Inflation Targeting: A Three-Decade Perspective 1

Transcription

1 Inflation Targeting: A Three-Decade Perspective 1 Salem Abo-Zaid and Didem Tuzemen 3 First version: July This version: September 1 Abstract Using cross-country data for period 19-7, we study the effects of Inflation Targeting on the levels and the volatilities of inflation, GDP growth and fiscal imbalances in both developed and developing countries. We find that the targeting developing countries manage to bring inflation to lower and more stable levels, and also experience higher and more stable GDP growth. As for developed nations, we find evidence for higher GDP growth and conduct of more disciplined fiscal policy after adopting the regime. The improvements in the budget balances are at least partly attributed to the attempts to achieve the announced inflation target. JEL Classification: E31, E, E, E Keywords: inflation targeting, monetary policy, fiscal policy 1 We would like to thank Carmen Reinhart for her invaluable comments. Department of Economics, University of Maryland, College Park, MD. abozaid@econ.umd.edu. 3 Department of Economics, University of Maryland, College Park, MD. tuzemen@econ.umd.edu. 1

2 1. Introduction Since the early 199 s, several countries followed New Zealand and started to conduct their monetary policies by means of inflation targeting (IT). Currently 7 countries rely on IT in setting a specific inflation rate, which guides the policy makers. As the adoption of the regime is becoming more common around the world, the merits of the regime are being debated more. While the advocates of the regime, such as Svensson (1997), Mishkin (1999), Neumann and Hagen () and King () argue that the regime helps the targeting countries to reach lower and less volatile inflation levels, some economists have opposing views. As a notable example, Ball and Sheridan () claim that the reduction in inflation during the past 1- years is not unique to the IT countries, since a similar pattern is observed in many non-targeting (NT) countries as well. Thus, they claim that low inflation in the IT countries cannot be credited to the IT regime, but it is due to some other common factor, such as the reversion of inflation rates to their historical lower means. Moreover, the doubts on the role of the IT regime in dampening the inflation rates become more significant when the developed countries are considered. This paper aims to shed light on the debated effects of the IT regime for both developed and developing countries, by conducting detailed comparisons between the evolutions of the major macroeconomic variables of these countries during the past three decades. Our work takes steps further beyond the existing literature in some important aspects: we discuss not only the behavior of inflation following the adoption of the IT regime, but also the effects of this regime on growth and fiscal imbalances. To the best of our knowledge, this paper is the first one to consider the relationship between IT and fiscal imbalances. Our main results suggest that the IT regime has important benefits for both developed and developing countries. The inflation rates are considerably lower when countries, especially developing ones, adopt the IT regime. We show that the regime has positive effects on GDP growth, for both developed and developing countries. Also, our results point to a clear convergence in the GDP growth rates among the

3 developing IT countries after the adoption of the regime. Lastly, we find that fiscal imbalances are significantly reduced when countries target inflation. In line with the previous studies, we first present the evidence regarding the inflation and the GDP growth rates in both the IT and the NT countries. Existing studies in the literature have been mostly interested in examining the effects of the IT regime on the inflation rates and on the volatilities of inflation and GDP growth. However, studying only the GDP growth volatility is not sufficient; we believe that it is equally important to examine the evolutions of the actual GDP growth rates following the adoption of the IT regime. This approach is especially useful when testing the claim that countries which attempt to stabilize the price levels are unable to achieve their goals in a costless fashion. More specifically, it is commonly thought that committing to an inflation target can be harmful for economic growth, given that the monetary authority s concern in this case is shifted towards inflation stability and away from economic activity. We challenge this belief by comparing the GDP growth rates in the IT and the NT countries. Our results show that IT has positive effects on GDP growth rates, for both developed and developing countries. The effect of the IT regime on the fiscal policy is another important issue to address. It has been the belief that higher budget deficits are associated with higher inflation rates as discussed in Sargent and Wallace (191) and Amato and Gerlach (). In particular, the common argument is that the government inflates in order to increase the inflation tax revenues and to lower the real value of its deficits. However, the adoption of the IT regime, which is typically announced to the public, may reduce these incentives for the government. Having the IT regime in place, the government may become more fiscally disciplined in order to conduct the monetary policy more successfully. That is, the government may put greater efforts to keep its deficits low, and thus, do its part in achieving the inflation target. We show evidence that this is likely to be the case, and we suggest that the adoption of the IT regime has favorable effects on The idea has been empirically challenged too. See Catao and Terrones (). See Mishkin and Schmidt-Hebbel (1). 3

