Coping with the recent financial crisis, did inflation targeting make any difference?

Transcription

1 Coping with the recent financial crisis, did inflation targeting make any difference? Armand Fouejieu A. * (Preliminary draft, January 2012) Abstract: Countries have faced one of the greatest economic shocks recently. The effects of the financial crisis went largely among the financial markets and hit the real economy. The aim of this study is to investigate whether inflation targeting helped countries which implement this monetary policy strategy to perform better during the 2008/2009 financial crisis. Based on the literature, we first present some arguments suggesting that inflation targeters can be expected to do better when facing a global shock. Our empirical investigation is conducted on a large sample of developed and developing countries. Difference in performances between targeters and non-targeters during the crisis are assessed in two ways: central banks performances in terms of inflation and interest rate and more general economic performances in terms of GDP growth. We apply difference in difference in the spirit of Ball and Sheridan (2005) and find that there is no significant difference between the two groups concerning inflation and GDP growth. However, the rise in interest rates during the crisis has been significantly less pronounced for targeters. JEL Classification: E00, E4, E6 Key words: inflation targeting, financial crisis, macroeconomic performances, difference in difference. (*) Laboratoire d Economie d Orléans (LEO), University of Orléans, CNRS, Rue de Blois, BP Orléans Cedex 2, France. armand.fouejieu-azangue@univ-orleans.fr. I am grateful to Yannick Lucotte, Jean-Paul Pollin and Patrick Villieu for their comments and remarks. I also thank Céline Colin, Christian Ebéké and René Tapsoba.

2 I- Introduction The global economy has recently faced a large shock due to a financial crisis originating from the US financial market and especially the US subprime market. What caused the 2008/2009 financial crisis is subject to debate among economists. For Stiglitz among others, monetary policy management during the economic boom preceding the crisis is responsible of the financial bubble which crashed. The five years preceding the crisis were characterized by a large increase in liquidity 1 coupled to relative good control of inflation, encouraging demand for credit, since risk premium and interest rate went down at the same time. Consumer indebtedness led to debt burden when interest rates were raised by central banks increasing liquidity being unsustainable for sound monetary policy. Conversely, Dooley (2010) argues that the crisis has nothing to do with monetary policy. He highlights the ineffectiveness of financial regulation and the lack in financial innovations control. Rose and Spiegle (2009) point out the role of financial system regulation during the global boom. For these authors, the regulatory framework could have encouraged the risk taking by financial institutions through the implicit designation of some of them as too big to fail. Financial institutions bypassed the regulatory system and had more and more risky activities, essentially by securitising loans in order to divert risks - without reducing them. The consequences of this crisis affect primarily the financial sector increasing uncertainties, decreasing assets prices, bankruptcies - and spread to the real economy through credit channel. On the real economy, consequences were among others, decrease in investment mainly due to restricted access to funding and rising interest rates -, rise in unemployment and a huge drop of GDP. Despite exceptional measures taken by governments and central banks, numerous economies suffered and continue to suffer from the 2008/2009 financial crisis. The aim of this paper is to investigate the comparative performances of inflation targeting during the recent financial crisis. It is a contribution to the relatively poor literature on the performance of inflation targeting when facing economic shocks. Neumann and Von Hagen (2001) compare targeters to non-targeters during the 1978 and 1998 oil prices shocks, on three criteria: inflation, long term interest rate and short term interest rate on 1 The huge increase in exchange reserves accumulation in emerging market economies can also explain this global liquidity increase. 2

3 interbank market. Using the difference in difference approach, the main conclusions are follows: the rise in inflation, long term and short term interest rate were lower for targeters - even if this difference is not statistically significant for inflation rate. Mishkin and Schmidt-Hebbel (2007) also compare targeters and non-targeters facing oil price shocks. Their main hypothesis is that, if inflation targeting increases central banks credibility in anchoring price expectations, one could expect targeters to perform better in terms of inflation and the consequences of shocks through exchange rate to be less pronounced. To test this hypothesis, impulse response functions and VAR approach are used. The results of Mishkin and Schmidt-Hebbel (2007) are in favour of targeters. In their paper, the authors also investigated how domestic interest rate reacts to international interest rate movements. This is to evaluate the independence and the credibility of national monetary policy vis-à-vis to world interest rate evolutions. Their main finding is that inflation targeting helps targeters domestic interest rate to be less sensible to international interest rate. In other words, inflation targeters central banks are more independent from world interest rate shocks. Carvalho (2010, 2011) analyses the difference between inflation targeters and non-targeters during the recent financial crisis. Carvalho (2011) is, at the best of our knowledge, the only study which empirically tests if inflation targeters outperformed the others during the recent financial crisis. He considers 51 countries, including 23 inflation targeters and find that countries which adopted the inflation targeting monetary strategy did better than their peers during the crisis. This paper provides a rigorous approach to investigate the difference between targeters and non-targers during the 2008/2009 financial crisis. First, we start by discussing the main raisons why inflation targeters can be expected to do better during the crisis. Indeed, based on the empirical literature, inflation targeting seems to be associated to better control of inflation, better fiscal discipline, less volatile exchange rate and more exchange reserves accumulation. Inflation targeting is also associated to more central banks credibility and more room for loosing monetary policy when necessary. As we will argue, these advantages related to inflation targeting can make a difference during a crisis. Second we use a more rigorous econometric approach in the spirit of Ball and Sheridan (2005) to empirically test our hypothesis. Targeters and non-targeters are compared in two ways during the crisis: (1) on central banks performances, based on inflation and interest rate; (2) in more general terms using GDP growth. To do so, the difference in difference approach is applied, avoiding selection bias and taking into account the bias 3

