How Good are Good Transitions For Growth and Poverty? Indonesia since Suharto, For Instance

Transcription

1 How Good are Good Transitions For Growth and Poverty? Indonesia since Suharto, For Instance Lant Pritchett Harvard Kennedy School September 24, 2010 Presented at the Indonesia Update, ANU Abstract. On May 22 nd 1998 no one knew what to expect about the future of Indonesia over any horizon, from days to years to decades. The previous day Suharto has resigned after over 30 years in power and all bets were off. From the vantage point of 2010 we know what did happen politically, elections were held in June of 1999 as promised and democracy has proved more robust in Indonesia than many have feared and, after the re-election of Yudhoyono in 2009 one can say that May 21 st 1998 began a transition to electoral democracy. In this paper I examine how other outcomes, particularly economic outcomes such as GDP growth and poverty, but also governance outcomes have evolved in this period of democracy. Overall, I argue that outcomes are roughly as good as should have been expected not that outcomes were terrifically better than in the pretransition period, but because the empirical data from other transitions should have warned against over-optimistic predictions on how much or how rapidly transitions to electoral democracy, a positive transition in its own right, should have created positive transitions in growth, in poverty reduction, or even in governance in practice. Preliminary and Incomplete: 1 9/28/2010

2 How Good are Good Transitions? Indonesia, For Instance Introduction Benchmarking is an essential component of any performance assessment. While the world of high finance is hardly the place to draw positive lesson just these days, the rewards to portfolio managers give a simple and clear example. People who manage allocated a portfolio within a certain asset class, say US bonds, or European equities, or emerging market bonds, or small cap equities or so on, are rewarded based on how well their investments do relative to an index of the entire class of assets. Although this compensation scheme has its problems, it makes sense as you cannot ask the person allocating your portfolio among various US municipal bond options to make the same returns as the person managing the value equity makes, as this holds the portfolio manager responsible for events beyond their control, like whether stocks outperformed bonds, not the results of their own actions or decisions. After the crisis, Indonesia did not return to the very rapid rates of economic growth that prevailed for over 30 years during the Suharto regime. Growth, which was 5.9 ppa from 1987 to 1997 was only 3.7 ppa from the nadir of the crisis in 1999 to Growth was slower by (roughly) 2.2. ppa. Is that a disappointment? Are those growth rates an indictment of the policies or economic management of the democratic governments? How one assesses performance depends on the benchmark. I argue that Indonesia past growth is not an appropriate benchmark for assessing the growth performance of Indonesia 1999 to 2008 as it ignores two key facts relevant to benchmarking performance. Preliminary and Incomplete: 2 9/28/2010

3 First, the literature on economic growth has documented that economic growth is episodic rather than steady and that there is very powerful regression to the mean in economic growth rates so that economies that have growth fast in the past are expected to slow down (not necessarily have slower than average growth, but to be slower than their previous pace). If one takes as a benchmark the expected rate of growth based on crossnational estimates of the magnitude of regression to the mean then, with just this factor alone, Indonesia performs right at the benchmark. Second, it is likely that a major political transition affects the rate of economic growth through a variety of channels. In analysis original to this paper I identify all country episodes of rapid democratization and examine the growth rates ten years before and ten years after this political transition. For countries that went into growth rates with an above average growth rate there appears to be a democratic transition affect that slows post-transition growth by about 2 ppa. Again, this affect alone could fully account for Indonesia s slower growth. The combination of the two effects implies that if the benchmark for growth 1999 to 2008 is countries with rapid growth up to 1997 and a large democratizing transition then the expected deceleration of growth is between 3.5 ppa and 4.3 ppa. By this benchmark Indonesia s actual post-transition growth rate of 3.7 ppa and deceleration of only 2.2 relative to the pre-crisis rates of 5.9 ppa are actually a substantial better performance than realistic benchmarks based on the performance of other countries. Using the same type of benchmarking exercise I also look at how Indonesia has done on various aspects of governance since the political transition. In that domain there are mixed messages. There has been no progress in bureaucratic capability but Preliminary and Incomplete: 3 9/28/2010

4 it is not clear from the benchmarking analysis that much progress should have been expected. On the other hand, there is more progress on the control of corruption than would have been expected. Finally, on poverty, although the headcount of consumption expenditure has declined, not nearly as fast as might have been expected. On this there is less ability to tell, either from theory or empirics, what should have been expected from democracy. Preliminary and Incomplete: 4 9/28/2010