4 reducing fiscal imbalances, and that it can be attributed to having a better cooperation between the fiscal and the monetary policy. Since the link between fiscal imbalances and the IT regime has never been studied before, our work serves as a first step to fill this important gap in the literature. This paper is organized as follows. Section presents a detailed literature review. Section 3 provides an overview of the data. Section reports the statistical facts about the macroeconomic variables in both the IT and the NT countries. Section presents the econometric methodology and results. Finally, Section concludes.. Related Literature The debate over the role of the IT regime in stabilizing inflation has been the subject of a voluminous number of studies. This section briefly presents some of these studies, with the focus being on the more recent ones. Among the earlier studies, Svensson (1997) claims that the IT regime helps in reducing inflation variability, and that if the regime is flexible, it can help in stabilizing output as well. Bernanke et al. (1999) outline the role of the IT regime in anchoring public s inflation expectations, and claim that the regime provides an explicit plan and a direction for the monetary policy-making, which in turn helps in stabilizing inflation. Mishkin (1999) argues that the IT countries manage to significantly reduce both the current and the expected rate of inflation beyond what would have been in the absence of the IT regime. He further emphasizes that, once the IT countries manage to reduce their inflation rates, these rates stay low and do not bounce up during periods of economic expansions. Following Mishkin (1999), King () shows that, in the case of the IT countries, the inflation rates are not only lower, but they are also less volatile and less persistent. Thus, he concludes that shocks to inflation at a given period in time have short-lasting effects on the inflation rates in the IT countries. Neumann and Hagen () argue that the IT regime helps the countries with inferior economic performances to catch up, and that it promotes

5 convergence across countries. Johnson () re-examines the role of the IT regime in effecting inflation expectations by using a sample of IT and NT OECD countries. He shows that the level of expected inflation in the IT countries falls when controlled for the country effects, the year effects, on-going inflation rates and the business cycles. He concludes that a very important success of the IT regime is this additional reduction in the level of expected inflation. Above mentioned IT-supporting studies have been challenged by the frequently cited and thoughtprovoking paper of Ball and Sheridan (), who claim that the IT regime is irrelevant for bringing down the inflation rates and volatilities when the developed IT countries are considered. Using a sample of IT and NT OECD countries they show that the inflation rates decline in the NT countries as well as the IT countries. They argue that the reason for the fall in the inflation rates is simply the mean reversion. Since the initial levels of the inflation rates have been higher for the IT countries, there is a natural tendency for the inflation rates to return back to their means. Thus, they conclude that the IT regime is irrelevant in stabilizing inflation rates in the countries they consider. Some recent studies support the results of Ball and Sheridan (). Using the data for 7 industrial countries, which adopted the IT regime in the 199 s, Lin and Ye (7) show that inflation targeting has no significant effect on the inflation rates or volatilities. Sims () argues that the IT regime may be more harmful than useful, if there is a substantial chance that the central bank cannot control the path of inflation. The most recent work in this line is Goncalves and Salles (), in which the validity of the results of Ball and Sheridan () are tested for the case of developing countries. Using the data for 3 countries (of which 13 are IT) for the period 19-, they show that the IT countries experience greater drops not only in the inflation rates, but also in the volatilities of the growth rates. Their results imply that the IT regime may have more significant effects on the macroeconomic performance of developing countries rather than developed ones, simply because the latter do not suffer from severe inflation or other macroeconomic problems to begin with.

6 3. Sample and Data Description 3-1. Sample As of today, there are 7 countries using the IT Regime. However, our sample covers 3 countries, since countries adopted this regime only recently. 7 Among the 3 targeting countries, 1 are developing and 11 are developed countries. In order to make comparisons, we compose two control groups, one for developing and another for developed countries. The control groups constitute of 1 developed and 17 developing NT countries. When choosing the control groups, we make an effort to cover a wide range of countries according to their locations, income levels and data availabilities. As for developed countries, this is a challenging task, since the number of countries in this group is relatively small. When choosing the control group for developed countries, we mostly follow the selection in Goncalves and Salles (). The analysis we carry out covers the period There are three main reasons for choosing this time period. First, there is almost full data availability for most of the countries in this time period. Second, several countries adopted the IT regime in the second half of the 199 s, so this time period gives us about a decade of Post-IT observations. Third, given that the inflation rates were very high during the 197 s, considering a longer sample of extremely high inflation rates would only bias our results. Although inflation rates were also high in the early 19 s, choosing a sample that goes back to 19 is important when assessing the effects of the IT regime for the early targeting countries, such as New Zealand, Israel and the UK. We check for the robustness of our results by considering a shorter Pre-IT period that starts in 197, which is the year when the inflation rates declined sharply in many developing countries. Finally, we do not include the data for -1 in our sample, since the severe economic crisis may heavily bias our results without adding any new insights. For the list of the IT countries, see Table A1 in the Appendix. 7 These countries and adoption years are Indonesia (), Slovak Republic (), Turkey () and Ghana (7). For the list of the countries in the control groups, see Table A in the Appendix.