4 which could arise from regression to mean - following Ball and Sheridan (2005). We use a large sample of developed and developing countries during the 2003/2009 period. The findings suggest that, even if there is no difference between targeters and non-targeters in terms inflation, targeters seem to have been more able to contain the rising interest rate during the recent financial crisis. Overall, in terms of GDP growth, there is no significant difference between the two groups, suggesting that, taking the economy as a whole, inflation targeting did not make a difference during the 2008/2009 financial crisis. The rest of the paper is organized as follows: section II presents supportive arguments that inflation targeting can make a difference during the crisis. Section III presents the empirical tests and results. In section IV, the robustness checks are conducted and section V concludes. II- Why targeters can be expected to do better during the crisis? While it is interesting to compare targeters to non-targeters during the crisis, it is maybe more important to highlight the reasons why one can expect inflation targeting to provide better macroeconomic conditions than other monetary policies, to cope with a financial crisis or a global shock. This is the aim of this section. The recent financial crisis takes its origins in the US financial system and spreads to the rest of the global economy due to the global financial integration. The consequences of the crisis first affect the financial sector but also the real economy, mainly through the credit channel. Real interest rate went up from 6% on average in 2003/2007 to 12% in , investment declined, unemployment raised and GDP growth dropped - from around 6% in 2007 to around 4% in 2008 and less than 0 in 2009 for the global economy. Facing a crisis with such effects, why inflation targeters can be expected to do better than the others? Our argumentation is based on two fields of the empirical literature: the literature on inflation targeting and the recent literature on the financial crisis. The first one highlights some differences between targeters and non-targeters. We refer to the second one to see how these differences can affect the countries resilience during the 4

5 crisis 2. Going through the empirical literature on the effects of inflation targeting, there are some structural macroeconomic differences which can favour targeters when facing a shock. This concerns fiscal and monetary policies management, and international economic variables. Regarding fiscal policy performances, Tapsoba (2010) highlights 3 channels through which inflation targeting can improve fiscal discipline. The first is related to the requirement of no fiscal dominance for the effectiveness of inflation targeting; the second 3 is related to the Olivera-Tanzi effect ; and the third is related to the flexible exchange rate that is associated to inflation targeting. In his empirical investigations on a large sample of developed and developing countries, Tapsoba (2010) finds that inflation targeting improves fiscal discipline especially when one takes into account the duration since the adoption of this monetary strategy. Lucotte (2010) investigates the consequences of inflation targeting adoption on fiscal discipline through tax collection in developing countries. Using a sample of 59 countries including 19 targeters, estimations based on the propensity score matching approach reveal that on average, public revenues are higher for inflation targeters. Tapsoba (2010) and Lucotte (2010) show that inflation targeting positively affects fiscal policy management - as compared to other monetary strategies. Inflation targeting is then associated to higher fiscal discipline. Therefore, we can expect inflation targeters to enter the crisis with better conditions concerning fiscal policy and especially, less indebtedness. This can be determinant during the crisis since the recent empirical literature shows that countries with higher government debt - especially external short term debt - are more affected by the crisis - see for example Blanchard et al (2010), Tsangarides (2010), Carvalho (2011). More generally, one can assume that countries with higher total debt will face more constraint during the crisis. For example, higher indebted countries will be less able to undertake necessary fiscal stimulus during the crisis, because debt service is already high and they cannot easily run up more debt. Since inflation targeting provides more fiscal discipline, we can expect targeters to have more sound fiscal policy during the period preceding the crisis, and then to have more 2 We follow the literature on the recent financial crisis which argues that the initial macroeconomic conditions of the economies determine how these countries faced the crisis. See for example Lane and Melesi-Ferretti (2010). 3 The Olivera-Tanzi effect refers to a context in wish high inflation tends to reduce the volume of tax collection and the real value of tax revenue collected by a government. 5

