COUNTRY ECONOMIC MEMORANDUM for São Tomé and Príncipe Background Notes Background Note 1 – Economic Growth and Volatility in São Tomé and Príncipe. Rafael Barroso Chenli You I. Introduction 1. The purpose of this background note is to give an overview of the literature on output volatility and economic growth, assess output volatility and its impact in São Tomé and Príncipe (STP). This note is organized in four sections, besides this introductory part. The second section reviews the literature on the impact of output volatility on economic growth. The third section discusses different measures of volatility, calculating volatility for STP across different periods, and compares them to peer countries. The last section offers some policy recommendations. 2. Output volatility and its relationship with growth have been a hot topic in economic research literature for a long time. There is significant controversy about how economic volatility1 affects economic growth. Although the link between economic growth and volatility is theoretically ambiguous, a negative impact of economic volatility on output growth dominates the empirical literature. This negative relationship also holds with newer and better datasets, advanced econometrics methodologies, and for specific country groups. 3. STP generally displays high levels of volatility across different measures throughout time, comparing to peer countries. There are various measures that have been used in the literature to measure output volatility. This background note looks at three categories of measures: (i) ordinary standard deviations of growth rates, (ii) forecasts obtained from regressions; and (iii) filter methodology. By calculating each of these measures, the note finds that STP’s volatility started at high levels in the period 1981-1993. After that, STP experienced sharp increases in these measures into even higher levels during the 1994-2009 period. During 2010-2017, STP was able to reduce volatility, but still exhibited higher volatilities than many other peer countries. 4. There are three mains messages in this note. The first one is that volatility affects growth as supported by the literature review and the econometric estimations carried out in this note. The second one is that STP is a volatile country, although volatility of GDP growth and inflation has declined over time  Senior country economist for STP and consultant respectively, both at the Macroeconomic, Trade and Investment Global Practice of the World Bank Group. 1 In this background note, economic volatility is referred as the volatility of important economic indicators, such as GDP, CPI, Current Account Balance, and Net Lending/Borrowing. 1 and are in line with peers. On the other hand, STP still faces higher volatility on current account balances and net lending/ borrowing than its peers. The third message is that, on average, a fifty percent increase in volatility translates into a 25 percent decrease in GDP per capita growth rates. Finally, policy measures aimed at diversifying exports in terms of goods and markets, reduce the reliance on external finance and fiscal rules can help cushion the volatility and reduce its impact. II. Literature Review 5. There is significant controversy about how economic volatility affects economic growth, as the link between growth and volatility is theoretically ambiguous. The impact could be positive due to the Schumpeterian “cleansing effect� of recessions (Caballero and Hammour, 1991), because of opportunity costs of productivity-enhancing reorganizations during recessions (Hall, 1991), or because volatile sectors command high investment rates following optimal portfolio theory (Imbs, 2007). On the other hand, a negative impact could also be possible, resulting from irreversibility of investment or diminishing returns to investment, or from credit market imperfections that constrain investments during recessions (Aghion and Howitt, 2006)2. 6. The empirical literature on the relation between output volatility and economic growth adds to this controversy. There have been a lot of empirical studies working on the relationship between volatility and growth, using cross-sectional and time series data for as many as 140 developing and developed countries. These studies have resorted to various model designs from Ordinary Least Squares with controls, Auto Regressive models, to Generalized Methods of Moments (GMM) methodology. For example, Ramey and Ramey (1995), Lensink et al. (1999), Martin and Rogers (2000), Fatas (2002), Rafferty (2005), Badinger (2010) and Posch and Walde (2011) find that output growth tends to be lower during periods of higher volatility. On the other hand, Kormendi and Meguire (1985), Grier and Tullock (1989), Caporale and McKiernan (1996), Fountas and Karanasos (2006) and Lee (2010) find that countries with higher output volatility tend to experience higher economic growth rates3. 