Macroeconomic effects of exogenous tax changes in a small open economy: narrative evidence from Croatia

ABSTRACT

Following the Romer and Romer narrative approach (RR), this paper investigates the effects of tax changes (personal income tax and value added tax in particular) on economic activity in Croatia. We use the narrative approach to identify exogenous tax shocks and construct a unique time series of these shocks for the 2004–2019 period. However, as Croatia is a small open economy, we adjust the original RR modelling approach by taking into account the relevance of external (demand) shocks. Our results indicate that positive tax shocks lead to a fall in private consumption and output in the Keynesian manner. We show that, compared to direct taxes, indirect taxes exercise a stronger effect on macroeconomic aggregates, but the effect of direct taxes is statistically significant throughout the whole time horizon and does not fade out. As the main contributions of this paper to the existing literature, we highlight the following: firstly, this paper sets forth the first estimates of the macroeconomic effects of tax changes based on the narrative approach in the case of a developing economy; secondly, this study extends the original RR approach by including the effects of external shocks, making this approach more suitable for the analysis of fiscal policy in small open economies.

KEYWORDS:

• Income tax
• fiscal multiplier
• narrative shocks
• GDP growth
• Croatia

JEL-CODES:

• C11
• C32
• E62
• H61

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. See for example, Grdović Gnip (2014, 2015), Ravnik and Žilić (2011), and Šimović and Deskar-Škrbić (2013).

2. For details see Caldara and Kamps (2017) and Martens and Ravn (2014), among others.

3. To be explicit, the first tax bracket is defined from zero to the amount of two personal allowances and taxes at the rate of 15%. The second tax bracket is defined from the level of two personal allowances up to the level of five personal allowances, and the applied tax rate is 25%. The third tax bracket is between five and fourteen personal allowances, being taxed at the rate of 35%, and the fourth tax bracket is applied on income larger than 14 personal allowances, taxed at 45%.

4. The roughly half is based on the weighting method of yearly estimates explained in Section 3.2.

5. For a detailed insight inot all changes that took place see Bejaković (2016).

6. Again, the roughly half is based on the weighting method of yearly estimates explained in Section 3.2.

7. Given a relatively small observation sample, we report the results with two lags. For quarterly data, four lags would be a more appropriate choice. We tried with four lags and the conclusions remained consistent, but given that we want the model to be as parsimonious as possible, we opt for presenting the results with up to two lags being included.

8. We test for the reverse Granger causality as well, i.e. if GDP Granger-causes tax shocks, and again find no evidence.

9. More than 35% of Croatia’s exports go to Germany, Italy and Slovenia, while the same three countries make more than 39% of Croatia’s imports (Eurostat).

10. Plots of cumulative impulse responses are shown in Appendix 2.

11. We use the following equation:

$Y_t=\mathrm{\Theta }+\mathrm{\Phi }{T}_t+\sum _{i=1}^P{\alpha }_i{Y}_{t-i}+{\epsilon }_t$

where $Y_t$ is a vector of tax shocks and the logarithm of real GDP, $\mathrm{\Theta }$ captures the constant, while $T_t$ represents the time trend. The term ${\sum }_{i=1}^P{\alpha }_i{Y}_{t-i}$ captures the lags of exogenous tax changes and lagged output growth, where the latter is included to control the usual dynamics of GDP.

12. In this case we estimate the following equation:

$Y_t=\mathrm{\Theta }+\mathrm{\Phi }{T}_t+\sum _{l=1}^P{\alpha }_l{Y}_{t-l}+\sum _{l=0}^Q{\beta }_l{X}_{t-l}+{\epsilon }_t$

where $X_t$ is a vector of exogenous variables that include Euro Area GDP and Euro Area prices, while other variables remain the same as under the closed-economy specification. Moreover, although our data are quarterly, we include two VAR lags in the system to make the model as parsimonious as possible, given a relatively short observation period, especially in the case when the exogenous block is included in a three-variable system. One additional reason is also that in this way estimates across all specifications are more comparable. As for the tax shock lags, we estimate the models up to eight tax shock lags.

Additional information

Funding

This work has been supported by the Croatian Science Foundation under project number IP-2016-06-4609.