ represents the budgetary item in current prices, $V_{t}^{j}$ is the appropriate aggregate base in constant prices, while $\beta_{1}$ measures the elasticity of $B_{t}^{j}$ with respect to $V_{t}^{j}$. In case the budgetary item and the aggregate base are cointegrated, we use the following error-correction specification:

  $$ \Delta lnB_{t}^{j} = \alpha + \delta \left( {lnB_{t - 1}^{j} - \gamma lnV_{t - 1}^{j} + \varphi + \ldots } \right) + \beta_{1} \Delta lnV_{t}^{j} + \beta_{2} \Delta lnV_{t - 1}^{j} + \delta t + \vartheta , $$

  (13)

  where $\delta$ is the loading factor, and parameters $\beta_{1}$ and $\beta_{2}$ short-run elasticities of the budgetary item with respect to the aggregate base. In both specifications, $\alpha$ represents the trend change in fiscal ratios, while $\delta t$ captures a change in this trend. One should note that $\delta t$ and $\Delta lnV_{t - 1}^{j}$ were kept in the specification only when they were statistically significant. We use an Engle-Granger model in order to establish the presence of cointegration and specify the error-correction term, as estimates from the Johansen model proved to be too volatile.

• (3) In the third step, one needs to estimate the cyclical component of the aggregate base. For this purpose we employ the Hodrick-Prescott filter in order to estimate the trend component of the aggregate bases, whereby the value of lambda was set using the Ravn-Uhlig frequency rule.

• (4) In the fourth step we calculate the cyclical component of budgetary items and subtract it from the budgetary item in order to obtain the cyclically adjusted series. In order to calculate the cyclical component of the analyzed budgetary item the following formula is applied:

  $$ B_{c,t}^{j} = B_{t}^{j} *\beta_{1} *m_{c,t}^{j} , $$

  (14)

  where $B_{c,t}^{j}$ is the cyclical component of the budgetary item, $B_{t}^{j}$ is the value of the budgetary item in current prices, $\beta_{1}$ represents the elasticity of the budgetary item with respect to the aggregate base, while $m_{c,t}^{j}$ stands for the aggregate base gap, i.e. the share of the cyclical component in the trend value of the aggregate base.