7. However, a negative impact of economic volatility on output growth dominates the most recent empirical research findings with more recent and improved panel datasets. Badinger (2010) proposes a new instrument4 to identify the causal effect of output volatility on economic growth, while its cross- section evidence from 128 countries points to a negative effect of volatility on growth. Dabusinskas, Kulikov, and Randveer (2012) investigates the impact of macroeconomic volatility on growth in a panel of 121 countries over the period 1980 to 2010 and confirms that macroeconomic volatility is negatively related to economic growth. Using data for 93 countries from 1965 to 2004, Fatas and Mihov (2013) presents evidence that volatility exerts a strong and direct negative impact on growth. Antonakakis and Badinger (2016) examines the linkages between output growth and output volatility in the G7 countries 2 Harald Badinger, 2010, “Output volatility and economic growth�, Economics Letters 106, 15–18. 3 For a comprehensive review and discussion of empirical studies, see Dopke (2004) and Norrbin and Yigit (2005). 4 This new instrument is based on “exogenous� volatility spillovers from abroad and thus entirely unrelated to a country's institutions and policies. This is one of the earliest researches to use this instrument to identify the effect of output volatility on economic growth. 2 over the period 1958–2013 and finds that volatility shocks lead to lower growth, while growth shocks reduce output volatility. 8. Recent research projects have also paid attention to specific country groups, such as small states, small island countries, and Sub-Saharan African (SSA) economies. Easterly and Kraay (2013) consider a large cross-section of 157 countries for which at least 10 years of annual data are available. Of these, 33 are small states defined as having an average population during 1960-1995 of less than one million. They conclude that small states are no different from large states, and so should receive the same policy advice of large states. Gounder and Saha (2007) study South Pacific Island Nations from 1971 to 2003 and concludes that economic growth is negatively affected by output volatility. Melina and Portillo (2018) work on annual data covering 109 countries of the period 1960-2007 and focus on 31 SSA economies. They find that SSA economies stand out by their macroeconomic volatility, while inflation and output tend to be negatively correlated. 9. The negative impact of volatility on economic growth hold consistent across different econometric approaches. Various advanced econometrics approaches have been employed to study the relationships between output volatility and economic growth, using cross-section and panel data models as well as time series analyses of individual countries. Among the research papers mentioned above, researchers have been using linear regressions with multiple controls and fixed effects, combination of different instrumental variables and panel estimation, Vector Auto-Regression (VAR)-based spillover index approach, GMM, among other identification strategies. 10. Although the controversy on the relation between volatility and growth has not been settled, recent research shows that, output volatility generally has a negative impact on economic growth. As a small island country from the SSA region, STP does not differ from the cases discussed by various research papers on small states, small island nations, and SSA economies. Consequently, the literature suggests that volatility affects growth in STP. III. Output Volatility in São Tomé and Príncipe 11. Various approaches have been used in the literature to measure output volatility. These measures can be classified into three categories: (i) standard deviation of growth rates, (ii) standard deviation of forecast residuals; and (iii) filters. The first category calculates the standard deviation of growth rate of GDP, CPI, Current Account Balance, and Net Lending/Borrowing. The second category measures the standard deviation of the residuals, obtained using regressions of these variables over a linear trend and mixed lags. The last category has two definitions. One is the standard deviation of the difference between series of the above-mentioned variables smoothed by the Hodrick-Prescott (HP) Filter or Baxter-King (BK) Filters and actual series, while the other is the average over years of the ratio of the absolute deviation between the observed value of these variables and the value filtered using a moving average process over years5. 12. Many research papers propose measuring volatility based on the standard deviation of the growth rate of a variable. This method is simple and straightforward; however, it assumes that the 5 For detailed definitions, see Cariolle (2012). 3 variables are stationary at first difference. In other words, this approach imposes restrictive hypotheses to the behavior of a series without any prior testing. 