This Special Feature analyses tax revenue buoyancy in 24 developing Asian economies for the period 1998 to 2020, assessing how tax revenue varied relative to changes in GDP. The chapter estimates tax buoyancy in the short and long run over this period and examines how the revenue impact of the COVID-19 pandemic in 2020 compared with longer-term trends.
Revenue Statistics in Asia and the Pacific 2024
Tax Revenue Buoyancy in Asia
Revenue Statistics in Asia and the Pacific
2. Tax revenue buoyancy in Asia
Abstract
Introduction
This Special Feature has been adapted by the Asian Development Bank (ADB) from “Buoyant or Sinking? Tax Revenue Performance and Prospects in Developing Asia”, an ADB Economics Working Paper written by Samuel Hill (Senior Economist at the World Bank), Yothin Jinjarak (Senior Economist in the ADB’s Macroeconomics Research Division) and Donghyun Park (Economic advisor at the Office of the Chief Economist and Director General, Economic Research and Regional Cooperation Department of the ADB).
To shed light on how well governments are positioned to meet the challenge of increasing tax revenue to finance the higher public spending required to achieve the Sustainable Development Goals, it is important to understand the responsiveness and efficiency of tax collection in developing economies, informed by empirically assessing the links between the tax base and tax revenues. In particular, tax buoyancy measures the response of tax revenues to changes in gross domestic product (GDP) and is therefore a key metric for understanding tax system performance and the outlook for revenues.
This chapter focuses on developing economies in Asia, where tax revenue as a share of GDP tends to be comparatively low, and estimate the short-run and long-run association between tax revenue and output with panel and time-series analysis. This chapter examines the following research questions with reference to 24 developing economies in Asia over the period from 1998 to 2020:
1. How buoyant was tax revenue before the COVID-19 crisis?
2. What was the impact of the COVID-19 pandemic on tax buoyancy? How large were the actual tax revenue losses in 2020 compared to the model estimates?
Tax buoyancy: Definition and selective literature review
Broadly defined as how nominal tax revenues (either in aggregate and/or by individual types of taxes) vary with changes in nominal GDP, tax buoyancy1 estimates inform analysis of long-term fiscal sustainability and the variability of tax revenues over short-term business cycle and provide a formal metric of structural changes in tax revenues.
Where tax buoyancy is estimated to be greater than one, tax revenues are rising more than proportionately to an increase in GDP. In this scenario, tax revenues are structurally increasing and sufficiently buoyant to support fiscal sustainability, even allowing for some increases in the spending as a share of GDP. However, this is not a stable long-run equilibrium, as taxes cannot continue to grow faster than GDP - the tax base - indefinitely. The tax system is also playing an automatic stabiliser role, providing a countercyclical fiscal impulse.
In contrast, when tax buoyancy is less than one, tax revenues are decreasing relative to GDP and weak taxes pose a risk to fiscal sustainability in the absence of spending cuts. The tax system also works against short-run macroeconomic stabilisation.
Finally, a buoyancy of one implies tax revenues are structurally stable, rising in tandem with GDP. Taxes are sufficiently buoyant to support fiscal sustainability so long as spending is not rising as a share of GDP, and the tax system has a neutral influence on short-run changes in output. However, it does not guarantee overall fiscal sustainability, which also depends on debt dynamics.
The analysis in this chapter is informed by, and builds on, previous tax buoyancy studies. In a sample of 30 Asia and Pacific economies spanning 1980-2017, (Jalles, 2021[1]) assesses the buoyancy of total tax revenues, personal income taxes, corporate income taxes, general sales taxes and trade taxes. They control for inflation and tax rates, and they draw on data from the World Bank’s World Development Indicators as well as the IMF’s World Economic Outlook (WEO) and Tax Policy Division databases. Applying an error correction model (ECM) specification with panel data and using the pooled mean group estimator, they estimate short-run tax buoyancy of one and long-run buoyancy of greater than one.