7 We follow Ball and Sheridan () and Goncalves and Salles () in determining the adoption year for the NT countries. For each control group, the adoption year is the average year of the IT regime adoption among the corresponding IT countries. Hence, the adoption years are 199 and 199 for developed and developing countries, respectively. 3-. Data Our data covers different variables of interest: inflation rates, GDP growth rates and government fiscal imbalances. The inflation rate is measured by the annual change in the Consumer Price Index (CPI). The GDP growth rate is the yearly growth rate in real GDP. Following Alesina and Tabellini (), we use the GDP share of government deficits as the proxy for fiscal imbalances. Our main data comes from the IMF s World Economic Outlook (WEO) database. 9 This database, however, does not report the fiscal data for developing countries. When constructing the data on fiscal imbalances, we benefit from the database used in Kaminsky, Reinhart and Vegh (), as well as the fiscal data reported on the websites of central banks and the OECD. 1. Data Facts and Statistical Results -1. An Overview of the IT Countries We first present some basic facts about the IT countries. Figure 1 summarizes the data for IT countries: years of the adoption of the IT regime and the average inflation rates in the two years preceding the adoption. 11 Some studies, such as Mishkin and Hebbel (1) consider only the average inflation rates 9 Database is provided at 1 For the OECD database, see 11 Peru is not shown here since the average inflation rate in the years preceding the adoption of IT is about 1%. 7

8 in the quarter before of the IT adoption. A quarter is a relatively short time period; therefore we look at the average inflation rates in the years preceding the adoption. The figure reveals some very important and interesting facts. First, starting with the adoption of the regime by New Zealand in 199, the shift towards the regime has been continuous along the past 17 years. Second, the average inflation rates in the eve of adoption vary a lot among countries. For example, several countries (mainly developed ones) choose to implement this regime when their inflation rates are already in low levels. Sweden, Thailand, Austria and Finland are some examples. Other countries, such as Peru, Israel and Chile adopt the regime while experiencing high inflation rates. Third, the group of the IT countries is geographically diversified: The adopters are located in Europe, North America and Latin America, as well as Asia, Middle East and Africa. Fourth, the IT countries show a significant diversification when the level of economic development is considered. CHL Figure 1: Average in the IT Countries: Average Inflation Rate (in percents) NZ ISR UK CAN SWE FIN SPA COL MEX POL HUN CZE KOR SA PHI ICE BRA THA NOR GHA TUR SVK IDN AUS SWI Source: Authors calculations.

9 -. Inflation In this subsection, we compare the data on inflation (levels, volatilities and the timing of changes) for the targeting countries with those for the corresponding control groups, considering developed and developing countries separately. The evolution of the inflation rates in all countries over the sample period are shown in Figures A1-A in the Appendix. Tables 1 and summarize the statistical information for developing countries. We observe that, prior to the adoption of the regime the inflation rates are significantly higher in the IT countries, especially between 197 and the adoption years. Although inflation drops dramatically in all countries, the average inflation rates in the IT countries fall below the average in the NT countries. 1 Table 1: Inflation - Developing IT Countries (19-7) Brazil Chile Colombia Czech Republic Hungary Israel Mexico Peru Philippines Poland South Africa Thailand Average Adoption Average Inflation Rate (in percentages) Pre IT IT Post IT Average Standard Deviation of Inflation (in percentages) Pre IT IT Post IT Average Coefficient of Variation of Inflation (in percentages) Pre IT IT Post IT Note: Pre IT: 19-Adoption ; 7-IT: 197-Adoption ; Post IT: Adoption -7. We take into consideration that the average inflation rates in the IT countries may be biased because of the extremely high inflation rates in Brazil and Peru. When these two countries are excluded from the 1 Typically the changes are from double-digit or triple-digit figures to single-digit figures. 9

10 sample, the group s average inflation rates become considerably lower: 3.3%,.% and.1%, for the Pre-IT, 197-IT and Post-IT periods, respectively. 13 Similarly, the average inflation rates for the control group may be biased due to the high inflation rates in Argentina, Bolivia and Bulgaria. Excluding these countries from our sample gives average inflation rates of 1.1% for the Pre-IT, 1.% for the 197-IT and 7.% for the Post-IT period. The exclusion of the outlier observations mostly affects the initial averages for both groups. Table : Inflation - Developing T Countries (19-7) Argentina Bolivia Bulgaria China Costa Rica Côte d'ivoire Dominican Republic Ecuador Egypt El Salvador India Malaysia Morocco Panama Tunisia Uruguay Venezuela Average 1 - Average Inflation Rate (in percentages) Pre IT IT Post IT Average Standard Deviation of Inflation (in percentages) Pre IT IT Post IT Average Coefficient of Variation of Inflation (in percentages) Pre IT IT Post IT Note: Pre IT: ; 7-IT: ; Post IT: Although the developing IT countries initially have considerably higher inflation rates, - almost twice as high compared to rates in the control group - these rates drop much faster than the rates in the NT countries, and the IT countries end up with lower inflation after the adoption of the regime. For example, Latin American IT countries, such as Brazil and Peru, manage to bring their inflation rates down to 13 Excluding data for Poland gives average inflation rates of.3%,.% and.3% for the Pre-IT, 197-IT and Post-IT periods, respectively. 1