13. The next set of measures of volatility is based on the residual or explanatory power of econometric regressions. These measures have the merit of being based on a less restrictive formalization of the process underlying the change in the trend of economic series, in which the variables are not necessarily assumed to be stationary at first difference. It obtains standard deviations of the residuals from three regression estimations: a regression on a linear trend, a regression on a rolling mixed trend, and a regression over three lags and a temporal trend. 14. The last set of measures is the filtered value of statistical series. This type of volatility indicators is based on cyclical fluctuations. The filtering technique has the advantage of not being based on an a priori formalization of the change in the series. It also offers the advantage of being sensitive to breaks in the trend over time. Problems of leakage or compression6 can nevertheless introduce or exclude undesirable or relevant components in order to study the effects of short-term fluctuations. Applied to developing countries, the choice of a smoothing parameter between six and ten initially appears preferable, since the trend in these countries seems to fluctuate more significantly. To measure the filtering values, this note utilizes two indicators: the standard deviation of the difference between series of the above-mentioned variables smoothed by the HP or BK filter and actual series, and average over years of the ratio of the absolute deviation between the observed value of these variables and the value filtered using a moving average process over years. 15. This background note investigates output volatility in STP through various indicators for the period 1981-2018. All the different measures for volatility focus on four variables: GDP growth rate, CPI, current account balance, and net lending/borrowing. The whole data sample is divided into four periods: 1981-1990, 1991-2000, 2001-2014, and 2015-2018, according to economic structure and performance of STP. 16. Volatility generally decreases through time for most of the peer countries. Figure 1 to 5 display a decreasing trend of GDP growth rate standard deviation through time (GDP per capita in Figure 5). Although the measures of volatility of CPI (Figure 6 to 9), Current Account Balance (Figure 11 to 15), and Net Lending/Borrowing (Figure 17 to 21) demonstrate some fluctuations (e.g. for some specific measures, some countries experienced increases from period 1981-1990 to period 1991-2000 and 2001-2014), most of the volatility measures decreased significantly into period of 2015-2018, but the short time interval doesn’t allow for more definitive conclusions on recent volatility. Volatilities of the peer countries are comparatively higher than the average levels of advanced economies and the world. But the CPI volatility of these economies are comparable to the average level of the world. 17. STP’s volatility in GDP growth rate has been decreasing through time and does not stand out among peer countries anymore. Figures 1 to 5 show that STP had one of the highest volatilities in the 6 According to Cariolle (2012), leakage effects occur when some of the long-periodicity variations can be attributed to the cyclical component, and compression effects occurs when part of the short-periodicity cyclical variation can be attributed to the trend. 4 sample in the 1980s, but it declined overtime following a worldwide pattern. However, the decline in STP was faster than in other countries and volatility now in STP is not much different from its peers. Figure 1 – Standard Deviation of GDP Growth Rate: Figure 2 – Standard Deviation of Regression Residual on 1981-2018 a Linear Trend (GDP Growth Rate): 1981-2018 Source: World Bank Staff Calculation (2019) Source: World Bank Staff Calculation (2019) Figure 3 – Standard Deviation of Regression Residual on Figure 4 – Standard Deviation of Regression Residual a Rolling Mixed Trend (GDP Growth Rate): 1981-2018 over Three Lags and a Temporal Trend (GDP Growth Rate): 1981-2018 Source: World Bank Staff Calculation (2019) Source: World Bank Staff Calculation (2019) 18. STP experienced very high levels of inflation volatility from 1981 to 2000, but now displays one of the lowest volatilities among peer countries. STP had very high volatilities of CPI during periods of 1981-1990, and 1991-2000, when the adoption of a centrally planned economy and then the move towards a market-based economy disrupted the country. However, volatility declined after 2000 reflecting more controlled monetary policy especially after the introduction of the exchange rate peg. 5 Figure 5 – Standard Deviation of the Cyclical Figure 6 – Standard Deviation of Regression Residual on Component (GDP per capita)7: 1981-2018 a Linear Trend (CPI): 1981-2018 Source: World Bank Staff Calculation (2019) Source: World Bank Staff Calculation (2019) Figure 7 – Standard Deviation of Regression Residual on Figure 8 – Standard Deviation of Regression Residual a Rolling Mixed Trend (CPI): 1981-2018 over Three Lags and a Temporal Trend (CPI): 1981-2018 Source: World Bank Staff Calculation (2019) Source: World Bank Staff Calculation (2019) Figure 9 – Standard Deviation of the Cyclical Figure 10 – Moving Average Method over Years (CPI): Component (CPI): 1981-2018 1981-2018 Source: World Bank Staff Calculation (2019) Source: World Bank Staff Calculation (2019) 7 Here we are using GDP per capita instead of growth rates. The reason is that according to the literature, this measure is using “level� of variables, instead of “growth rate�, because it calculates the standard deviation of the “difference between series�. So, growth rate, or say “changes�, has already been taken into account. 