5 I. Democratic Transitions and Economic Growth It is tough to make predictions, Especially about the future. Yogi Berra (Baseball player) As the famous New York Yankee and noted savant pointed out, it is much easier to predict the past than the future. I therefore start with the easy part, reviewing Indonesia s growth performance before, during, and after the economic crisis and political transition of 1997/98. But even about the past, the much trickier question is, is the growth performance since the crisis and political transition better or worse that was expected? That requires some statement of what was expected should we have expected Indonesia s growth to return to the same pace as before the crisis? Should we have expected growth under democracy to have accelerated, which we might have expected if a democratic government could provide better or more stable economic policy. Should we have expected growth to decelerate, which we might have expected if democracy has a difficult time in providing the impetus for rapid growth, given the conflicting pressures. I.A) Indonesia growth performance, in pictures Figure 1 shows five periods of Indonesia s growth in GDP per capita since 1960 (which is where the standard available data series begin). From 1960 to 1967 there was essentially zero growth, a lack of growth was accompanied by other economic disruption, such as shortages and rapid inflation. Moreover, it is worth remembering that in this period Indonesia was a quite poor country. For instance, according to the purchasing power adjusted estimates from the Penn World Tables 6.3 Indonesia in the early 1960s Preliminary and Incomplete: 5 9/28/2010

6 had per capita GDP lower than that of Mali or Ethiopia today, Indonesia s GDP per capita in the 1960s was that of Somalia in the 1970s before its descent into chaos. Even countries thought of as very poor today, such as Bangladesh or Nepal currently have twice the GDP per capita that Indonesia had in the 1960s. After the political chaos and violence that occasioned the beginning of the New Order government under Suharto the economy began to grow. From 1967 to 1987 growth was 4.8 ppa (percent per annum). This put Indonesia among the growth stars of the time. If one takes the period from 1987 to 1997 growth accelerated to an even more rapid pace, with Indonesia experiencing nearly 6 ppa growth. The financial crisis that began in Thailand and spread to other East Asian countries resulted in a particularly severe crisis for Indonesia. The Rupiah collapsed (before any inflation), most of the banking sector was illiquid and insolvent, and new investment ground to a halt. This was followed by inflation, a rapid rise in rice prices, a huge rise in poverty (see section on poverty below). However, somewhat amazingly to those of us living through it, by 1999 the crisis was stemmed and the economy stabilized at a much lower level. On June 7, 1999 the first post-new Order national elections were held and the democratic transition moved along. From 1999 to 2008 the economy has grown at a pace of 3.7 ppa. I only take the data through to 2008 primarily because this is what is available on a standardized basis for comparisons with other countries which I use below. This means I am not attempting an update on GDP per capita and in particularly on how the global crisis, this time caused by the rich countries, affected Indonesia--but a retrospective on the decade following the transition. Indonesia took until 2004 to recover to the 1997 pre-crisis level, Preliminary and Incomplete: 6 9/28/2010

7 which seemed, at the time, slow relative to the pace of recovery East Asian countries. For instance, Korea regained its 1997 peak already by 1999 after a severe but short jolt, Thailand had recovered by 2002 (before the onset of the more recent troubles there). By 2008 Korea GDPPC was 47 percent above its 1997 level, Thailand s 26 percent, while Indonesia s only 19 percent. In 2008 GDP per capita was almost exactly 40 percent higher than at the nadir of the crisis (and almost 20 percent higher than the pre-crisis peak). Whether this growth performance has under or over performed expectations depends on what expectations were. One, super-optimistic, scenario would be a quick recovery (a V shaped recovery to the 1997 level in two years) plus a resumption of the rapid pre-crisis growth rate of 5.9 percent. In that scenario output would have increased Preliminary and Incomplete: 7 9/28/2010

8 nearly 75 percent in the ten years from 1999 to Another possible scenario would be that the economy would start back again on the rapid growth path of 87 to 97 (but no particularly rapid crisis recovery). But 1987 to 1997 growth was rapid even by long-term Indonesia averages, so another would be just to return to the 1967 to 1987 growth rate. This exceeds the actual, but not by a tremendous amount. Moreover, in the four years from 2004 to 2008 growth was actually at the same pace as (4.5 ppa versus 4.8 ppa). Finally, another scenario would be that after the crisis Indonesia would grow at an average pace, so Figure 2 also shows the median growth of GDP per capita of all other countries in Indonesia s income range. Compared to that standard Indonesia did well, growing at 3.7 ppa compared to the median country growth of 2.9 ppa, but few countries strive to have merely ordinary growth performance. Preliminary and Incomplete: 8 9/28/2010

9 I.B) What is the expected outcome of a big and good political transition The scenarios in Figure 2 are entirely mechanical and really do not address the question of whether growth was as expected or whether growth in the democratic posttransition era has been disappointing. Having lived in Indonesia from 1998 to 2000 I can reassure you the expected growth rate after the crisis was not a well-defined concept. No one knew what the trajectory of the other events, social, political and policy, was going to be. Under many of the political and social outcomes that seemed possible, if not probable, in 1998 military intervention, increased separatism and unrest, ethnic violence, unstable democracy, various stripes of populism the growth expectations were very negative. The more interesting, if only because it is tractable, question is, if you had known the political and social trajectory that has actually happened which, relative to the more pessimistic expectations and downside scenarios at the time is enormously positive what would have been your growth expectations? If you had been reassured in 1998 by a visitor from the future that national elections would be held peacefully in June 1999 and that these would be reasonably free and fair, that the party that had controlled Indonesia for 30 years would peacefully cede power to the elected government, that Indonesia would then have ordered democratic transitions from June 1999 onwards. Moreover, let s even assume you knew in 1998 who the elected leaders would be: Gus Dur, Megawati, Yudohyono (all well known political figures and none of them among the more feared outcomes of democracy). What would have been growth expectations Preliminary and Incomplete: 9 9/28/2010