11 single-digit levels; while the NT ones, such as Ecuador and Venezuela continue to experience high inflation, although their initial inflation rates are significantly lower than those in both Brazil and Peru. Hence, even when a similar group of countries is considered, the IT countries seem to be more successful in lowering the inflation rates. This is in favor of the claim that the adoption of the IT regime has a significant impact on inflation in developing countries, mainly by moving inflation to lower levels, despite the inferior performances of these countries prior to the adoption of the regime. Another important observation follows when convergence is considered in the case of the developing IT countries. Despite the differences in the initial rates and the rates throughout the adoption years, most of these countries successfully move their inflation rates towards a lower level of around %. Perhaps the most notable examples are Brazil and Peru, which manage, within a short period of time, to bring their inflation rates close to the group average. On the contrary, convergence is not common in the case of the control group. Costa Rica, Dominican Republic, Ecuador and Venezuela continue to have average inflation rates significantly higher than the group average even after 199. Next, consider the volatility of inflation. The average volatility drops to a lower mean of 3.% in the case of the developing IT countries, while the developing NT countries have a higher average volatility of.% in the Post-IT period. But, there is no clear pattern: the average volatility for the NT countries is higher during the Pre-IT period, but it is lower during the shorter period of 197-IT. Excluding the data for the outlier countries from each group gives us a pattern: the average volatility for the IT countries is considerably higher before the adoption of the IT regime, but it is significantly lower afterwards. In particular, the average volatilities in the IT countries drop from 3.1% in the Pre-IT period, and 3.% in the 197-IT period, to only.% in the period after the adoption. At the same time, the volatilities for the NT countries drop from 11.3% (Pre-IT period) and 1.% (197-IT period) to.% (Post-IT period). In other words, although the initial average volatilities for the IT countries are almost twice more than those for the NT countries, the final volatility is only about one half. 11

12 The last result should be interpreted with caution. Typically, the standard deviation tends to be higher when the levels (hence the mean) is higher. It might be the case that the volatilities for the IT countries are higher, simply because these countries have higher inflation rates before the adoption of the regime. In order to check for the robustness of our results, we calculate the Coefficient of Variation (CV) of inflation. 1 We find that the average CVs of inflation in the developing IT countries are slightly lower than that those in the control group. Moreover, while the Pre-IT average CV in the IT countries drops by 3%, the control group has a smaller drop of about %. 1 These observations confirm that the developing IT countries have been more successful than the NT countries, in both lowering and stabilizing their inflation rates. How about the timing of the declines in the inflation rates? For some countries, the inflation rates decline right after the adoption of the IT regime. The most notable examples are the Czech Republic, Mexico and Hungary, where the inflation rates drop in the first year of adoption. Similarly, inflation in Colombia declines sharply in the year of adoption, after moderate reductions in the preceding years. We now consider the developed countries (Tables 3 and ). In the IT countries, the inflation rates decline significantly after the adoption of the regime. In particular, the average inflation rate in the IT countries following the adoption is.3%, which is less than one third of its Pre-IT level of 7.7%. Excluding Iceland - which, relative to the average, has extremely high inflation rates both before and after the adoption - gives a similar result: the average inflation rate drops from.% (Pre-IT) to.1% (Post-IT). Table confirms what has been claimed by several studies: the inflation rates decline substantially in the developed NT countries as well. In the Post-IT period, the average inflation rate in the developed NT countries is %, which is close to the average rate of.3% observed in the developed IT countries. In fact, excluding Japan (which had a deep deflation during the 9 s) from the sample gives an average inflation 1 CV is defined as the standard deviation to mean ratio. 1 Excluding the outlier countries has no significant effect on these results. 1

13 rate of.% for this group. Therefore we conclude that although the IT countries begin with higher initial rates - an average of 7.7% compared to.% in the NT countries they do manage to bring their inflation rates to the similar levels observed in the NT countries. Table 3: Inflation - Developed IT Countries (19-7) Australia Canada Finland Iceland Korea ew Zealand orway Spain Sweden Switzerland United Kingdom Average Adoption Average Inflation Rate (in percentages) Pre IT IT Post IT Average Standard Deviation of Inflation (in percentages) Pre IT IT Post IT Note: Pre IT: 19-Adoption ; 7-IT: 197-Adoption ; Post IT: Adoption -7. Table : Inflation - Developed T Countries (19-7) Austria Belgium France Germany Italy Japan etherlands United States Denmark Portugal Average 1 - Average Inflation Rate (in percentages) Pre IT IT Post IT Average Standard Deviation of Inflation (in percentages) Pre IT IT Post IT Note: Pre IT: ; 7-IT: ; Post IT: As previously mentioned, in some studies, the adoption of the IT regime by a country is claimed to be harmful for output growth, as the central bank conducts a relatively more tightening policy in order to 13