6 scope to implement necessary adjustments during the crisis. Graph 1 presents the average government debt and the short term external debt in percentage of GDP. As expected, the two ratios are lower for targeters. Graph 1: 1.a: Government debt (% GDP) 1.b: Short term external debt (% GDP) Source : author calaculations About international economic variables, Rose (2006) analyses the implications of inflation targeting adoption for the exchange rate volatility, external reserves accumulation, sudden stops of capital flows and current account balance. 23 inflation targeters and 45 non-targeters are confronted using data from 1990 to The results suggest that inflation targeting tends to reduce the exchange rate volatility; countries which adopt inflation targeting seem to be less exposed to sudden stops of capital inflows 4. For external reserves and current account balance, it seems to be no significant difference between the two groups. In Lin (2010), questions are the same to those of Rose (2006) mentioned above. Using propensity score matching, his empirical analysis is based on 74 developed and developing countries including 23 inflation targeters - with data from 1985 to The results are follows: for developing countries, inflation targeting reduces the exchange rate volatility and increases external reserves 4 Note that in many case the coefficients are not statiscally significant but appear with the right sign. 6

7 accumulations. Conversely, for developed countries, inflation targeting is associated to an increase in exchange rate volatility and less external reserves accumulation. The conclusions of Rose (2006) and Lin (2010) that inflation targeting is associated to lower exchange rate volatility and higher external reserves accumulation - at least for developing country - can make a difference when analysing how countries face crisis shocks. As pointed by Calvo (2010), international reserves play a role during the crisis since we do not have a global lender at last resort. In normal time, when everything is going well for the economy, we do not worry about external reserves. During a global economic downturn and a credit crisis as the one the world faced in 2008/2009, external reserves are vital since central banks have to deal with credit problems going beyond the countries boundaries. Indeed, in that context, it is less easy to get foreign currency in return for national currency. Calvo (2010) further notices that the European Central Bank got a currency swap arrangement with the Federal Reserve to be sure to have enough dollars; highlighting the importance of external reserves during a global crisis. Reserves can also serve as guarantee and allow countries to access funding on international markets at a lower rate. In that case, reserves can be considered as a sign of countries solvability. International reserves also allow countries to continue to insure its imports during a crisis - at least the most important ones - without compromising some production sectors which rely on specific imports. Exchange rate volatility - which could be considered as an indicator of domestic currency stability on the international markets - can also make a difference on the countries resilience during the crisis. As our empirical results will show, the more exchange rates are volatile during the pre-crisis period, the more the negatives effects of the crisis. With this in mind, if inflation targeting favours external reserves accumulation and exchange rate stability, or, but at a lesser extent, current account balance equilibrium and less exposition to sudden stop of capital flows, we can expect targeters to enter the crisis with better external position and then to be more resilient. Graph 2.b shows that on average, inflation targeters faced less current account balance deficit than non-targeters. Graph 2.a tends not to confirm the empirical results found in the previously cited papers. Indeed, external reserves expressed in month of import are lower for targeters during the five years preceding the crisis. On the monetary policy point of view, as pointed by Carvalho (2010), on average, interest rates should be higher for inflation targeters during the economic boom that preceded the crisis, since monetary policy should be more responsive to the increasing liquidity and 7

8 inflation. Thanks to this higher interest rate, domestic investors are less prompted to acquire high-yields, but dubious foreign assets - since their funds are quiet well remunerated in the domestic economy -; so that targeters preserve their financial system from external financial shocks 5. Graph 2: 2.a: External reserves in month of imports 2.b: Current account balance (% GDP) Source : author calculations During the crisis there are some characteristics related to inflation targeting strategy which could be expected to help targeters to better manage shocks. Higher interest rates for targeters before the crisis give to their central bank more scope to reduce their rates during the crisis in avoiding the zero nominal interest rate boundary. Empirical studies tend to support that inflation targeting enhances central bank s credibility Johnson (2002) shows that inflation targeting reduces inflation expectations, Levin et al (2004) show that inflation targeting succeeds in disconnecting the current inflation expectations from the past inflation realisations. Indeed, inflation targeting monetary strategy relies mainly on central bank independence, transparency and responsibility; three characteristics which provide more credible monetary policy. Since during the crisis one of the main concerns was increasingly uncertainties and market failures on the financial market - moral hazard and adverse selection -, it can be assumed that the more central 5 Graph 4 in our empirical investigations clearly shows that real interest rates were higher for targeters during the five year preceding the crisis. 8

9 banks decisions and interventions to deal with the crisis will be credible, the more these interventions will have the desired effects on the financial market and the real economy. If inflation targeters central banks are more credible than others, we can expect them to do better during the crisis. Also thanks to their higher credibility, as suggested by Carvalho (2010), emerging countries central banks which adopted inflation targeting have more room for monetary easing during the crisis without compromising their inflation objective. Central bank credibility can also help to deal with deflation phenomenon that can rise from the global crisis 6. Overall, if inflation targeters central banks are more credible, we can expect them to be more effective than their peers in dealing with different challenges they faced during the crisis. All these arguments, mainly suggested by the literature, are in favour of inflation targeters to outperform their peers during a shock like the recent financial crisis. In the next section we empirically test this hypothesis. III- Empirical analysis Inflation targeters and non-targeters will be confronted on two main ways. First we will test the difference in monetary policy effectiveness during the crisis, using two variables: inflation and interest rate. Second, the two groups will be compared on the basis of their general economic performance using GDP growth. Without any other mentions, our data are from the World Development Indicators 2011 The World Bank. Before we start our empirical investigations, let us present the methodological approach that will be applied, the sample and the period we focus on. Methodological approach, period and sample Methodology We are interested in the following question: did inflation targeting make a difference during the recent financial crisis? The difference in difference approach is used to give an answer. It seems to be appropriated here since the crisis can be considered as an 6 This view is supported by the governor of the Canadian central bank, Mark Carney, cited in Carvalho (2010), p. 4. 9