For advanced economies, (Lagravinese, Liberati and Sacchi, 2020[2]) examine 35 OECD countries for the period 1995-2016, assessing the buoyancy of total revenues, total taxes, personal income taxes, corporate income taxes, and general sales taxes. They control for unemployment, inflation, and various policy variables, using data from OECD Revenue Statistics, and OECD National Accounts. They too apply an ECM specification with panel data and use a Dynamic Common Correlated Effects estimator and instrumental variable (IV). They report estimates of short-run and long-run tax buoyancies generally less than one.
Empirical strategy and findings
In its simplest form, the estimation of tax buoyancy includes the growth of nominal tax revenues as the dependent variable and the growth of nominal GDP and lags of both variables as the determinants. The equation below describes this relationship, where β refers to the long-run tax buoyancy and θ refers to the short-run tax buoyancy (OECD, 2023[3]).
From here, the equation that explores the nexus between tax revenue growth and GDP growth - i.e. tax buoyancy - can be augmented to include other control variables, including proxies for changes in policy variables, the business cycle and exogenous shocks.
Tax buoyancy is typically estimated following an ECM approach, which is preferred for two reasons (Dudine and Jalles, 2018[4]; Lagravinese, Liberati and Sacchi, 2020[2]). First, the natural logarithm of both tax revenues and GDP are an integrated series and it is hypothesised that there exists a cointegrating relationship between them. Second, the approach allows for the separate estimation of short-run and long-run tax buoyancies, which may naturally differ. It is reasonable to assume that tax buoyancy will be one over a sufficiently long horizon, given that taxes cannot indefinitely grow faster or slower than GDP. However, in the short run, tax policy features, such as allowances to carry forward losses, may result in revenues deviating from changes in activity (Creedy and Gemmell, 2008[5]).
In the ECM equation, a common interpretation of short-run tax buoyancy is how effectively taxes act as an automatic stabiliser, while the long-run coefficient is an indicator of fiscal sustainability. By including a time dummy variable for 2020 in the analysis, it is also possible to infer how government responses to COVID-19 affected tax revenues and draw some conclusions on the outlook for revenue mobilisation in the medium to long term.
Data and descriptive statistics
To maximise country and temporal coverage, including observations for 2020, the data for this analysis is drawn from a variety of sources, mainly IMF, OECD, and ADB databases. As described in more detail in (Go et al., 2022[6]), tax revenue data are carefully validated for consistency across countries and through time. Tax revenue in domestic currency is drawn primarily from OECD’s Global Revenue Statistics database, supplemented by IMF Government Financial Statistics. GDP data is from the IMF’s WEO database.
Given the focus on tax-buoyancy dynamics, the sample is restricted to economies for which tax revenue and GDP data are continuously available up to and including 2020. The final sample covers 24 developing economies in Asia and the Pacific from 1998 to 2020, giving a total of 552 observations.
Nominal taxes and nominal GDP in domestic currency are converted to their growth rates, using log differences, i.e. growth: . Both series in the log-differences passed the panel unit-root tests (Im, Pesaran and Shin, 2003[7]; Levin, Lin and Chu, 2002[8]).
Estimation results
To estimate tax buoyancy in developing Asia, an ECM of total tax revenue and nominal GDP, both in time-series and panel, is estimated using data for 24 developing Asian economies for the period 1998-2020. Table 2.1 shows the panel results for long- and short-run tax buoyancy using the mean group estimator (Hill, Jinjarak and Park, 2022[9]).
Table 2.1. Tax buoyancy estimates for developing Asian economies using mean group estimator, 1998-2020
Note: The sample covers 24 economies from 1998-2020. The dependent variables is the natural log of tax revenues in local currency units. The explanatory variable is the natural log of gross domestic product (GDP) in local currency unit. The dummy variable takes value 1 for the year 2020 and value 0 for other years.
Source: ADB Staff estimates from OECD Revenue Statistics (accessed 15 Sept 2021); IMF Government Finance Statistics (accessed 22 Oct 2021).
The estimation yields two sets of coefficients, the instantaneous impact of changes in GDP on tax revenues (short-run tax buoyancy) and the long-run relationship between GDP and tax revenues (long-run tax buoyancy). During a major downturn like the COVID-19 pandemic, tax buoyancy may be affected differently, including due to changes in policy or increased tax evasion (Sancak, Xing and Velloso, 2010[10]). To investigate the impact of COVID-19, the baseline specification is augmented to include a dummy variable, which takes the value of 1 for the year 2020 and zero otherwise.