14 achieve the targeted level of inflation. In this subsection we challenge this claim by comparing the GDP growth patterns in the IT and the NT countries. The evolution of the GDP growth rates in all countries in our sample are shown in Figures A-A in the Appendix. Table summarizes our results for the developing IT countries, while Figure presents the average GDP growth rate of each country along the investigated period. Some observations are worth-noting. First, with the exclusion of Colombia and Thailand from the sample, the average GDP growth rate for the developing IT countries becomes considerably higher for the Post-IT period compared to the Pre-IT level. 1 In particular, the cross country average rises by about %, from.% to.%. Second, the growth rates for some countries, such as Columbia, the Czech Republic and Philippines change course immediately after the adoption of the IT regime and show an upward trend afterwards. Table : - Developing IT Countries (19-7) Brazil Chile Colombia Czech Republic Hungary Mexico Peru Philippines Poland South Africa Thailand Israel Average Adoption Average Rate (in percentages) Pre IT IT Post IT Average Standard Deviation of (in percentages) Pre IT IT Post IT Note: Pre IT: 19-Adoption ; 7-IT: 197-Adoption ; Post IT: Adoption -7. The change in the average GDP growth rate is considerably lower when the developing NT countries are considered: corresponding to a change from 3.% in the Pre-IT period to.3% in the Post-IT period as shown in Table and Figure 3. With the exclusion of China - which has a significantly higher average 1 The decline in the growth rate for Thailand is not surprising given the severe crisis this country faced in the late 199 s. 1

15 GDP growth rate compared to the group average - from the sample, the change in the average GDP growth rate for the developing NT countries turns out to be an increase from 3.1% to 3.9%, which is only one half the size of the increase in the average growth rate for the developing IT countries. In addition, the IT countries have lower GDP growth volatility than the NT countries in the Post-IT period. Table : - Developing T Countries (19-7) Argentina Bolivia Bulgaria China Costa Rica Côte d'ivoire Dominican Republic Ecuador Egypt El Salvador India Malaysia Morocco Panama Tunisia Uruguay Venezuela Average 1 - Average Rate (in percentages) Pre IT IT Post IT Average Standard Deviation of (in percentages) Pre IT IT Post IT Note: Pre IT: ; 7-IT: ; Post IT: Figure : Average Rates - Developing IT Countries (19-7) Before (left panel) and After the Adoption (right panel) of the IT regime Source: Authors calculations. Note: The horizontal line shows the cross-country average. Our results for the shorter time period of 7-IT are in line with those for the longer Pre-IT period. In the case of the developing IT countries, the average GDP growth rate increases significantly, while the 1

16 average standard deviation of GDP growth gets cut by more than a half. On the contrary, the developing NT countries experience only a small increase in their average GDP growth rate and a moderate decline in their average GDP growth volatility. Figure : Average Rates - Developing T Countries (19-7) Before (left panel) and After the Adoption (right panel) of the IT regime Source: Authors calculations. Note: The horizontal line shows the cross-country average. In light of the above analysis, we would like to comment on the common belief that a commitment to the IT regime has inferior effects on economic growth. Given that the IT countries exhibit higher GDP growth rates after the adoption compared to the Pre-IT adoption period, and given the fact that their GDP growth rates catch up with those of the NT countries, we do not observe any negative effect of the IT regime on GDP growth. Moreover, when the increase in the average GDP growth rate of the developing IT countries- corresponding to a change from.% (Pre-IT) to.% (Post-IT) - is considered, it is possible to conclude that the IT regime has a positive effect on GDP growth. When we exclude China from the sample and repeat the same analysis, we find that the average GDP growth rate for the developing IT countries is even slightly higher compared to that for the developing NT countries. Figure also shows that most of the developing IT countries converge to a higher average GDP growth rate after the adoption of the IT regime, which is not observed in the case of the developing NT countries 1