10 exogenous shock and treated as an event study. Particularly, we are going to estimate the effect of inflation targeting on change in our main variables during and before the crisis. Formally, the equation can be written as: Y cr Y pre = α + βit + θx + ε (1) Where Y is the output variable on which we want capture the impact of inflation targeting, with cr indicating the crisis period and pre the pre-crisis period. IT is a dummy variable taking the value of 1 if the country is inflation targeter and 0 otherwise, X is the vector of control variables and ε is the error term. As pointed by Ball and Sheridan (2005), the β coefficient can be biased, especially if Y pre is correlated to IT. The idea is quite simple. Let us consider the example of interest rate. As we already discussed above, interest rate is higher for targeters in the pre-crisis period because of this monetary strategy. So, the difference in interest rate - during and before the crisis - will tend to be lower for targeters and the coefficient β will produce spurious effect - especially, it will tend to overvalued the performance of IT in containing the rising interest rates during the crisis - since we are not taking into account these initial differences due to inflation targeting. β is biased in this case. In order to deal with this possible bias, we follow the recommendation of Ball and Sheridan (2005) 7 and introduce Y pre as a control variable in the equation. Therefore, the final generic model we will test is: Y cr Y pre = α + βit + θx + φy pre + ε (2) In equation (2) the coefficient β measures the effect of inflation targeting on the change in Y, given some initial condition. For example, let Y represents GDP growth. If β is significant and positive, it means that inflation targeters did better during the crisis than non targeters with the same average GDP growth in the pre-crisis period. Now we get the real effect of inflation targeting during the crisis. Analysis period and sample We use data from 2003 to The pre-crisis period consists of the five years preceding the crisis 2003 to 2007, while the crisis period is the year As Graph 3 shows, 7 For more details and mathematical explanations, interested reader should refer to the appendix on the methodology in Ball and Sheridan (2005). 10

11 the effects of the crisis were most pronounced in 2009 the global GDP growth was negative as compared to about 4% in 2008 and The real interest rate reached his higher level around 12% and inflation rate shows its lower level for the last decade. For robustness check, we will take the two years 2008/2009 as crisis period, since the literature generally considers the beginning of the financial crisis in September 2008 the failure of Lehman Brothers. Concerning the sample, we refer to Lin (2010) and Roger (2009) for inflation targeters 30 countries, see Appendix 1. We drop those which abandoned the strategy Spain and Finland, in , and those which adopted inflation targeting during our analysis period in other to avoid the selection bias Indonesia, Romania, Slovakia 8 and Guatemala which adopted inflation targeting in 2005; Turkey and Serbia in 2006; Ghana in Graph 3: Global evolution of GDP growth, real interest rate and inflation Source: author calculations Finally our targeters group consists of 21 countries, all of which started implementing inflation targeting before our analysis period. For the control group non-targeters -, we applied a selection criterion in the spirit of that of Lin and Ye (2009) based on the GDP per capita. In order to get some homogeneity between the two groups, we keep in the control group only countries with the average GDP per capita at least as large as the poorest inflation targeter during the period 2003/2007. This criterion is applied for the whole sample as well as for subsamples of developed and developing countries. By 8 Slovakia joined the European Union in

12 applying this selection method which is not too restrictive, we want to keep a large sample of countries in order to make our conclusions as general as possible. Note that for each of our estimations, we also drop countries for which data on the dependant variable inflation rate, real interest rate or GDP growth - are not available. Nevertheless, depending on data availability for control variables, the number of observations will vary from one regression to another. We now turn to the empirical tests. Testing comparative performances of central banks We want to compare central banks performances during the financial crisis. To achieve this objective, targeters and non-targeters are confronted on two main indicators: inflation rate and real interest rate. Applying the above presented methodology, the sample consists of 110 countries 9. Inflation rate is usually considered as the outcome of monetary policy. Note that during the crisis the concern was deflation scare since inflation falls down for the global economy. Therefore, rather than assessing the role of inflation targeting in lowering inflation as it is generally the case in literature -, the purpose here is to know whether inflation targeting has helped targeters to be less affected by the decrease in general prices level during the crisis. Considering equation (2) where Y is inflation rate, we expect β to be positive, suggesting lesser decrease of inflation for targeters as compared to non-targeters during the crisis. As control variables, we introduce - on average for 2003/ : GDP growth, as it is recognized that countries which grow faster tend to face higher inflation pressure; M2 aggregate 10 in percentage of GDP, as a measure of liquidity; Inflation volatility, approximated by the standard deviation of annual inflation rate for the 2003/2007 period - inflation volatility can be considered as an indicator of domestic economy stability -; Imports in percentage of GDP, in order to control for imported inflation; Credit provided by banking system in percentage of GDP, used as proxy of financial development; Government budget balance captures countries fiscal discipline. 9 Philippines is the targeter with the lower average GDP per capita in 2003/2007, USD. So we drop from our control group countries with average GDP less than 1110 USD. See Appendix 2 for basic sample countries list. 10 It could be better to use M3 which is certainly a more complete measure of liquidity, but this variable is much less available especially for developing countries. 12