The short-run tax coefficient is 1.3; if the impact of the pandemic crisis is taken into account, this falls to 1.1. The long-run tax coefficient is 1.1, suggesting fiscal sustainability in the sample of economies during the period 1998-2020. Regression results show that both estimated short-run and long-run tax buoyancies in developing Asia are very close to one and statistically significant. The results also indicate that the pandemic had both direct and indirect negative impacts on the region’s tax revenues. The coefficient on the dummy variable for 2020 is statistically significant and indicates that, after controlling for the economic downturn in the region as whole, the pandemic subtracted a tenth of a percentage point from tax revenue growth.
Figure 2.1 plots the long-run tax buoyancy coefficients. The results show that estimated long-run tax buoyancy was around one in most economies, indicative of fiscal sustainability. The estimates are relatively large for some small economies, where GDP and tax revenues are often volatile.
Figure 2.1. Long-run tax buoyancy coefficients, 1998-2020
Note: Upper and lower bounds of estimates from time series Autoregressive Distributed Lags and panel mean-group estimator.
BAN = Bangladesh, BHU = Bhutan, CAM = Cambodia, FIJ = Fiji, FSM = Federated States of Micronesia, GEO = Georgia, INO = Indonesia, KGZ = Kyrgyz Republic, KIR = Kiribati, LAO = Lao People's Dem. Rep., MAL = Malaysia, MLD = Maldives, NEP = Nepal, PAK = Pakistan, PHI = Philippines, PNG = Papua New Guinea, PRC = People's Rep. of China, RMI = Marshall Islands, SAM = Samoa, SIN = Singapore, SRI = Sri Lanka, THA = Thailand, TON = Tonga, VAN = Vanuatu, VIE = Viet Nam.
Source: Hill, Jinjarak, and Park (2022[9]), "Buoyant or Sinking? Tax Revenue Performance and Prospects in Developing Asia", ADB Economics Working Paper No. 656, Asian Development Bank, Manila.
Impact of COVID-19 on tax revenues
In response to the COVID-19 pandemic, many countries implemented fiscal stimulus in the form of tax cuts and exemptions. These measures were often intended to be temporary and, if they turn out to be temporary, then pre-COVID-19 estimates of tax buoyancy are a reasonable starting point for judgments about revenue mobilisation during and after the COVID-19 pandemic. However, there may be sound economic reasons or political pressure for governments to extend or even entrench measures. If tax cuts become permanent, then tax revenues may be structurally lower and the starting point for improving revenue mobilisation may be worse.
This section further explores the impact of COVID-19 on tax revenue with data up to and including 2020. More specifically, it seeks to identify excess tax losses, which requires comparing actual tax revenues in 2020 with those expected based on 2020 GDP outturns. The causal impact of COVID-19 is thus a comparison between the actual tax revenue in 2020 and the model predictions of 2020 tax revenue.
Figure 2.2 shows the gap between actual and model estimated tax revenues in 2020 for the economies in the sample, expressed as a percentage of 2019 GDP. In most but not all economies, the decline in tax revenues in 2020 was greater than predicted by the model estimates. On average (GDP weighted), it is estimated that developing Asian economies endured excess tax revenue losses because of COVID-19, over and above what was expected because of the decline in GDP, equal to half a percentage point of 2019 GDP.
Figure 2.2. Excess tax revenue losses in 2020
Percentage of GDP in 2019
Note: Excess tax revenue losses refer to actual values minus model estimates. Negative values are tax loss beyond what would normally be expected in the GDP downturn. Data exclude Federated States of Micronesia whose revenue loss is estimated at 19.8% of GDP.
BAN = Bangladesh, CAM = Cambodia, GEO = Georgia, GDP = gross domestic product, INO = Indonesia, KGZ = Kyrgyz Republic, KIR = Kiribati, LAO = Lao People's Dem. Rep., MAL = Malaysia, MLD = Maldives, NEP = Nepal, PAK = Pakistan, PNG = Papua New Guinea, PRC = People's Rep. of China, RMI = Marshall Islands, SAM = Samoa, SIN = Singapore, SRI = Sri Lanka, THA = Thailand, VAN = Vanuatu, VIE = Viet Nam.
