Separating Wheat from Chaff in PPP Assessments of Exchange Rate Misalignment
In a recent paper, Cheung, Chinn, and Nong (2016), or CCN, have updated and extended their ongoing analysis examining exchange rate misalignments based on deviations from international patterns involving purchasing power parity (PPP). For the important case of China, they find that the renminbi was extremely undervalued in 2005, but then rose to an appropriate level by 2011 and proceeded to become about 25 percent overvalued by 2014 (p. 32). However, it is highly unlikely that the Chinese currency has become seriously overvalued, because the medium-term outlook for China’s current account is an ongoing surplus of 1 to 2 percent of GDP (IMF 2016) rather than a plunge into excessive current account deficits.
The identification of a fundamental equilibrium exchange rate (FEER) at which the medium-term current account can be expected to lie within an acceptable range (Williamson, 1983; Cline, 2008) will usually provide a more reliable basis than the PPP approach for examining misalignments. Using the FEER approach, my most recent estimates find that the Chinese renminbi is presently neither undervalued nor overvalued (Cline 2016). Differing databases and differing model specifications tend to give widely varying assessments in the PPP approach, posing serious problems of reliability for any single set of estimates. The approach has not provided good guidance in important historical episodes such as Japan’s high surpluses from the mid-1980s to the early 2000s, despite measured overvaluation on a PPP basis.
The PPP approach is usually premised on the Balassa-Samuelson effect whereby the ratio of the market exchange rate to the PPP exchange rate tends to rise over time for developing countries as their per capita incomes increase.1 (CCN call this the “Penn Effect.”) One can then estimate cross-country regressions to identify the “normal” trendline indicating the amount by which the market exchange rate is weaker than the PPP exchange rate, with this gap typically being found to narrow as per capita income rises. CCN follow Kessler and Subramanian (2014) in finding that for developing and emerging-market economies, this line is actually a mild U-shaped curve, such that the usual Balassa-Samuelson progression only begins after per capita income at the market exchange rate exceeds about 4 percent of the US level.2 An individual developing country is then diagnosed as being undervalued if its exchange rate is even weaker than this curve would indicate, but overvalued if the reverse is true.
In general there are good reasons why a FEER approach should be expected to be more reliable than the PPP approach to assessing exchange rate misalignment. It is the external balance that provides the most direct indication of whether the currency is misaligned. My colleague John Williamson and I have long used a relatively accommodative range of ± 3 percent of GDP as the bounds within which the current account imbalance should be constrained, with the lower bound premised on long-term external debt sustainability for the country in question and the upper bound reflecting the need for global adding-up and avoidance of beggar-thy-neighbor behavior. The Balassa-Samuelson PPP line will not necessarily guarantee that the external imbalance will remain within such a range. The presence of trade barriers can impede realization of the expected relationship. Moreover, the extent of the tradable/nontradable productivity differential in relationship to per capita income is not necessarily tight. These and other reasons lie behind the result that the cross-country scatter plot of the PPP exchange rate looks much more like a very fat upward-tilted sausage (see Cheung, Chinn, and Nong 2016, p. 34), rather than a narrow upward-sloping line or U-curve.
Williamson and I earlier questioned the usefulness of PPP comparisons for purposes of assessing currency misalignment (Cline and Williamson 2008). At that time, Cheung, Chinn, and Fujii (2007) were estimating that the renminbi should be appreciated by 100 percent to eliminate undervaluation, whereas my estimate for 2007 was that the needed appreciation was about 15 percent on a real effective exchange rate (REER) basis and about 36 percent bilaterally against the US dollar (p. 133). Williamson and I pointed out that it was well known that the survey estimates of China’s domestic prices were extreme outliers on the low side. As Jeffrey Frankel (2008, p. 157) subsequently noted, “Several weeks after the Peterson Institute conference that led to this volume … the authors were proven spectacularly right … [by new data released by the Asian Development Bank and World Bank]”. Thus, one ongoing problem is the risk of mistaken measurement of domestic prices.
A more fundamental problem is that the FEER needs to be judged in terms of the prospective current account outcome. If an unsustainably large deficit is already present or in the pipeline in view of lags from the exchange rate to trade, then the exchange rate is overvalued, regardless of what the PPP comparison would say. Conversely, if the economy is on track to maintain or develop an extremely large current account surplus (such as China’s 10 percent of GDP surplus in 2007) at the present real exchange rate, then the currency is undervalued, even if the exchange rate looks unduly strong in terms of the actual exchange rate compared to what the development-PPP curve would predict.
Thus, Williamson and I cited the classic case of Japan, in which experience from the mid-1980s to early 2000s showed that “the international basket of goods and services can cost about 50 percent more in Japan than in the United States with external equilibrium [defined at a surplus of 2 percent of GDP] still being maintained” (p. 137). Similarly, whereas Kessler and Subramanian (2014), like CCN now, estimated that in 2011 China was no longer undervalued, their model also estimated that Turkey was seriously undervalued (by 11 percent), even though Turkey had a dangerously high current account deficit of about 10 percent of GDP, a patently implausible finding for purposes of assessing a FEER.3
Now consider the CCN findings for China. In their preferred model (quadratic, developing-country panel) and preferred data set (Penn World table [PWT] 8.1), they find that the renminbi was 49 percent below the PPP-based trendline in 2005, but that by 2011 the currency was 3 percent above the line. Considering that the average current account balance in the two years following each of these dates fell sharply from 9.2 percent of GDP in 2006–07 to 2.1 percent in 2012–13 (IMF 2016), their model deserves credit for informational value. Indeed, the IMF’s own projections in the spring of 2011 anticipated a Chinese surplus of averaging 6.6 percent of GDP in 2012–13 (IMF 2011). In qualitative terms, the CCN model would have done better than that of the Fund, except for the problem of delay in data availability (the PWT 8.1 data only became available in 2013; Feenstra et al. 2013).