10 conditional on knowing that the political transition from 30 years of authoritarian oneparty rule had gone so smoothly? It might not seem unreasonable to expect that this good political transition to democracy would be accompanied by good outcomes on the economic front: that growth and poverty reduction would be as rapid, or perhaps even more rapid, than during the Suharto years. It is in this light that Indonesia s post-crisis, post-transition growth could be seen as disappointing. However, is this really what the experience and data from other countries around the world would really lead us to expect? Have other countries that have had very rapid transitions to democracy experienced more or less rapid growth after their political transition. This question I can answer. The concept of governance is hugely complicated, but can be divided into two conceptually distinct components, the polity how political leadership arrives to power and the second is administrative performance of the government itself (to which we return in the next section). One commonly used indicator of the polity is a measure called, unsurprisingly, POLITY which ranks countries on a scale of autocracy from zero to negative 10 and democracy from zero to positive 10. The simple sum of those two indicators gives an empirical ranking that ranges from negative 10 (completely autocratic) to positive 10 (completely democratic). Just to illustrate the range of the ranking selected East Asian examples are given in Table 1. As can be seen this rating has nothing intrinsic to do with economic performance (both China and Vietnam are rated as -7) not even with quality of governance in the sense of administrative capability to Preliminary and Incomplete: 10 9/28/2010

11 carry out governmental functions (Singapore is a -2 and Malaysia in 2007 a 3 while both have very capable administrations). Table 1: Selected East Asian examples of the POLITY IV rankings, from -10 (autocracy) to +10 (Democracy) Country Year POLITY Indonesia China Vietnam Singapore Malaysia Philippines Indonesia Japan This POLITY ranking captures Indonesia s massive, rapid, and sustained transition to democracy. Figure 3 shows Indonesia s POLITY ranking from 1960 to Obviously to Indonesia experts this adds no new information, but does show the ranking follows the known major events of Indonesia s political trajectory in a not unreasonable way the Sukarno era was rated as -5, during the entire Suharto period from 1967 to 1997 was rated a -7, improving dramatically to a positive 6 in 1999 with the June elections and then increasing to 8 with the election of 2004 that brought in SBY as president. Preliminary and Incomplete: 11 9/28/2010

12 Figure 3: Indonesia s political transition to democracy as rated in the POLITY data POLITY Raanking Year Source: POLITY data from WDR 2011 data. Knowing that Indonesia has been a democracy since 1999 however does very little to help in what one should have expected from its growth rate. Figure 4 is the simple display of countries average POLITY score and their growth rate over the period (each country is identified with a three letter code and Indonesia is highlighted inside the circle). The median growth rate of the imperfect democracies in this period (defined as those countries with an average POLITY score above 5 but less than 10) is 2.95 ppa, which puts Indonesia s growth rate over this period (in this data set) of 3.3 ppa pretty squarely in the middle of growth performance. But, there is really no strong association between growth and POLITY score, the autocracies do slightly better, 4.3 ppa, the countries in between do worse, 1.39 ppa but mostly there are countries at a large range of economic growth in each POLITY category 1. 1 There is a substantial literature by economists on the question of whether democracies tend to have higher or lower growth rates than non-democracies. The current conventional wisdom is that there is very little connection between the average growth of democracies and non-democracies as distinct groups but that there is a much higher variance in growth rates among non-democracies, with really rapid growth (e.g. China, Vietnam, Indonesia under Suharto) but also very slow growth and even growth implosions. That said, one can find specifications in which democracy appears to matter. For instance, Persson and Preliminary and Incomplete: 12 9/28/2010

13 The simple comparison democracies grow faster or slower than non-democracies does not capture the possibility that transitions themselves might have impacts. In this case while in the long-run democracies might be capable of sustaining rapid growth the transition period itself creates an adjustment period of slow growth. To examine this question we need to compare growth before and after rapid, large, political transitions from autocracy to democracy, for which we need to identify large transitions. First, I searched over the POLITY combined democracy indicator (sum of autocracy and democracy) to identify all instances in which the POLITY index changed by more Tabellini (2006) regress annual growth rates on a dummy variable that =1 if the POLITY score is above zero and 0 if POLITY is below zero and show this is associated with more rapid growth. This is a particular functional form of a step function in which an improvement in democracy from Malaysia s 2006 value of 4 to Netherlands at 10 would have no effect and the reduction in autocracy in Jordan from -9 to -2 should have no impact, but moving across exactly zero should have all of the impact. Preliminary and Incomplete: 13 9/28/2010