17 (Figure 3). This comparison between the developing IT and NT countries suggests that for the IT countries, sharing the same monetary regime renders them to display some similarities not only in the behavior of the inflation rates, but also in the real economic performance. This conclusion is in line with the results in Neumann and Hagen (). Next, we report the main results for the developed countries in Table 7 and Table. The cross country average GDP growth rate for the developed IT countries slightly increases from.% (Pre-IT) to 3.1% (Post-IT). When both Spain and Finland are excluded from the sample, the Post-IT average growth rate becomes almost identical to the Pre-IT average. This pattern differs from the one observed in the case of the developed NT countries. For this group, the average GDP growth rate actually declines from.% (Pre-IT) to.1% (Post-IT). 17 Given that the average GDP growth rate increases for the developed IT countries and it slightly declines (or remains unchanged) for the developed NT countries, it is misleading to believe that the IT regime has an adverse effect on output growth. On the contrary, the regime has a positive effect on the level of GDP growth, but it is more pronounced in the case of developing countries. For both the developed IT and NT countries a declining pattern is observed when the GDP growth volatilities are considered. The average volatility of the GDP growth drops significantly for some of the developed IT countries, such as Australia and the United Kingdom. There is a slight decline for some, such as Sweden, and no change for others, such as New Zealand. Only for Korea, we observe an increase in the volatility of GDP growth, which can be explained with the huge collapse in When Korea is excluded from the sample, the group s average GDP growth volatility drops nearly by 3%, going down from.3% to 1.%. Despite the decrease in the average GDP growth volatility for the developed IT countries, GDP growth is still more volatile in the IT countries than the NT countries. This result is 17 Here, it should be noted that Japan had a long decade of deflation, which can be a source of bias. When Japan is excluded from the group, we find that the average growth rate remains unchanged. 1 See Figure A in the Appendix. 17

18 mainly due to the fact that the IT countries experience higher volatilities before the adoption of the regime. Table 7: - Developed IT Countries (19-7) Country Australia Canada Finland Iceland Korea ew Zealand orway Spain Sweden Switzerland United Kingdom Average Adoption Average Rate (in percentages) Pre IT IT Post IT Average Standard Deviation of (in percentages) Pre IT IT Post IT Note: Pre IT: 19-Adoption ; 7-IT: 197-Adoption ; Post IT: Adoption -7. Table : - Developed T Countries (19-7) Austria Belgium France Germany Italy Japan etherlands United States Denmark Portugal Average 1 - Average Rate (in percentages) Pre IT IT Post IT Average Standard Deviation of (in percentages) Pre IT IT Post IT Note: Pre IT: ; 7-IT: ; Post IT: We close this subsection by re-stressing that we do not find any evidence that would support the belief that a commitment to an inflation target comes at the cost of a lower GDP growth and/or higher growth volatility. We actually find the reverse to be true. The IT countries, particularly developing ones, enjoy 1

19 higher GDP growth rates and lower GDP volatilities compared to the NT countries. Our results not only support those of Gonvales and Salles (), but also take a step further to show that the positive effects of adopting the IT regime are both on the levels and the volatilities. -. Fiscal Imbalances In this subsection we investigate the possible effects of the IT regime on fiscal imbalances. Our study is the first one to consider the fiscal effects of the IT regime. Previous studies have been silent on the fiscal side; therefore the analysis conducted here is aimed to fill a very important gap in the literature. We need to note that there are some data limitations when fiscal policy is considered. As mentioned earlier, we measure fiscal imbalances by the GDP share of government deficits (and allow for negative values). Although the data on the budget imbalances are available for all developed countries in our sample, this is not the case for some developing countries. Therefore, we use shorter time periods for the analysis of some these developing countries. The fiscal imbalances for all countries of interest are shown in Figures A9-A1 in the Appendix. As seen in Table 9 and Table 1, starting from very similar levels of average GDP share of government deficits, both the developing IT and NT countries experience declines in their deficits, albeit moderately. The decreases in the fiscal imbalances in the developing NT countries seem to be higher than that for the developing IT countries. Note, however, that the results may be biased because we include Philippines in the IT group and Egypt in the control group. Excluding these two countries gives the following results: the average of fiscal imbalances in the developing IT countries changes from -.1% to -3. %, while a similar change is observed in the case of the developing NT countries (going from -3.% to -3.%) Also, excluding Hungary from the IT group and India from the NT group shows that the average GDP share of government deficits for the IT group drops from.% to 3.%, while that of the NT group drops from 3.% to 3.%. Again, the changes are very similar. 19

20 Table 9: Fiscal Imbalances - Developing IT Countries (19-7) Brazil Chile Colombia Czech Republic Hungary Mexico Peru Philippines Poland South Africa Thailand Israel Average Adoption Average GDP Share of Government Deficits (in percentages) Pre IT IT Post IT Average Standard Deviation of GDP Share of Government Deficits (in percentages) Pre IT IT Post IT Note: Pre IT: 19-Adoption ; 7-IT: 197-Adoption ; Post IT: Adoption -7. Table 1: Fiscal Imbalances - Developing T Countries (19-7) Argentina Bolivia China Costa Rica Côte d'ivoire Dominican Republic Ecuador Egypt El Salvador India Malaysia Morocco Panama Tunisia Uruguay Venezuela Average 1 - Average GDP Share of Government Deficits (in percentages) Pre IT IT Post IT Average Standard Deviation of GDP Share of Government Deficits (in percentages) Pre IT IT Post IT Note: Pre IT: ; 7-IT: ; Post IT: Following the adoption of the IT regime, fiscal imbalances become more stable as seen in panel of Table 1. The average volatility gets cut by almost two thirds. A similar pattern is observed for the NT group, but in this case the decline in the volatility is considerably lower. In order to check for the robustness of this result we exclude Israel (IT) and Egypt (NT) from the sample since both countries exhibit very high volatilities. The initial average volatility in the developing IT countries falls only