13 The last two variables are expected to be positively correlated to the dependent variable and the others negatively. We also introduce high income dummy which is equal to 1 if the country is a high income country and 0 otherwise. Graph 4.a gives the evolution of inflation rate for the two groups during the last decade. It clearly appears that on average inflation is lower for targeters. The graph also confirms our hypothesis that targeters interred the crisis with better conditions - with lower inflation rate. However in 2009, there is no significant difference between the two groups. Graph 4 4.a : Inflation rate 4.b : Real interest rate Source : author calculations Table 1 presents our estimation results. As it can be seen in column (1), the coefficient associated to IT is significant with the expected sign. When controlling for initial conditions in column (2) introducing the pre-crisis inflation rate as control variable -, the effect of IT vanishes. The variable inflation pre exhibits significant and negative coefficient in almost all the regressions, suggesting that countries with higher inflation rate in the pre-crisis period are those which faced higher drop in inflation during the crisis. To summarize the results concerning inflation, there is no significant difference between targeters and non-targeters during the crisis 11. Among the control variables, only M2 in percentage of GDP coefficient is significant and negative as expected; meaning 11 Note that inflation volatility seems to be significantly lower for targeters in the last decade and especially in See graph in appendix 3. 13

14 that countries with more liquidity before the crisis were the ones which faced the greater inflation decrease during the crisis. Table 1: Testing the impact of inflation targeting on change in inflation during the crisis Dependent variable : change in inflation rate (difference between 2009 and 2003/2007) (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) IT 1.686** (1.999) (1.018) (0.747) (0.671) (0.991) (1.048) (1.189) (1.485) (1.089) (-0.306) GDP growth (-1.430) Gov budget balance (0.600) M2 (%GDP) * * (-1.953) (-1.829) Inflation volatility (0.0328) (0.274) Dom credit by BS (-0.687) High income (-1.604) Imports (%GDP) (0.0492) Inflation pre ** ** * ** ** ** ** * (-2.278) (-2.054) (-1.859) (-2.440) (-1.083) (-2.305) (-2.355) (-2.285) (-1.914) Constant *** * (-3.154) (0.619) (1.177) (0.617) (1.529) (0.569) (0.886) (1.090) (0.534) (1.943) Nbr of observations Inflation targeters Adjusted R-squared Note: Robust t-statistics in parentheses *, **, *** indicate the statistical significance at 10%, 5% and 1% respectively. What about real interest rate? Using real interest rate here can be debatable, so let us explain very briefly this choice. If one wants to assess how aggressive central banks react during the crisis, and precisely whether targeters central banks lower their interest rate more than non-targeters, the use of short term interest rates central banks reference rates - is likely more appropriated. But here, we are interested not only on this reference rate, but above all, on how central banks decisions related to their rates affect the cost of funding for borrowers on the real economy ie the lending real interest rate. By using this variable, we also assess at some extent the credibility of central banks, since less credible central banks decisions will have less repercussions on the real rates in a context of economic shock and uncertainties. During the crisis, rising uncertainties was one of the greatest concerns, especially on the financial market. In that context, central banks 14

15 credibility is crucial for measures taken to reassure markets and investors 12. One can assume that, the more central banks will credibly intervene during the crisis by lowering its rate the more the financial markets will follow by lowering theirs 13. Graph 4.b shows the average evolution of real interest rate for the two groups. As already mentioned, during the economic boom, targeters have had on average higher real rate than non-taregeters - from 2003 to In contrast, it is clear that during the crisis, the real interest rate raised more sharply for non-targeters For estimations, we run equation (2) where Y is now real interest rate. β is expected to be negative expressing lower increase of the real interest rate for targeters. As control variables, we considered on average for the 2003/2007 period : Inflation rate, Credit provided by banking system in percentage of GDP and a proxy for Bank competition bank concentration The first variable is expected to be positively correlated to the dependant variable, since increase in inflation rate is associated to lower real interest in the pre-crisis period and higher change in real interest rate during the crisis. The second and third variables are expected to be negatively correlated to the dependant variable, since financial development could have help to cope with the crisis and bank competition to contain the rising interest rate in the banking system. We also use as control variable, an indicator measuring the Financing via international markets in 2007 in percentage of GDP. This variable is expected to be negatively correlated to the dependant variable, since higher financing via international market could have reduced the domestic real interest rate before the crisis. Finally, and maybe more importantly, we control for both Inflation and Risk premium on lending during the crisis in order to control for uncertainties on the financial market. This allows us to capture more precisely the effect 12 It is obvious that governments also have decisive role to play, for example in sustaining aggregate demand. 13 Note that for our sample, data shows that inflation targeters central banks lower their nominal rate by about 2% on average more than non-targeters between 2008 and We also run regressions with Money market rates - from IMF International Financial Statistics and find that IT is associated to lower nominal interest rate. These results can be provided upon request. 14 On average, real interest rate raised from 4% in 2003/2008 to 11.5% in 2009 for non-targeters. For the same periods, the rate was 6.7% and 7% for targeters. 15 Data are from Beck et al (2009) database on financial structure. 15