Source: Hill, Jinjarak, and Park (2022[9]), "Buoyant or Sinking? Tax Revenue Performance and Prospects in Developing Asia", ADB Economics Working Paper No. 656, Asian Development Bank, Manila.
Conclusion
To estimate tax buoyancy in developing Asia, an ECM of total tax revenues and nominal GDP is estimated for a sample of 24 developing Asian economies for the period 1998–2020, using both time-series and panel data approaches. The estimation yields two sets of coefficients: the instantaneous impact of changes in GDP on tax revenues (short-run tax buoyancy); and the long-run relationship between GDP and taxes (long-run tax buoyancy.) To investigate the impact of COVID-19, the analysis includes a dummy variable, which takes the value of 1 for 2020 and zero otherwise.
Regression results from the panel data analysis show that both short-run and long-run tax buoyancies in developing Asia as a whole are very close to one and statistically significant. The results also indicate that the pandemic subtracted a tenth from tax revenue growth after controlling for changes in GDP. To explore tax buoyancy at the country level, the same model is estimated for individual economies. Consistent with regional level analysis, long-run tax buoyancy coefficients are found to be close to one in most economies.
Using coefficients from country-level equations, a simple counterfactual analysis is performed to estimate excess tax revenue lost in 2020 because of the pandemic, reflecting the decline in revenue over and above what would normally be expected given the GDP downturn. Actual revenues in 2020 are compared with estimations of predicted revenues. Based on GDP-weighted figures, it is estimated that, on average, developing Asian economies endured excess tax revenues losses equal to half a percentage point of 2019 GDP because of COVID-19. This is consistent with a negative association observed between the size of COVID-19 fiscal stimulus measures and estimates of tax buoyancy.
References
[5] Creedy, J. and N. Gemmell (2008), “Corporation tax buoyancy and revenue elasticity in the UK”, Economic Modelling, Vol. 25, 1 January, pp. 24-37.
[4] Dudine, P. and J. Jalles (2018), “How buoyant is the tax system? New Evidence from a large heterogeneous panel”, Journal of International Development, Vol. 30, pp. 961–991.
[6] Go, E. et al. (2022), Developing Asia’s fiscal landscape and challenges, Asian Development Bank, Manila.
[9] Hill, S., Y. Jinjarak and D. Park (2022), “Buoyant or Sinking? Tax Revenue Performance and Prospects in Developing Asia”, ADB Economics Working Paper No. 656.
[7] Im, K., M. Pesaran and Y. Shin (2003), “Testing for unit roots in heterogeneous panels”, Journal of Econometrics, Vol. 115, pp. 53–74.
[1] Jalles, J. (2021), “Tax capacity and growth in the Asia-Pacific region”, Journal of the Asia Pacific Economy, Vol. 26/3, pp. 527-551.
[2] Lagravinese, R., P. Liberati and A. Sacchi (2020), “Tax buoyancy in OECD countries: New empirical evidence”, Journal of Macroeconomics, Vol. 63/103189.
[8] Levin, A., C. Lin and C. Chu (2002), “Unit root tests in panel data: Asymptotic and finite-sample properties”, Journal of Econometrics, Vol. 108, pp. 1–24.
[3] OECD (2023), Revenue Statistics 2023: Tax Revenue Buoyancy in OECD Countries, OECD Publishing, Paris, https://doi.org/10.1787/9d0453d5-en.
[10] Sancak, C., J. Xing and R. Velloso (2010), “Tax revenue response to the business cycle”, IMF Working Paper, Vol. WP/10/71.
Note
← 1. The literature on fiscal policy over the business cycle draws a distinction between estimating “tax elasticity” and “tax buoyancy”. Whereas the former is commonly used to refer to the direct impact of changes in tax revenue caused by changes in GDP (or a specific tax base), the latter captures (but does not distinguish between) other influences, including discretionary policy changes.