14 than 5 units in a single year (say from -7 to -1 or from 2 to 8). These are candidates for large democratic transition. Then there is a decision tree to classify and time to the transition, especially to cope with countries with multiple transitions that is described in Table 2, with examples (starting with Indonesia). The results of this classification scheme are 52 episodes of large democratic transitions (see Appendix Table A.1 for the complete list). Table 2: Classification of candidate large democratic transitions (all years with a greater than 5 point absolute change in POLITY rating) Empirical criteria Decision Rule Country example If country has only one candidate episode If the country has two episodes Keep that episode with transition dated to year of new value Indonesia 1998=-6, 1999=5, Δ=11, transition year is 1999 episodes are more than 3 years apart if episodes are closer than 3 years and in the same direction if episodes are closer than 3 years and in the opposite direction Keep both episodes Taiwan, 1986=-7, 1987=-1, 1991= =7, two transitions 1987 and 1992 Classify as one episode, dated to the earlier period Madagascar 1990=-6, 1991=2, 1992=9, one transition dated to 1992 Do not keep as an episode Armenia, 1995=3, 1996=-6, 1997=-6, 1998=5 (no sustained democratic transition, brief cycle into autocracy) If a country has three or more episodes then each was classified separately depending on the timing and direction of changes (20 countries) to keep and date distinct episodes. Once the episodes of large democratic transitions had been identified the next step is calculate growth rates before and after the episode. But as we want to capture the medium term we calculate the 10 year growth rate up to 3 years before the transition and the growth rate from one year after the transition for 10 years (or until the data ends). Preliminary and Incomplete: 14 9/28/2010

15 Figure 5 illustrates the timing of the growth episodes, using Indonesia s transition in 1999 as an example year growth three years pretransition (e.g (up to) 10 year growth post transition (e.g =3.3 ppa) T-3-10 (1986) T-3 (1996) T (1999) T+1 (2000) T+1+10 (2007) Figure 5: Illustration of the calculation of the pre and post large democratic transition growth rates (with example dates from Indonesia) Table 3 gives the first key result of the empirical analysis of the transitions, which is that countries that have experienced large democratic transitions beginning from a preceding period of above average growth (higher than the cross-national average of 2 ppa) experienced sharp decelerations of growth rates in the 10 years following the (year after) the transition. So, according to the PWT6.3 data on the growth rate of purchasing power parity adjusted real GDP per capita, Indonesia s growth rate from 1986 to These timing assumptions are not innocuous. Often a political transition is preceded by a large fall in GDP per capita, sometimes as the result of the chaos of the political transition itself. If one then calculates the growth before the transition to include this fall, which could be the result of the transition itself, then it will look like the political transition accelerated growth. This is why I go back some years before the transition so that the pure disruption effects are not counted as part of the pre-democratic period. Rodrik and Wacziag (2005) for instance do an analysis similar to examine growth impacts of democratic transitions and date the transitions to just before the transition. They find similar results overall, of the nine countries they identify with democratizing transitions begun from above 2 ppa growth the average deceleration was 3.53 ppa which is exactly what we find in Table 3. But in some instances the differences in timing produce clearly different country results. For instance, we both find a democratic transition in the Philippines dated to Since GDP fell from 1984 to 1986 by their dating growth in the Philippines accelerated whereas my timing compares growth of 1974 to 1984 and finds growth decelerated. Preliminary and Incomplete: 15 9/28/2010

16 was 5.54 ppa, but over the period 2000 to 2007 was only 3.28 ppa. This is the growth deceleration observed in the analysis above, with growth 2.2 ppa slower in the democratic, post-suharto era than in the authoritarian period. This, first-cut, analysis suggests Indonesia has had less deceleration of economic growth than Indonesia should have expected if expectations were grounded in the experience of other countries. First, of the 22 country episodes of large democratic transition from above average growth all but one experienced a growth deceleration (Korea 1987 is the sole exception with an acceleration of only.22 ppa). The combination of high initial growth and democratic transition makes some deceleration all but inevitable. Second, the magnitude of the decelerations experiences was very large: the median deceleration across these 22 countries was 2.99 ppa and the average deceleration 3.5 ppa. So if Indonesia had experience the typical (median) or average (mean) deceleration of this group of countries its growth rate would have been 2.55 ppa (5.54 less 2.99) or 2.01 ppa (5.54 less 3.53) rather than the 3.3 ppa growth it actually experienced. So, while one framing is that Indonesian growth has been disappointing in the democratic era, this is only relative to an arbitrary expectation that it would maintain the same growth rate. But there is no evidence to suggest that zero deceleration is a reasonable expectation for post democratic transition growth. Preliminary and Incomplete: 16 9/28/2010

17 Table 3: Countries with large democratic transitions starting from above average growth (higher than 2 percent) before the transition Country Year of large democratic transitions Magnitude of POLITY increase Growth pretransition (T to T-3) Growth posttransition (T+1 to T+10 (or end of data) Change in growth pre and post transition (sorted) GRC % 0.02% -7.17% IRN % 0.11% -7.01% PRT % 1.48% -5.63% TWN % 3.95% -2.52% TWN % 5.78% -0.64% NGA % -2.44% -8.25% ECU % -1.66% -7.36% COG % 0.57% -5.11% IDN % 3.28% -2.26% DOM % 1.35% -4.14% KOR % 5.57% 0.22% THA % 0.82% -3.85% MNG % 2.09% -2.30% BGR % -0.10% -4.12% PAN % 1.68% -2.23% BEN % 1.30% -2.32% PAK % 1.32% -2.18% URY % 3.16% -0.27% BRA % -0.34% -3.65% PRY % -0.75% -3.45% BOL % 0.27% -2.09% ROM % 0.85% -1.28% Median, high initial growth transitions 5.01% 1.08% -2.99% Average, high initial growth transitions 4.82% 1.29% -3.53% Source: Author s calculations with POLITY and PWT6.3 data. The data in Table 3 does not suggest that democracy or democratic transitions necessarily reduce growth uniformly as this only analyzed those countries which entered their transition from rapid growth. If we take the polar extreme of countries that entered an episode of democratic transition from negative growth rates in the previous decade Preliminary and Incomplete: 17 9/28/2010