16 of monetary policy strategy on the real rates. The estimations results are presented in Table 2. In column (1), the IT variable exhibits a negative and significant coefficient as expected. Controlling for differences in initial conditions in column (2), the coefficient is still significant but expresses a lower effect - the magnitude goes from 6.26 to , suggesting an overvaluation of the impact of IT in the first regression. Seven out of the eight regressions reveal a significant effect of inflation targeting during the crisis, with a coefficient averaging In other words, during the crisis real interest rate increased by about 4.3% more for non-targeters as compared to targeters. The only regression - column (4) - where IT does not have any effect is the one using credit provided by banking sector as control variable. One can assume that this variable is a transmission channel. Overall, even if there is no significant difference between targeters and non-targeters for inflation rate during the crisis, targeters clearly exhibit lower increase in real interest rate, suggesting at some extent the success of central banks in managing the crisis. Table 2: Testing the impact of inflation targeting on change in real interest rate during the crisis Dependent variable : change in real interest rate (difference between 2009 and 2003/2007) (1) (2) (3) (4) (5) (6) (7) (8) (9) IT *** *** ** ** * *** *** ** (-3.357) (-2.763) (-2.115) (-0.935) (-2.267) (-1.966) (-2.777) (-3.145) (-2.133) Inflation 0.626* 1.002** (1.908) (2.399) Dom credit BS *** (-3.379) (-0.863) Fcing int markets (-1.424) High income (-0.987) Risk premium 1.836*** 1.750*** (4.146) (6.090) Inflation (-0.833) Bank competition (1.591) Real interest rate pre ** ** *** *** ** *** ** *** (-2.529) (-2.315) (-2.809) (-2.695) (-2.547) (-5.111) (-2.185) (-7.369) Constant 6.185*** 10.19*** 6.874* 14.99*** 12.85*** 11.00*** 4.344*** (3.733) (4.056) (1.812) (4.590) (3.818) (3.822) (3.568) (1.086) (0.429) Nbr observations Inflation targeters Ajdujted R Note: Robust t-statistics in parentheses *, **, *** indicate the statistical significance at 10%, 5% and 1% respectively 16

17 Difference in GDP growth Now, facing the crisis, we would like to compare inflation targeters to non-targeters in more general perspectives. For this purpose, it seems appropriated to compare the two groups on the differences in GDP growth during and before the crisis. Equation (2) is now considered with Y representing GDP growth. β is expected to be positive, suggesting a lesser decrease in GDP growth for inflation targeters. The basic sample consists of 114 countries including 21 inflation targeters. As controls variables, we first consider the most mentioned in the recent literature on the 2008/2009 crisis see for example Blanchard et al (2010), Tsangarides (2010), Carvalho (2011). Generated in mean for the 2003/2007 period, the first set of control variables consist of: Trade openness, measured as the sum of exports and imports in percentage of GDP. The more the country is opened to the rest of the world, the more it will be affected by international shocks like the 2008/2009 crisis. Real effective exchange rate volatility, approximated by the standard deviation of annual real effective exchange rate for the 2003/2007 period. This variable captures the stability of countries currency on the international market. Countries with more volatile real effective exchange rate could be more affected by an international shock. Inflation volatility, approximated as explained above. More volatile economies will be less able to manage shocks like those generated by the recent crisis. Short-term external debt in percentage of GDP is obtained by the ratio of total short term external debt to countries GDP, both in US dollars. This variable captures the countries financial exposition as pointed by Blanchard et al (2010). The larger the short-term debt in the initial period, the larger the country could be affected by the adverse shift in capital flows during the crisis, since larger current account deficit requires more capital flows. Exchange reserves expressed in month of imports. As we already discussed, exchange reserve can have a crucial role when countries are facing an international shock. The last variable is expected to be positively correlated to the dependent variable while the others negatively. We also control in our regressions for: Real effective exchange rate in 2007, to control for the relative currency appreciation before the crisis. Exports in percentage of GDP on average for the 2003/2007 period - are used as another measure of trade openness. It can however, at some extent capture the countries dependence vis-à-vis of partners economy health. Indeed, the more partners economies will be affected, the more exports to these 17