6 Figure 11 – Standard Deviation of Current Account Figure 12 – Standard Deviation of Regression Residual Balance Growth Rate: 1981-2018 on a Linear Trend (Current Account Balance): 1981-2018 Source: World Bank Staff Calculation (2019) Source: World Bank Staff Calculation (2019) 19. On the contrary, STP has high volatility in current account balance and net lending/borrowing throughout the whole period: 1981-2018. Figure 12-22 display different measures of volatility for current account balance and net lending/borrowing. We can see that STP starts the period 1981-1990 with high volatilities in current account balance and net lending/borrowing comparing to other peer countries. During the period of 1991-2000, all these volatility measures increased significantly in STP and became the highest among all the benchmarking economies. After entering 2000-2014 period, STP experienced a significant improvement with obvious decreases in these volatility measures, reaching back to similar or even lower levels than the 1981-1990 period. Unfortunately, during the 2015-2018 period, STP’s volatility still remains one of highest levels among benchmarking economies for current account balance and net lending/borrowing. Figure 13 – Standard Deviation of Regression Residual Figure 14 – Standard Deviation of Regression Residual on a Rolling Mixed Trend (Current Account Balance): over Three Lags and a Temporal Trend (Current Account 1981-2018 Balance): 1981-2018 Source: World Bank Staff Calculation (2019) Source: World Bank Staff Calculation (2019) 7 Figure 15 – Standard Deviation of the Cyclical Figure 16 – Moving Average Method over Years Component (Current Account Balance): 1981-2018 (Current Account Balance): 1981-2018 Source: World Bank Staff Calculation (2019) Source: World Bank Staff Calculation (2019) Figure 17 – Standard Deviation of Net Lending/ Figure 18 – Standard Deviation of Regression Residual Borrowing Growth Rate: 1981-2018 on a Linear Trend (Net Lending/Borrowing): 1981-2018 Source: World Bank Staff Calculation (2019) Source: World Bank Staff Calculation (2019) Figure 19 – Standard Deviation of Regression Residual Figure 20 – Standard Deviation of Regression Residual on a Rolling Mixed Trend (Net Lending/Borrowing): over Three Lags and a Temporal Trend (Net 1981-2018 Lending/Borrowing): 1981-2018 Source: World Bank Staff Calculation (2019) Source: World Bank Staff Calculation (2019) 8 Figure 21 – Standard Deviation of the Cyclical Figure 22 – Moving Average Method over Years (Net Component (Net Lending/Borrowing): 1981-2018 Lending/Borrowing): 1981-2018 Source: World Bank Staff Calculation (2019) Source: World Bank Staff Calculation (2019) 20. In conclusion, STP successfully decreased GDP growth rate and CPI volatility, reaching comparatively low levels among peer countries; however, STP still maintain very high levels of current account balance and net lending/borrowing volatility. This background note looks into different variables to provide a multidimensional perspective for STP’s output volatility: GDP growth rate, CPI, current account balance, and net lending/borrowing. To obtain robust results, it uses three different volatility measures: ordinary standard deviations of growth rates, forecasts obtained from regressions, and filter methodology. For each variable, different measures return similar results. It has also separated the whole sample into three periods to achieve detailed observations across time. According to the figures and discussions above, volatilities in GDP growth rate and CPI have been decreasing through time for STP and become one of the lowest among benchmarking economies. However, STP still maintains high volatility in current account balance and net lending/borrowing. 21. The impact of volatility on growth can be measured through econometric regressions too. To explore the volatility more in detail, this note assembled a panel data set of 194 countries from 1981- 2018. According to the literature8, it splits the whole period into “5-year� subsamples to calculate the volatility, using the standard deviation of the GDP growth rates. It runs regressions of average growth rates on log of volatility and other control variables, with and without country fixed effects. It can be seen that volatility has been decreasing through time from the initial period: 1982-1986. There is an increase in the level of volatility during the period of 2006-2011, due to the financial crisis in the middle of this period. After that, volatility went back to a lower level again, following the downward sloping trajectory. 