21 slightly (to 3.%) with no change in the Post-IT average. As for the developing NT countries, the initial average volatility drops from 3.1% to.%. Thus, with the major outliers excluded, the previous conclusion is unchanged. Finally, despite the fact that the IT countries start off with different standard deviations - ranging from.% to.% - these values become very similar after the adoption of the IT regime. This pattern is not observed for the NT group. Next, we turn to developed countries. Figure A9 and Figure A1 show sharp decreases in the GDP share of government deficits for both the developed IT and NT countries. The IT countries have considerably lower deficit shares compared to the NT countries. In fact, most of the IT countries succeed to turn their deficits into surpluses in the last decade. The best examples are Australia, Canada, New Zealand and Korea. In this regard, the United Kingdom and Switzerland stand as the major exceptions. Except for Denmark, the developed NT countries still display considerable higher levels of deficits. More interestingly, for several developed IT countries, the change to surpluses occurs immediately after adopting the IT regime. Canada, Australia, Spain, Finland and Sweden are just some examples. 1 Although the GDP shares of government deficits show a declining trend for the developed NT countries as well, there are still high fluctuations. The trends for Germany, France, Portugal and the United States illustrate this fact clearly. Summing up, these countries manage to cut the GDP share of government deficits by about %, but they fail to turn them into surpluses, and hence the actual shares fluctuate around a new lower average. Changes in the levels and the volatilities of fiscal imbalances in the developed IT and NT countries are reported in Table 11 and Table 1. The average of Pre-IT deficits in the IT countries is -1.%, and for the Post-IT era it becomes a moderate surplus of.%. Exclusion of Norway, which has a surplus of more Denmark showed a consistent budget surplus in the past years. 1 Notice that for Finland and Spain, the sharp improvements in the budget imbalance become moderate after abandoning the regime in favor of joining the EMU. 1

22 than %, gives a significantly higher Pre-IT average deficit of about -.%, and a Post-IT balance of about -1%. When the developed NT countries are considered, the average GDP share of government deficits gets cut by almost a half, changing from -.% (Pre-IT) to -.% (Post-IT). The high initial average for this group is highly biased due to the existence of Italy and Belgium in the sample, since these countries experience deficits of around 1% during the Pre-IT period. With the exclusion of these two countries, the Pre-IT average becomes considerably lower (-3.3%), with no change in the Post-IT average (-.%). The results regarding the standard deviations of the fiscal imbalances differ highly across the developed IT and NT countries. More specifically, the average standard deviation for the IT countries remains higher than that for the NT countries. Moreover, it can be inferred that the volatility of the imbalances increases slightly for the IT countries, whereas it declines moderately for the NT countries. Yet, this result is not robust to the exclusion of the outliers in the two groups. For example, once Belgium and Denmark are excluded from the developed NT countries group, the Pre-IT average volatility becomes slightly lower. Table 11: Fiscal Imbalances - Developed IT Countries (19-7) Australia Canada Finland Iceland Korea ew Zealand orway Spain Sweden Switzerland United Kingdom Average Adoption Average GDP Share of Government Deficits (in percentages) Pre IT IT Post IT Average Standard Deviation of GDP Share of Government Deficits (in percentages) Pre IT IT Post IT Note: Pre IT: 19-Adoption ; 7-IT: 197-Adoption ; Post IT: Adoption -7. Standard deviations of the imbalances for these countries are about twice as big as the average standard deviation for the whole group.

23 Lastly, our figures show that several developed IT countries have continuous reversals in their deficits, which causes the actual levels to significantly differ from their mean. Clearly, this leads to high variations. On the contrary, in many developed NT countries, the fluctuations are smaller and hence volatilities are much lower. 3 We repeat the calculations by considering only the deviations from the moving averages. This method delivers an interesting result: the average standard deviations for the two groups of countries increase only slightly. More specifically the changes are from 1. to. for the IT countries, and from 1. to 1. for the NT countries, which are very similar. Table 1: Fiscal Imbalances - Developed T Countries (19-7) Country Austria Belgium France Germany Italy Japan etherlands United States Denmark Portugal Average 1 - Average GDP Share of Government Deficits (in percentages) Pre IT IT Post IT Average Standard Deviation of GDP Share of Government Deficits (in percentages) Pre IT IT Post IT Note: Pre IT: ; 7-IT: ; Post IT: We close this subsection by summarizing our main results. The developed IT countries succeed to improve their fiscal imbalances considerably, with several of them ending up with surpluses. Although there is a cost of a slight increase in the volatility, the success of turning the deficits into surpluses is notably important. On the contrary, the improvements for the NT countries are not sufficient to balance their budgets. We believe that this important difference between the developed IT and NT countries is due 3 In other words, the NT countries now have lower means and the actual values vary around these means. On the contrary, the IT countries exhibit a trend in their series thus making simple averages inadequate measures. We also consider the means for the two groups. With this method, we find that the average GDP share of government deficits for the IT countries drops down to.1% from 1.%, and that for the NT countries drops down to.% from.%. 3