18 economies will decrease. GDP per capita on average in 2003/ is used as an indicator of countries economic development. Developing countries are expected to be less affected by the recent financial crisis. The Exchange rate regime, represented by a dummy variable taking the value of 1 if the regime is flexible and 0 otherwise we use the de facto classification in Atish et al (2010). This variable is crucial for our analysis. At least in theory, adoption of inflation targeting is associated to flexible exchange rate regime. However, there are countries with flexible exchange rate, but which are not inflation targeters. Moreover, exchange rate regime could have made a difference on how countries dealt with the crisis. Taking this into account, we need to control for the exchange rate regime in all our regressions to make sure that the coefficient β captures the real and only the effect of inflation targeting. The three first variables are expected to be negatively correlated to the dependent variable and the last positively. As it appears on Graph 5, on average GDP growth was lower for targeters during the precrisis period. The crisis period does not exhibit a significant difference between the two groups. Table 3 presents our estimations results. In column (1) we run a basic regression controlling neither for GDP growth in the pre-crisis period, nor for exchange rate regime. As we argued, we need to control for initial conditions and exchange rate regime to make sure that the coefficient associated to the IT dummy is not biased. This is done in column (2) up to (11). Inflation targeting is not significant in any of the eleven regressions we ran 16. Concerning the control variables, coefficients associated to trade openness, exchange reserves, short term external debt and GDP per capita are significant with the expected sign, suggesting that these variables have been determinant in how countries have been affected by the crisis. It seems to be no effect of exchange rate regime. This finding confirms the conclusions of Tsangarides (2010). 16 Note that we ran the eleven regressions without controlling for exchange rate regime and GDP pre. For three of them, the coefficient associated to IT was significant and positive as one can expect. But, when we introduce the exchange rate regime variable, this effect vanishes, confirming the relevance to control for this variable. These auxiliary regressions can be provided upon request. 18

19 Graph 5: GDP growth for the whole sample (targeters and non-targeters) Source : author calculations Table 3: Testing the impact of inflation targeting on change in GDP growth during the crisis (sample of developed and developing countries) Dependent Variable : change in GDP growth (difference between 2009 and 2003/2007) (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) IT (0.980) (-0.276) (1.002) (0.629) (0.309) (-0.271) (-0.187) (0.234) (-1.004) (-0.540) (-0.544) REER (1) (-1.266) REER volatility (-0.563) Inflation volatility (1.232) Trade openness ** (-2.145) (-0.450) Export (%GDP) (-1.523) Exchange reserves 0.237** (2.176) (1.094) Short T ext debt ** * (-2.517) (-1.892) GDP per capita -8.65e-05** ** (-2.402) (-2.481) ER regime (0.294) (-0.627) (-0.167) (0.650) (-0.203) (-0.217) (0.0590) ( ) (1.507) (0.558) GDP growth pre ** *** *** *** ** ** *** ** *** *** (-2.387) (-3.530) (-3.381) (-3.408) (-2.251) (-2.098) (-2.866) (-2.599) (-3.442) (-3.605) Constant *** *** ** *** * ** *** (-9.630) (-3.284) (0.937) (-2.095) (-3.148) (-1.893) (-2.040) (-2.714) (-0.514) (-1.616) (0.180) Nbr observations Inflation targeters Adjusted R Note: (1) Real effective exchange rate Robust t-statistics in parentheses *, **, *** indicate the statistical significance at 10%, 5% and 1% respectively 19

20 For the whole sample, our regressions reject the hypothesis that inflation targeters outperformed their peers during the recent financial crisis. Since in the literature on inflation targeting empirical analyses sometimes find different conclusions depending on the samples of developed or developing countries, we conducted the same investigation on these subsamples. We use the World Bank countries classification to build our subsamples. Graph 6.a and 6.b give the picture of the differences in GDP growth for developed and developing countries, targeters and non-targeters. As for the whole sample, it is clear that on average, targeters exhibit lower GDP growth during the precrisis period. We can also confirm the idea that developed countries have been more severely hit by the crisis. Developing countries subsample Following the methodology defined to build our control group, the basic subsample of developing countries consists of 68 countries including 8 inflation targeters 17. Running the same regressions as for the whole sample, we find no robust effect of inflation targeting see table 4. Only two out of the eleven regressions show a significant and positive effect of inflation targeting. We cannot rigorously conclude in favour of inflation targeting, given this lack of robustness. Developed countries subsample Using the same approach to determine the control group, the developed countries basic subsample consists of 45 countries including 13 inflation targeters 18. When running our basic regressions and control neither for exchange regime nor for initial GDP growth, inflation targeting dummy shows a significant and positive effect for nine out of the ten regressions. See Table in Appendix 4. However, this effect vanishes when the exchange rate regime dummy variable is introduced. Finally, we reach the same conclusion for developed countries: there is no effect of inflation targeting - See Table As previously, for developing countries, the poorest targeter is the Philippines. 18 The poorest targeter is Poland with average GDP per capita of US dollars. So we drop from our developed control group, countries with average GDP per capita bellow 5305 USD. 20