18 then growth accelerated substantially after the transition. The negative growers in the period leading up to the POLITY change actually accelerated by 3.3 ppa (median) to 3.7 ppa (mean) in the ten years after their large democratic transition. Actually, the average growth change of all democratizing episodes is only -.3 ppa, a very modest deceleration (consistent with Rodrik and Wacziag 2005). Table 4: Countries with large democratic transitions starting from negative growth before the transition Country Year of large democratic transitions Magnitude of POLITY increase Growth pretransition (T-10-3 to T-3) Growth posttransition (T+1 to T+10 (or end of data) Change in growth pre and post transition (sorted) FJI % 1.43% 1.99% MWI % 0.96% 1.80% MOZ % 6.58% 7.54% ZAR % -7.17% -6.10% POL % 4.75% 5.84% NIC % 1.03% 2.31% CAF % -0.60% 0.79% SLE % 4.59% 6.16% ETH % 1.93% 4.05% ZMB % -1.54% 1.10% HRV % 5.07% 8.28% LBR % 5.75% 9.73% DJI % -2.29% 2.65% GUY % 1.05% 6.45% Median, low growth -1.48% 1.24% 3.35% Mean, low growth -2.22% 1.54% 3.76% The data say the countries that had above average growth before their large democratic transition had very substantial growth decelerations in the medium run (10 years) after that transition. There are two immediate questions about whether this has Preliminary and Incomplete: 18 9/28/2010

19 anything to do with transitions at all, and if so whether it had anything to do with democratic transitions. The first issue is that, contrary to popular narratives in which growth is a persistent feature of countries, in fact growth rates are very volatile and exhibit very little persistence 3. There is strong regression to the mean so that countries with rapid growth in one decade are predicted to have slower growth in the next decade still faster than average, but slower than their previous pace. This means that part of the deceleration of rapid growth countries (and of the recovery of low growth countries) in Table 3 and Table 4 is just built into the natural lack of growth persistence and the definition of the categories: rapid growth countries are expected to decelerate whether they have a political transition or not and low growth countries are expected to accelerate growth whether they have a transition or not. The third step therefore is to calculate the growth transition of countries with no political transition. This is a little bit tricky because the political transitions of each country is at a different year so comparing before and after for countries with no transition around some arbitrary year (say, 1999) might be misleading as it is affected by secular differences in growth rates. Instead, for each country with no political transition episode I pick a random year and do the same calculation with the timings as illustrated in Figure 5 relative to that randomly chosen year as for the countries with political transitions 4. I can then average these countries in the categories just as for the transition 3 This was originally pointed out by Easterly, et. al. in 1993 and is one of the few stylized facts of the literature on economic growth that has remained robust as more and more data has come in. The cross national correlation of country growth rates across decades has remained quite low (in the.10 to.20 range, depending on the sample). 4 I then iterated this procedure 25 times (so that each iteration each country had a different year around which the growth difference was calculated) and the results reported in table 5 are the average over all iterations. Preliminary and Incomplete: 19 9/28/2010

20 countries. As expected, it really is the case that countries with above average growth are expected to have their growth decelerate and with below zero growth are expected to have their growth accelerate (that is, there is regression to the mean ) even around some arbitrarily chosen year. The average deceleration for countries with growth above 2 ppa is 1.8 ppa, which returns them roughly to the cross-national average. The average acceleration for countries with less than zero growth is larger, 4.6 ppa, which is larger in part because these country s growth rate is further from the cross-national average than the high growth countries (as they are defined as those not just below the average, but below zero). The deceleration of the growth rate of countries that started with above average growth and did have a large democratizing transition of 3.5 ppa is substantially larger than the counter-factual of the growth deceleration of countries with no transition 1.8 ppa. So the difference in difference the change in growth rates before and after (difference) between the democratizing countries and no transition countries (difference) is itself ppa countries with large democratizing transitions, on average, decelerated by 1.76 ppa more than countries with no political transition. So, crudely put, the growth deceleration of rapidly growing economies following a large democratizing transition seems to be about half natural deceleration due to regression the mean and half due to the democratizing episode itself. Preliminary and Incomplete: 20 9/28/2010