22. Econometric estimations show that volatility has a negative impact on growth for all countries and also for the Lower Middle-Income Small States that are comparable to STP. All the specifications return negative and significant coefficients, where the magnitudes range from -0.105 to -0.92. On average, the coefficients are generally around -0.5, which means that 50-percent increase in volatility translates into 0.25-percentage-point (50*0.5/100) lower annual per capita growth. 8 For example, Ramsey and Ramsey (1995) and Dabusinskas, Kulikov, and Randveer (2012). 9 IV. Policy Recommendations 23. To improve GDP growth rate, economic policy has to target on decreasing volatility. According to our regression results discussed in section 22, volatility has significantly negative impact on economic growth. On average, a 50-percent decrease in volatility translates into approximately 0.25-percentage- point higher annual per capita growth. In other word, decrease in volatility will sufficiently improve economic growth, especially for small states that are in the lower middle-income country group. Therefore, it is crucial for STP to implement policies to maintain a stable GDP growth rate in the future and minimize the economic volatility. This will be beneficial to the country’s economic development in the long run. 24. International policy collaboration and experiences are also important for decreasing volatility and improve economic growth in STP. As a small state in the Sub-Saharan Africa (SSA) region, STP is very vulnerable to international and regional macroeconomic shocks, as well as significant industrial shock such as oil industry. Consequently, collaborating with international organizations and regional economic organizations and learning from experiences of similar countries will provide substantial resources and lessons for STP to overcome potential economic shocks that could bring output volatilities, and therefore improve the economic growth in the long term. References Aghion, Philippe, and Howitt, P., 2006. “Appropriate growth policy: A unifying framework,� Journal of the European Economic Association 4(2-3): 269-314. Antonakakis, N., Badinger, H., 2016. "Economic growth, volatility, and cross-country spillovers: New evidence for the G7 countries," Economic Modelling, Elsevier, vol. 52(PB), pages 352-365. Badinger, H., 2010. Output volatility and economic growth. Econ. Lett. 106 (1), 15–18. Dabusinskas, A., Kulikov, D., Randveer, M., 2013. "The impact of volatility on economic growth," Bank of Estonia Working Papers wp2012-7, Bank of Estonia, revised 04 Feb 2013. Dopke, J., 2004. How Robust is the Empirical Link between Business-Cycle Volatility and Long-Run Growth in OECD Countries? International Review of Applied Economics, 18:1, 103-121. Easterly, W., & Kraay, A. 2000. Small states, small problems? Income, growth, and volatility in small states. World Development, 28(11), 2013-2027. Fatás, 2002. "The Effects of Business Cycles on Growth," Working Papers Central Bank of Chile 156, Central Bank of Chile. Fatás, A., Mihov, I., 2003. The case for restricting fiscal policy discretion. Q. J. Econ. 118 (4), 1419–1447. Fountas, S., and Karanasos, M., 2006. "The relationship between economic growth and real uncertainty in the G3," Economic Modelling, Elsevier, vol. 23(4), pages 638-647. Grier, K., and Tullock, G., 1989. “An empirical analysis of cross-national economic growth,� 1951- 1980, Journal of Monetary Economics, 24, issue 2, p. 259-276. 10 Gounder, R. and Saha, S., 2007. Economic Volatility, Economic Vulnerability and Foreign Aid: Empirical Results for the South Pacific Island Nations. Project Report. Massey University, New Zealand. Hnatkovska, V., Loayza, N., 2004. "Volatility and growth," Policy Research Working Paper Series 3184, The World Bank. Imbs, J., 2007. Growth and volatility. J. Monet. Econ. 54 (7), 1848–1862. Kormendi, R., and Meguire, P., 1985. “Macroeconomic determinants of growth: Cross-country evidence,� Journal of Monetary Economics, 16, issue 2, p. 141-163 Lensink, R., Bo, H., and Sterken, E., 1999. « Does uncertainty affect economic growth? An empirical analysis, » Review of World Economics, 135, issue 3, p. 379-396. Loayza, Norman V., Rancière, R., Servén, L., Ventura, J., 2007. Macroeconomic Volatility and Welfare in Developing Countries. World Bank. Martin, P., Rogers, C., 1995. "Long-Term Growth and Short-Term Economic Instability," CEPR Discussion Papers 1281, C.E.P.R. Discussion Papers. Melina, G., Portillo, R., 2018. Economic Fluctuations in Sub-Saharan Africa. IMF Working Paper, WP/18/40. Norrbin, S.C., Yigit, F.P., 2005. The robustness of the link between volatility and growth of output. Rev. World Econ. Weltwirtschaftliches Arch. 141 (2), 343–356. Posch, O., and Walde, K., 2011. “On the link between volatility and growth,� Journal of Economic Growth, Volume 16, Number 4, Page 285 Ramey, G., Ramey, V.A., 1995. Cross-country evidence on the link between volatility and growth. Am. Econ. Rev. 85 (5), 1138–1151. 11