24 to the adoption of the IT regime, which brings superior discipline in budget managing. As for developing countries, the changes in the GDP shares of government deficits for both the IT and the NT countries are similar.. Econometric Analysis Up to this point we presented the basic statistical features of the major macroeconomic variables and compared the results for the IT and the NT countries. In this section we test our results with econometric methods. To do so, we follow Ball and Sheridan () and Goncalves and Salles (), and adopt the Diffs-in-Diffs strategy. Our aim is to see whether inflation targeting affects the levels as well as the volatilities of inflation rates, GDP growth rates and the GDP shares of government deficits of the countries we consider. As in the previous sections, we analyze the results for developed and developing countries separately. We start by defining the initial and the final averages of the variables of interest. The initial average stands for the -year mean of the variable between 19 and 19, whereas the final average stands for the mean of the variable in the last years of the sample. We check for the robustness of our results in two ways: first by changing the definitions of the initial and the final averages, and second by excluding the variable specific outliers from the sample. We define the alternative initial average as the mean of the variable of interest between 19 and the adoption year, and the alternative final average as the mean of the variable between the adoption year and 7. While the year of adoption varies among the IT countries, for the NT countries we continue to use years 199 (developed countries) and 199 (developing countries). The general econometric specification is as follows: Goncalves and Salles () use single data points for the initial and the final values of the variables of interest. We find this method to be problematic since there are large differences in these levels when the time series for each country are considered separately.

25 x = c + β x i + γ D + ε where x is the difference between the final and the initial average of the variable of interest. x i stands for the initial average of the variable. We include this variable in order to control for the possible mean reversions, as clearly explained in Ball and Sheridan (). D is a dummy variable that takes value of one if the country adopts the IT regime, and zero otherwise. This variable is the most important one in assessing the differences between the IT and the NT countries. Finally, C and ε are the constant and the error terms, respectively. -1. Estimation Results for Inflation We report the estimation results in Tables 13 and 1. For all countries, the initial average inflation rate seems to affect the change in average inflation between the initial and the final periods defined above. The coefficient estimate for the initial average inflation rate is found to be negatively significant at 1% significance level in the regressions for both developed and developing countries. For the developed countries, the adoption of the regime has no effect on the change in the average inflation rate. This is due to the fact that developed countries, whether they adopt the IT regime or not, experience similar patterns in the movements of the inflation rates. On the contrary, for developing countries, adopting the IT regime does matter. The coefficient estimate for the dummy variable in this regression has a negative sign and it is statistically significant at 1%. In particular, a developing country adopting the IT regime observes a 3. percentage point higher drop in its inflation rate compared to a developing NT country. Therefore, we find support for our earlier conclusion that adopting the IT regime helps developing countries in lowering their inflation rates. When the volatility of inflation is considered, the coefficient estimate for the initial average is also negatively significant at 1% in both regressions. Yet, the adoption of the regime does not have a statistically significant effect on the volatility of inflation.

26 The coefficient estimate for the dummy variable in the regression for developing countries has the expected negative sign but it is statistically insignificant. We find the same coefficient to be positive, but again statistically insignificant in the regression for developed countries. These results are robust to the changes in the definitions of the initial and the final average values. Table 13: Diffs-in-Diffs Regression Results - Inflation Rate Dependent Variable: Change in Average Inflation Rate Developed Countries Developing Countries Coefficient Std. Dev. Coefficient Std. Dev. Constant IT Dummy * 1.77 Initial Average -.93***.1-1.1***.7 R Note: Significance levels: * 1%, ** %, *** 1%. Table 1: Diffs-in-Diffs Regression Results - Volatility of Inflation Dependent Variable: Change in Average Volatility of Inflation Developed Countries Developing Countries Coefficient Std. Dev. Coefficient Std. Dev. Constant IT Dummy Initial Average -.99***.1-1.***.3 R Note: Significance levels: * 1%, ** %, *** 1%. -. Estimation Results for Estimation results are reported in Tables 1 and 1. The coefficient estimate for the initial average of the GDP growth rate is found to be negatively significant at 1% significance level for all countries.