21 Table 5: Difference in the growth decelerations/accelerations of countries with episodes of large political transition, positive and negative, versus countries with no episode Growth before the transition (T-13 to T- 3) Large democratizing transitions (POLITY change more than 5, positive) Before After Difference of before and after growth of countries with transition Difference of before and after growth of countries with no large political transition Difference in difference: change in growth of democratizing episode countries versus countries with no episode (III less IV) I II III IV V High (g>2) 4.8% 1.3% -3.5% -1.8% -1.76% Medium (0<g<2) 1.3% 2.0% 0.7% 0.3% 0.43% Negative (g<0) -2.2% 1.5% 3.8% 4.6% -0.87% Source: author s calculations. A second counter-factual is to compare the growth deceleration of countries with large democratizing episodes to countries with large shifts in the POLITY ranking away from democracy and towards autocracy. Exactly the same procedure as described in Table 2 was followed in defining the timing of country episodes of moving away from democracy. The question then is whether countries that began political transitions with above average growth had larger decelerations if that transition was democratizing versus moving away from democracy. (The list of all autocratizing episodes is in Appendix Table A.2). Turns out that the countries with above average growth had much less deceleration after an autocratizing episode than the countries that had a democratizing episode only 1.4 ppa deceleration. The deceleration was about 2.2 ppa less for these countries than for democratizing countries. This does not imply Preliminary and Incomplete: 21 9/28/2010

22 autocratizing episodes helped in some absolute sense as remember the no episode countries decelerated by only 1.8 ppa so the rapidly growing autocratizing countries had only slightly less than the natural deceleration. Preliminary and Incomplete: 22 9/28/2010

23 Table 6: Difference in the growth decelerations/accelerations of countries with episodes of large political transition, positive and negative, versus countries with no episode Growth before the transition (T-13 to T-3) Large democratizing transitions (POLITY change more than 5, positive) Large autocratic transitions (POLITY change more than 5, negative) Before After Change Before After Change Difference between democratic episode and autocratic episode High (g>2) 4.8% 1.3% -3.5% 3.5% 2.2% -1.4% -2.2% Medium (0<g<2) 1.3% 2.0% 0.7% 1.2% 1.8% 0.6% 0.2% Negative (g<0) -2.2% 1.5% 3.8% -1.9% 1.2% 3.1% 0.6% Two final benchmark comparisons are to use simple regressions that associate the change in countries growth rates with their previous growth rate (and level of income). The first simple comparison is to just take all the developing countries (with GDP per capita lower than PPP$10,000 in 19987) and regress the change in growth rates between 1987 to 1997 and 1999 and 2007 for all countries. In this regression Indonesia s predicted deceleration is 1.97 ppa versus the actual of 2.15 ppa. So regression to the mean alone can explain all of Indonesia s growth deceleration. The second simple regression is to regress the change in growth rates for all 52 countries with large democratizing transitions on their initial growth rate (and level of income at the beginning of the post transition growth period). This is the predicted deceleration for those countries which experienced democratic transition. From this the predicted deceleration for Indonesia is 4.26 ppa which combines the regression to the mean and the large democratizing transition. I.C) Summary of the post-transition growth performance Preliminary and Incomplete: 23 9/28/2010

24 Table 7 brings together the main empirical results of my attempt to benchmark Indonesia s growth performance since the economic crisis that began in 1997 and the political transition that began in The naive expectation that growth would return to the previous high levels might imply disappointment with Indonesia growth as growth decelerated by 2.28 ppa. However, any number of a variety of other ways of setting an empirical benchmark for the performance of Indonesia s economy suggest that the economy has actually outperformed what would have been expected as there are two forces leading to a deceleration in growth: first, regression to the mean and second, a large democratizing transition. The simplest possible regression of growth rates on countries growth rates actually predicts growth for Indonesia of 3.47 ppa versus the actual of So typical regression to the mean fully explains Indonesia s growth deceleration. But, the analysis of large democratizing episodes suggests that countries that go into these episodes with above average growth (nearly) always experience significant deceleration of growth. The average deceleration among the 22 countries was 3.53 so the predicted for Indonesia if it experience the average deceleration was growth of only 2.01 ppa so Indonesia outperformed in growth by 1.27 ppa. Preliminary and Incomplete: 24 9/28/2010

25 Finally, a simple regression of post-transition growth rates on all large democratizing transitions and the previous rate of growth actually predicts that Indonesia s growth was predicted to slow to only 1.28 ppa through the combined effects of regression to the mean and democratic transition. This benchmark therefore completely reverses the sense that democracy in Indonesia has been disappointing in its growth performance, relative to the benchmark Indonesia has actually managed to sustain much higher growth than other, reasonable, benchmarks would have predicted. Table 7: Summary of Indonesia s post crisis, post transition growth relative to benchmarks (1999 Predicted Indonesia Over/under Definition of benchmark Actual performance to benchmark Indonesia s pre-crisis growth rate, to ( ) 3.70 ( ) Median, all developing countries, Regression to the mean predicting changes growth versus based on developing countries Countries with large democratic transitions starting with above average (>2 ppa) growth (table 3, predicted= ) Predicted value from regression among 52 countries with large democratizing transitions Source: Author s calculations ( ) 3.47 ( ) 2.01 ( ) 1.28 ( ) 3.29 ( ) 3.29 ( ) 3.28 ( ) 3.28 ( ) II. How fast does democracy improve Governance A distinct question is whether large democratizing transitions tend to improve other aspects of governance measured as the capability of the administration for implementation (as opposed to policy making). This is more difficult empirically as there Preliminary and Incomplete: 25 9/28/2010

