Handicapping China’s--and India’s--Economic Futures (II)

choi

Lant Pritchett and Larry Summers, Asiaphoria Meets Regression to the Mean. National Bureau of Economic Research, October 2014 (NBER Working Paper No 20573)
www.nber.org/papers/w20573

Quote:

"From 1967 to 1980, Brazil’s economy grew at 5.2 percent per year. While many people might have identified macroeconomic and structural imbalances putting that growth at risk of a recession or cyclical slowdown, no one in 1980 was predicting that for the next 22 years—from 1980 to 2002—per capita growth would be almost exactly zero. We conjecture that nearly any assessment of the risk of such an extended slowdown using existing statistical methods for forecasting long-run growth would have found a 5.2 percentage point deceleration an extremely improbable outcome. In this section we examine episodes of growth to argue that, while not our modal forecast, the likelihood of a slowdown much larger than the regression to the mean—a so-called 'sudden stop'—is a possibility.”  page 30

“Hausmann, Pritchett,and Rodrik (2005) document the existence of frequent growth accelerations of substantial magnitude (more than 2.5 percent per year) to rapid growth. They show the timing of these growth accelerations are typically not well explained by standard growth determinants (eg,good policy) or changes in the standard growth determinants (eg,policy reform).”  page 31

“India’s [GDP] growth regime is an annual growth rate of 2.09 percent from 1950 to 1993, quite near the world average of 2.15 percent. This is followed by an acceleration of growth to 4.23 percent from 1993 to 2002, then a second acceleration of growth from 4.23 to 6.29 from 2002 to 2010. In this set of episodes India hasexperienceda period of acceleratedgrowth for 17 years (1993to2010) at a pace of 4 percent or higher.”  page 32

China’s GDP “produces three accelerations in a row. Growth from 1968 to 1977 was 4.33 percent per year, accelerating to 7.61 percent from 1977 to 1991 and accelerating yet again in 1991 to 8.63 percent until 2010. * * * China has had growth rates of over 6 percent for 33 years starting in 1977,and this data set ends in 2010.”  page 33

“Speculation about how much longer China’s and India’s current episodes of rapid growth might last and what might happen after those episodes through comparison with all other experiences of country accelerations into rapid growth is informative but not dispositive. Table 9 shows all 28 growth recorded accelerations that resulted in episodes of growth higher than 6 percent per year (which is roughly two standard deviations above the cross-national mean).This table reveals how unusual China’s (and to a lesser extent India’s) current growth experiences are in three ways.”  page 36

"First, episodes of super-rapid growth (>6 percent) tend to be extremely short-lived. The Kar et al (2013) method of dating growth episodes mechanically does not allow for episodes of less than eight years. The median duration of a super-rapid growth episode is nine years, only one year longer than its possible minimum. There are essentially only two countries with episodes even close to China’s current duration. Taiwan had a growth episode from 1962 to 1994 [total 32 years] of 6.8 percent (decelerating to growth of 3.5 percent from 1994 to 2010). Korea had an episode from 1962 to 1982 followed by another acceleration in 1982 until 1991 when growth decelerated to 4.48 percent—a total of 29 years of super-rapid growth (> 6 percent)—followed by still rapid (> 4 percent) growth. So, China’s experience from 1977 to 2010 already holds the distinction of being the only instance, quite possibly in the history of mankind, but certainly in the data, with a sustained episode of super-rapid (> 6 ppa) growth for more than 32 years [and continuing].

“Second, the end of an episode of super-rapid growth is nearly always a growth deceleration. Of the 28 episodes of super-rapid growth, only two ended with a shift to higher growth: Korea in 1982 and China in 1991. So, again, China is remarkable in that its acceleration to super-rapid growth in 1977 was followed by another acceleration in 1991.”  page 36 (continuing discussing Table 9)

"Third, the typical (median) end of an episode of super-rapid growth is near complete regression to the world mean growth rate. The median of the growth episode that follows an episode of super-rapid growth is 2.1 percent per year. So the ‘unconditional’ expectation (or central tendency) of what will happen following an episode of rapid growth, conditional on a shift in growth, is a reversion to not just somewhat slower growth but actually a massive deceleration of 4.65 percentage points. Such a slowdown is more than twice the cross-national standard deviation of growth rates of roughly 2 percent. A deceleration of that magnitude would take India’s current growth episode of 6.29 to 1.64 percent and China’s from 8.63 (in the episode since 1991) to 3.98 percent."  page 37

"The results in Table 9 are not an artifact of classifying just super-rapid (>6 percent) growth. If we look at all episodes of growth greater than 4 percent (one standard deviation above mean) we would find many more episodes but similar results about duration and deceleration in all three regards. The 70 episodes of growth above 4 percent (inclusive of those above 6 percent) also have a median duration of nine years. One does find more examples of extended rapid growth greater than 4 percent—Singapore with 30 years at 4.17 percent from 1980 to 2010, Indonesia with 29 years at 4.71 percent from 1967 to 1996, Thailand with 29 years at 4.91 percent from 1958 to 1987 (followed by an acceleration), Vietnam with 21 years (and ongoing) at 5.54 percent. But still, other than the combination of Thailand’s episodes (the first of which was at much lower rates than China’s and the end of which precipitated the East Asian crisis of 1997), none of the episodes of even rapid growth (>4 percent) is of longer duration than China’s. In the 70 episodes of rapid growth (>4 percent), there are only four cases in which the episode ended with a growth shift that was an acceleration (China in 1991, Korea in 1982, Thailand in 1987, and Botswana in 1982). Finally, the median growth in the episode following the rapid growth episodes is 1.85percent. Again,the growth following an episode of rapid growth is, on average, full regression to the mean."  page 39

Note:
(a) This posting concerns a single issue the Pritchett-Summers paper seeks to address: “4. How long do episodes of rapid growth usually last? How do they end?”

(b) The paper spends four pages (pages 39-42) talking about middle-income trap.   However, page 6 (in the Introduction) tips the hand: "Section 4 also demonstrates that, in considering China’s prospects for continued rapid growth,the much-discussed middle income trap(if, properly measured, it even exists) is less a fundamental empirical issue than a simple regression to the mean."
(i) The authors have their doubts that the trap exist.
(ii) Even if the trap exists, its predictive value is not as good as the authors’.