21 Graph 6: GDP growth samples of developed and developing countries 6.a: Developed countries 6.b: Developing countries Source : author calculations Note that for both Table 4 and 5 control variables are the same as in Table 3 for each column. We choose not to report them for more clarity since we want above all to focus on our variable of interest. Table 4: Testing the impact of inflation targeting on change in GDP growth during the crisis (developing countries subsample) Dependent Variable : change in GDP growth (difference between 2009 and 2003/2007) (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) IT *** 3.130*** (-0.500) (-0.934) (3.548) (3.410) (0.102) (-1.518) (-1.271) (-0.577) (-1.004) (-0.633) (-0.544) ER regime * (0.148) (-1.203) (-1.733) (0.306) (-0.175) (-0.319) (0.302) ( ) (0.534) (0.558) (0.148) GDP growth pre * ** ** *** *** (-1.388) (-1.529) (-1.498) (-1.783) (-1.398) (-1.279) (-2.373) (-2.599) (-2.824) (-3.605) (-1.388) Nbr of observations Inflation targeters Adjusted R Note: Robust t-statistics inter parenthesis *, **, *** indicate the statistical significance at 10%, 5% and 1% respectively To sum up, we investigated the effect of inflation targeting on the economies resilience during the 2008/2009 financial crisis. The results suggest that for the whole sample as well as for subsamples of developed and developing countries, inflation targeting did not make a significant difference during this crisis. 21

22 Table 5: Testing the impact of inflation targeting on change in GDP growth during the crisis (developed countries subsample) Dependent Variable : change in GDP growth (difference between 2009 and 2003/2007) (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) IT 2.848** (2.363) (0.176) (0.142) (0.425) (0.160) (0.237) (0.195) (0.752) (0.410) (0.344) (0.223) ER regime ** (1.249) (0.570) (2.115) (1.285) (1.003) (1.118) (0.111) (0.780) (0.800) (0.756) GDP growth pre ** *** *** *** ** ** *** ** ** *** (-2.508) (-7.928) (-4.923) (-4.614) (-2.232) (-2.171) (-3.560) (-2.202) (-2.451) (-4.049) Nbr of observations Inflation targeters Adjusted R Note: Robust t-statistics inter parenthesis *, **, *** indicate the statistical significance at 10%, 5% and 1% respectively Investigating conditional effects of inflation targeting One can argue that inflation targeting could have the expected effect depending on some characteristics peculiar to economies implementing this monetary strategy. In other words, the impact of inflation targeting can be nonlinear or conditional. We test this hypothesis by analysing different types of conditionality: The inflation targeting effect could depend on the degree of countries economic globalisation. The most economically globalized countries are the most affected by the crisis and can be the ones on which the impact of inflation targeting will be more perceptible. We use the KOF index of economic globalization - Dreher (2011). The inflation targeting effect could depend on the countries indebtedness. The idea is the same as previously. We use the government total debt as a share of GDP. The inflation targeting effect could depend on the relative importance of exchange reserves. We argued that exchange reserves could have helped countries to deal with the crisis. By doing so, it can favour the positive effect expected from inflation targeting. The inflation targeting effect could depend on the countries financial development, as the literature supports that sufficient financial development is necessary for the effectiveness of this monetary strategy. Domestic credit provided by the banking system in percentage of GDP is used as a proxy of financial development. 22

23 The inflation targeting could depend on the degree of trade openness as we show that the most open economies have been the most affected by the crisis. To test these nonlinearities, we use the interactive variables approach. Each of the previously cited variables 19 is interacted with the inflation targeting dummy. Results are presented in Appendix None of the IT coefficients is significant. Coefficients associated to our interacted variables are significant for two of them. The interaction between IT and exchange reserves suggests that the positive effect of exchange rate reserves is more important for targeters. The more the reserves, the more inflation targeters did better during the crisis. Conversely, the coefficient associated the interaction between IT and trade openness is negative suggesting that inflation targeting exacerbates the negative effect of trade openness during the crisis. So far we have investigated linear and nonlinear - or conditional - effects of inflation targeting on change in GDP growth during the crisis and found no significant impact. This finding can raise some interrogations since, as we discuss in section II, targeters seem to have entered the crisis with some better structural economic conditions less government and short term external debt, better current account balance, lower inflation rate, higher and less volatile interest rate. In the next subsection, we briefly discuss this issue. Why did targeters fail to perform better? The first point that can be made in explaining our somewhat disappointing results is about the scale of the crisis. When facing a crisis with such effects, countries are affected regardless of different strategies implemented in their economies. Tsangarides (2010) empirically studies the difference between countries with flexible and fix exchange rate regime during the recent financial crisis. He concludes that flexible exchange rate, which was expected to do better since the literature supports that flexibility can be a shock absorber -, did not make any difference. The shock has been sizeable and indifferently affects the most developed and integrated countries. 19 These variables are generated in mean for the 2003/2007 period. 20 We ran two types of regressions: first without any additional control variables, and second with some additional controls. Conclusions are the same. The other regressions can be provided upon request. 23