26 are fewer widely accepted empirical indicators of governance that have a long enough span of coverage so as to allow before and after analysis. The most plausible set of indicators are those from the International Country Risk Guide (ICRG) which has produced ratings on a number of dimensions including political, economic, and financial risk for use by international business for a large number of countries since I use four of the component indicators from the Political Risk section Democratic Accountability, Bureaucratic Quality, Control of Corruption and Law and Order. These rank countries from a scale of 0 (worst) to 6 (best) except for bureaucratic quality where the scale is only 0 to 4. In this instance I compare the actual value in the year before the transition to the value 10 years following the transition 5. The first component is really just to reassure myself that the comparisons between the two sources, POLITY and ICRG, make sense and hence I examine whether countries with a large democratizing POLITY transition have a reported increase in Democratic Accountability according to ICRG. Indeed, the changes for democratizing transitions are positive and very large while those for autocratic transitions are negative. One interesting question is whether democracies have, on average, more capable administrations, as there if often the argument that authoritarian regimes can make the trains run on time. In the case of Indonesia ICRG rates that there has been neither improvement nor deterioration in bureaucratic quality since 1998 ranked a 2 on a o to 4 scale in both periods. Interestingly, democratizing transitions are not really expected to lead to that much improvement as those countries with transitions had an average improvement of only.4 units larger than countries with no transition at all. 5 Analysis using the pre-transition average to 10 years post-transition gave quite similar results. Preliminary and Incomplete: 26 9/28/2010

27 After the transition, and especially in recent years, there has been an effort to check corruption. According to the ICRG ratings Indonesia has been quite successful as control of corruption has improved from 1.25 to 3.83 a massive increase (more than two crossnational standard deviations). Interestingly, the data does not predict nearly that much improvement only a.32 greater gain for large democratizing transitions than others. Table 8: Democratizing transitions in POLITY and changes in other governance indicators Indicators from the Political Risk section of the International Country Risk Guide Democratic Accountability (Range: 0 to 6) Bureaucratic Quality (Range: 0 to 4) Corruption (Range: 0 to 6) Law and Order (Range: 0 to 6) Indonesia Change 1998 to 2009 Difference in countries with large POLITY transitions and to countries without transitions (before and after 1999) Democratic transitions (POLITY change >5) Autocratic transitions (POLITY change<-5) Cross national std. dev. of variable in A final comparison of interest is law and order. On this score there has been some improvement, but quite modest. Interestingly, on this indicator progress has been actually less than predicted among transition countries, rising by.5 compared to an expected rise among large democratic transitions of 1.08 units. III. Democracy and Poverty A final domain over which one might have expected a democratic transition to be a good transition is poverty reduction. Here one might have thought that a democratic Preliminary and Incomplete: 27 9/28/2010

28 government, motivated by elections and disciplined by democratic accountability (which did, as we saw, actually increase) would be able to engineer larger poverty reductions than during the previous authoritarian regime. That was not so, nor, I would argue, was there any good reason to expect it to be so. Figure 6 shows the evolution of the headcount poverty rate the proportion of the population below an absolute threshold of household per capita consumption expenditures. The series breaks in 1996 as the poverty line calculation was re-done so the first series and second series are comparable to within but not to each other. This shows the extremely rapid reduction of poverty from 1976 to 1996 from 40 percent points of the population to only 11.6 percent in only 20 years. This is one, if not the most rapid reduction in mass destitution in all of history. This is of course interrupted by the crisis, with poverty increasing quite dramatically, then resuming a downward trend, but less rapidly than prior to the crisis and with an actual reversal in 2005 to Figure 6: Evolution of poverty, 1976 to 2010 (with break in series at 1996) Poverty headcount rate Year New Series Old series Preliminary and Incomplete: 28 9/28/2010

29 The question is whether the downward trend good enough news or whether it is disappointing which again depends on what one expected to happen. Here there is insufficient comparable cross-national, time-series data on poverty headcount rates to do the episodic analysis of democratizing episodes that were possible for growth or governance indicators. Instead, I discuss three scenarios to illustrate that part of the puzzle is not just that growth was slower in the democratic period but also the responsiveness of poverty reductions to growth (the poverty elasticity) appears to have fallen. Figure 7 shows the evolution of the poverty headcount rate with the actual data (the new series of Figure 6) and three scenarios. Figure 7: Actual headcount poverty and alternative three scenarios for the path of poverty reduction Poverty headcount rate Yerar Actual At elasticity, actual growth At elasticity, growth Elasticity=2 (int'l average), actual growth The first scenario asks what the evolution of the poverty rate would have been had poverty had the same responsiveness to growth (elasticity) as in the 1976 to 1996 period. Table 9 shows the calculation of the poverty elasticity of growth which is simply the ration of the percentage reduction in the poverty headcount rate to the percentage Preliminary and Incomplete: 29 9/28/2010

