Discussion of The return to capital in China by Chong-En Bai, Chang-Tai Hsieh, Yingyi Qian and Zhenjie Qian.

Transcription

1 Discussion of The return to capital in China by Chong-En Bai, Chang-Tai Hsieh, Yingyi Qian and Zhenjie Qian. Olivier Blanchard December 2006 This is a very useful paper. There is a widely held belief that the return to capital in China is low, driven down by an investment rate in excess of 40% of GDP. The paper looks at the data carefully, and concludes that this is not the case: The rate of return decreased in the 1990s, but is now stable, at a high 20%. The approach of the paper is straightforward. Simplifying a bit: It defines the rate of return to capital as the ratio of profit to the value of the capital stock, minus the depreciation rate: R P Y Y W N P K K δ It then rewrites this expression as: P Y Y W N P K K δ = P Y Y W N P Y Y P Y Y P K K δ = α P K K/P Y Y δ where α is the share of capital in output. Turning to the data, the authors find that the ratio of the value of capital to the value of output is relatively low, and the share of capital in output relatively high. Together, these two facts imply a high rate of return on capital. In 2005 for example, the capital output ratio was 1.72, the share of capital was 0.58, the depreciation rate was 0.1, implying a rate of return of 23%. Discussion, Brookings Panel on Economic Activity 1

2 Do the computations leave room for disagreement? The authors learly do as careful an empirical job as one can do given the somewhat shaky data. Still, the high value of the capital share is striking, standing at about 15 to 20 percentage points above values of the capital share in countries at similar levels of development (I am using here the values for the capital share constructed by Gollin 2002.) To those of us who (for no good reason) like to see production as largely Cobb-Douglas, such a dramatic difference in capital shares is puzzling. Thus, at the presentation of the paper at the Brookings meeting, I tried my best to come up with various stories for measurement error, from the misallocation of self-employment income, to the different incentives to mistate profits by TVEs, by state firms, or by multinationals. In this final draft, the authors have done a great job of maiming if not killing most of my tentative hypotheses... I still want to take up two issues. The first is that of self-employment. In most low- and middle-income countries, the reported capital share is indeed very high, often above 0.6. But, as is well understood, a large part of what is classified as capital revenues is in fact selfemployment income, which should be largely counted as labor income. When Gollin makes this adjustment, the resulting capital share drops down, often by more than 20 percentage points. I was suspecting that the same was likely to be true in China. It turns out however that, according to the Chinese statistical authorities, until 2005, all self-employment income was included in labor income. In other words, if anything, the reported capital share understated true capital income, at least until I must admit to be stumped; all i can do is a bad loser, and suggest that somebody look over the shoulders of the Chinese statistical authorities to see how the statistics are put together... The second is the implication of the reported capital share for another standard exercise, namely the computation of TFP growth. The table below summarizes the conclusions from a number of studies, which differ in a number of empirical dimensions, but typically use capital shares for the construction of the Solow residual roughly similar to those used by Hsieh et al. The general pattern is that of a decrease in measured TFP growth over time, with very low measured TFP growth in recent years. 2

3 Table 1. TFP computations, by various authors Period Annual TFP growth (percent) Maddison Zheng Jinghai Angang Zheng OECD Kujis Wang Source; Blanchard-Giavazzi 2006 (updated) That true TFP growth would have slowed down in recent years is, on the face of it, implausible. State firms are being restructured. FDI has been very high, and we know from a number of studies that foreign firms, or joint ventures with foreign firms, have higher productivity than domestic firms. So, again, measurement error seems like a plausible explanation. And a natural possibility in this context is again an overstatement of the capital share. To see why, suppose that true TFP is given by: and measured TFP is given by: where α K and S = g Y α K g K (1 α K )g N Ŝ = g Y αˆ K g K (1 αˆ K )g N αˆ K are the true and the measured shares respectively. Then: Ŝ = S + (α K αˆ K ) (g K g N ) Thus, as (g K g N ) is typically positive, if the capital share is overstated, measured TFP growth will understate true TFP growth. And the downward bias will be stronger the larger (g K g N ), thus the larger the investment rate. For 3

4 (g K g N ) = 10% for example, an overstatement of the capital share by 20 percentage points will lead to an underestimation of 2% a substantial magnitude. One can therefore interpret declining measured TFP growth as an indication that the capital share is overstated. Again, this is no proof, just suggestive evidence. Let me end with a remark about differences in rates of return across firms. The paper offers a decomposition by provinces and by sector primary, secondary, and tertiary. (One may question estimates of the rate of return in the primary sector, given the difficulty of separating between revenues going to land and revenues going to reproducible capital). While this is interesting, two of the central questions in China have to do with other decompositions. The first is the rate of return to spending on health; the provision of health care, especially in the provinces, has sharply declined, and it is widely believed that the social rate of return there is very high, justifying a reallocation of investment towards that sector. Clearly, this falls outside of the scope of this paper. The second is differences in rates of return by ownership. Many observers suspect that the rate of return in state firms is considerably lower than in new firms. As the authors indicate, there is no way to construct such rates of return with the data they use. But we have some information from estimates using individual firm data from the industrial sector (OECD 2005). The evidence is presented in the table below: Table 2. Rate of return, percent (Profit plus interest as percent of physical capital stock and inventories, net of depreciation and taxation) Source: OECD Private Individually controlled Non mainland controlled State-controlled Direct control These rates of return reflect both the inclusion of inventories in capital, and taxation. Thus, they are lower than but consistent with the central estimates 4

5 of the paper. What the table shows however is a clear difference across ownership types. Private firms have much higher rates of return than state firms. The average rate of return for state firms is still high (and appears to have increased since 1999); but this hides substantial heterogeneity: The median rate of return is only 1.5%, and 35% of state firms report negative profits. This, again, has obvious implications for China s growth strategy. References Blanchard, O. and F. Giavazzi, Rebalancing growth in China. A three-handed approach, in China and the World Economy, July-August 2006, 14-4, OECD Economic Survey of China,