收入不平等、政府支出与实际汇率——基于跨国面板数据的研究

已有较多研究讨论了实际汇率的决定因素,而从收入不平等角度出发的研究并不多。本文搜集和整理了172个国家和地区1970年到2016年的跨国面板数据,分析了收入不平等对一国实际汇率的影响,并引入政府支出探究了收入分配对非贸易品部门和实际汇率的影响机制。实证检验结果表明,对于非OECD国家,收入不平等和实际汇率显著负相关,即收入越不平等,实际汇率高估越严重,而在OECD国家中这一现象并不存在。进一步的影响机制分析发现,对于非OECD国家,一国收入不平等加剧会导致该国政府支出增多,从而扩大了非贸易品部门规模,导致非贸易品的相对价格上升,使得实际汇率高估。

Income Inequality,Government Expenditure,and Real Exchange Rate: A Study Based on Cross-Country Panel Data

The stability of the real exchange rate is important for economic development. It affects trade volume and investment in addition to the stability of the country's financial and currency markets. The impact of income inequality on the real exchange rate, however, has been underexplored in the literature. To study this, we collect and combine different data sets to form a cross-country panel data covering 172 economies from 1970 to 2016. Armed with this data set, we test whether income inequality impacts the real exchange rate. We find that income inequality negatively covaries with the real exchange rate within non-OECD countries but does not significantly relate to the real exchange rate within OECD countries. In other words, in non-OECD countries, the greater the income inequality, the more overvalued the real exchange rate. To verify the robustness of the results, we perform three tests, and we find that the results are still valid. First, we replace the key variables with indicators from different databases. For example, we use the Gini coefficients in the WDI database and the WIID (World Income Inequality Database) to replace the original income inequality indicators (in the SWIID). The real exchange rate indicator used in the benchmark regression is calculated by the price level measured based on actual output in PWT8.2. We also use indicators calculated using the other two price indices (based on domestic absorption and actual consumption respectively) and the real effective exchange rate published by the WDI and IMF. Second, we perform the regressions based on samples in different periods and varying degrees of income inequality, and obtain results consistent with the main analysis. Third, the real exchange rate also has some impact on income distribution, so the above regressions may have an endogeneity problem. In the absence of suitable instrumental variables, we use the difference GMM and system GMM to re-estimate.
We then explore the channels through which income inequality affects the real exchange rate. One strand of the literature shows that income inequality influences government expenditure. When income inequality is high, the government redistributes income from taxes to mitigate social polarization. Another strand of the literature shows that government expenditure affects the real exchange rate. We conjecture that changes in income inequality affect the demand and relative price of non-tradable goods, thereby affecting the real exchange rate. To confirm our conjecture, based on our empirical model, we find that for a non-OECD country, as its income inequality increases, the share of government expenditure relative to GDP increases. This result is still valid after controlling various factors and using GMM. We also find that when government expenditure increases in non-OECD countries, the non-tradable sector expands and the relative price of non-tradable goods increases. This relationship is not found in OECD countries. Taking a further step, we confirm this main channel and rule out others using the intermediate effect model and the regression-decomposition method of Blanchard et al. (1993).
This paper makes three main contributions. First, our new data set covers 172 economies over more than 40 years, from 1970 and 2016, with rich variations in income inequality, which makes it possible to divide countries into groups without significantly reducing variation. With such a remarkable data set, our statistical results are robust across different specifications. Second, reversal causality between changes in real exchange rates and income inequality cannot be ruled out by regular OLS regressions. The dynamic panel data method used in this paper confirms that our result is not driven by reversal causality. Third, the discovery of a government expenditure channel through which income inequality impacts the real exchange rate complements the literature.
Our results show that a more equal income distribution is beneficial to economic growth because it stabilizes the real exchange rate (Rodrik, 2008); conversely, polarized income distribution leads to persistent overvaluation of the real exchange rate, resulting in financial crises (Gourinchas and Obstfeld, 2012). The policy implication is that mitigating income polarization is fundamental to stabilizing the real exchange rate and enhancing competitiveness. More importantly, if income polarization grows, government spending will increase, which will lead to domestic currency overvaluation. This indicates that policy makers should take into account the impact of income inequality, because it has a long-term impact on the relative price of non-tradable goods and the real exchange rate.