30 increase in GDP per capita. Overall during the 1976 to 1996 period (using the old poverty headcount series) the average (median) elasticity was (-1.24). Table 9: Poverty reduction, economic growth, and implied poverty elasticity in the period Year Average Median Headcount poverty rate Percentage Change in headcount poverty rate GDP PC (1976=100) Percentage Change in GDP per capita Implied Growth elasticity (ratio of col. II to col. IV) I II III IV V % % % % % % % % % % % % % % Table 10 shows the same calculation for the period 1996 to 2008, which is more complicated by the huge rise and fall in poverty rates during the crisis. I take two periods as the start to finish elasticities: 1996 (pre-crisis) to 2008 or 2000 (post crisis) to Both give the same result that the elasticity of poverty reduction with respect to growth declined to only around.5 (compared to 1.15 or 1.2). But, as both the table and Figure 7 illustrate, this was mainly the result of the anomalous increase in poverty while GDP per capita was declined during as otherwise the elasticities are quite similar to those 1976 to Preliminary and Incomplete: 30 9/28/2010

31 Table 10: Poverty and economic growth and implied elasticities, 1996 to 2008 Year Headcount poverty rate Percentage Change in headcount poverty rate GDP PC (1976=100) Percentage Change in GDP per capita Implied Growth elasticity (ratio of col. II to col. IV) I II III IV V % % % % % % % % % % % % % % % % % % Average Median End to end, % 35.2% End to End, % 23.10% The second scenario shows what would have happened if there had been the faster growth rate of and the elasticity of the period. In this instance poverty would be lower by 5 percentage points in 2008 than it actually was, which, given Indonesia s population is almost 12 million fewer people in poverty. Finally, even in the 1976 to 1996 period Indonesia s poverty elasticity was lower than might have been expected from international norms, or from the mechanical calculations of these elasticities from assumptions about the distribution of income and the poverty line. The fourth scenario shows what poverty would have been had there been the actual growth rates but a poverty elasticity equal to 2 (a large increase over the previous period). Preliminary and Incomplete: 31 9/28/2010

32 However it is sliced, it appears that poverty declined over the entire period by less than would have been expected even given the lower overall economic growth rate though nearly all of this is because of the sudden rise from and the subsequent lack of the return to the previous trend. This brings us back to the same question: was there any reason to expect that the democratic transition would have led to more rapid poverty reduction? Changes in the standard consumption expenditures based poverty head count rates can be exactly decomposed into three elements: (a) the pure growth effect, shifts in consumption expenditures (net of taxes or transfers) keeping the shape of the distribution across households constant, (b) the real income distributional effect, shifts in the consumption expenditures favorable to the poor due to shifts in the distribution (included in this are shifts in relative prices, such as goods consumed disproportionately by the poor becoming relatively more expensive) and (c) shifts in net transfers which would be the impact of either larger social transfers or improvements in their targeting. We have seen above that there is no reason to expect more rapid growth following democratizing transitions for growing countries so this rules out the first effect, but even compensating for that poverty did not fall as fast so it might appear that factors (b) and (c) worked against more rapid poverty reduction. The better distribution of income would have been the name for more pro-poor growth, which one might have expected from a democratic government. However, the principal difficulty is that in spite of the huge number of references to pro-poor growth or inclusive growth or broad-based growth 6 there is no firm, empirically based, guidance as to which actual governmental actions or policies might produce such growth. 6 Terms which get 130,000, 611,000 and 475,000 Google hits each. Preliminary and Incomplete: 32 9/28/2010

33 Since the distribution of income across households is the result of the complete general equilibrium outcome of an economy involving millions of firms and individuals taking investment, production, and consumption decisions it is unlikely that any simple formula are likely to emerge. In fact, as with any complex problem, it is perfectly possible for well-meaning but naïve actions to have counter-productive results. For instance, it might seem that since the main asset of the poor is (unskilled) labor that pro-labor actions would be pro-poor, and that strengthening labor production would therefore be propoor. But, it is possible that the increased labor protections are only effective in the formal sector, where wages are already high, which causes a reduction for demand for labor in the formal sector which then expands the supply of labor in the informal section which then reduces the relative wages in the informal sector, which actually increases poverty. Note that I am not asserting that this is the case, I am just giving an example in which a democratic government could adopt a seemingly pro-poor action (that was also in some quarters politically popular) than nevertheless actually would, through the repercussions of the actions of economic agents have exactly the opposite effect. Another example, might be raising the price of rice by say, banning imports. While it might seem pro-poor as the poor are rice farmers, in fact many of the poor are net consumers of rice so that an increase in the price of rice through a ban on imports again an action both politically popular and seemingly pro-poor would actually increase poverty. Again, the point is not about particular actions but rather that one can hardly have expected democratic governments to have been more pro-poor in the absence of reliable guidance as to what actions would in fact be pro-poor and that those would be politically more popular in a democratic arena. Preliminary and Incomplete: 33 9/28/2010