But the degree of reliability seems to be low. CCN find that use of the PPP data in the World Bank’s World Development Indicators, in their preferred quadratic-developing model, gives far different estimates (undervaluation of only 6 percent in 2005 and overvaluation of 26 percent in 2011). Similarly, if the linear model and full sample are used instead of the quadratic-developing model, the PWT 8.1 data show the renminbi still undervalued by 11 percent in 2011 (p. 32).
A particular problem of interpretation is that even if one accepts the CCN preferred conclusion that the renminbi was appropriately valued in 2011, it does not follow that at present the currency has become seriously overvalued, because there has been a real effective appreciation of 23 percent from 2011 to the first eight months of 2016.4 The reason is that the extent of the prospective current account imbalance should be the key to determining over- or undervaluation, and the PPP approach does not necessarily provide a good guide to the likely current account. The IMF’s latest medium-term forecast places China’s surplus at 0.5 percent of GDP in 2021 (IMF 2016). If instead the renminbi were overvalued by 23 percent, one would anticipate a sizable deficit, on the order of 3 to 5 percent of GDP.5 A key reason that the renminbi seems unlikely to have become seriously overvalued is that China’s current account has manifested an upward drift of about 0.6 percent of GDP annually for a given level of the real effective exchange rate (Cline 2012). From 2011 to 2021, that drift would boost the current account by 6 percent of GDP if there were no offsetting appreciation.
The challenge in this area is how to make meaningful use of the PPP information, given the lags in data availability as well as the extreme variability depending on the model and vintage of the data set. The CCN results for China in 2011 in comparison with the Fund’s current account surplus projections (which later had to be substantially downscaled) suggest that at least the PPP approach may provide a useful basis for further investigation of the prospective future current account path if the PPP-based model implicitly disagrees sharply with the Fund’s baseline.
Author's note: For comments on a previous draft, I thank without implicating Olivier Blanchard, Joseph Gagnon, and Nicholas Lardy.
Balassa, Bela. 1964. The Purchasing-Power Parity Doctrine: A Reappraisal. Journal of Political Economy 72, no. 6 (December), pp. 584–96.
Cheung, Yin-Wong, Menzie D. Chinn, and Eiji Fujii. 2007. The Overvaluation of Renminbi Undervaluation. NBER Working Paper 12850. Cambridge, MA: National Bureau of Economic Research.
Cheung, Yin-Wong, Menzie Chinn, and Xin Nong. 2016. Estimating Currency Misalignment Using the Penn Effect: It’s Not as Simple as It Looks. Available at: http://www.ssc.wisc.edu/~mchinn/CNY_misalignment.pdf. Summarized at: http://voxeu.org/article/estimating-currency -misalignment-using-penn-effect.
Cline, William R. 2008. Estimating Consistent Fundamental Equilibrium Exchange Rates. Working Paper WP 08-6 (July). Washington: Peterson Institute for International Economics.
Cline, William R. 2012. Projecting China’s Current Account Surplus. Policy Brief PB 12-7 (April). Washington: Peterson Institute for International Economics.
Cline, William R.. 2014. Estimates of Fundamental Equilibrium Exchange Rates, May 2014. Washington: Peterson Institute for International Economics.
Cline, William R. 2016. Estimates of Fundamental Equilibrium Exchange Rates, May 2016. Washington: Peterson Institute for International Economics.
Cline, William R., and John Williamson. 2008. Estimates of the Equilibrium Exchange Rate of the Renminbi: Is There a Consensus and, If Not, Why Not? In Debating China’s Exchange Rate Policy, ed. Morris Goldstein and Nicholas R. Lardy. Washington: Peterson Institute for International Economics, pp. 131–54.
Feenstra, R. C., R. Inklaar, and M. Timmer. 2013. PWT8.0 – A User Guide. Gronigen, NL: Growth and Development Centre, University of Gronigen.
Frankel, Jeffrey A.. 2008. Comment: Equilibrium Exchange Rate of the Renminbi. In Debating China’s Exchange Rate Policy, ed. Morris Goldstein and Nicholas R. Lardy. Washington: Peterson Institute for International Economics, pp. 155–65.
IMF (International Monetary Fund). 2011. World Economic Outlook Database April 2011. Washington.
IMF (International Monetary Fund). 2016. World Economic Outlook Database April 2016. Washington.
Kessler, Martin, and Arvind Subramanian. 2014. Is the Renminbi Still Undervalued? Not According to New PPP Estimates. RealTime Economic Issues blog (May 1, 2014).
Williamson, John. 1983. The Exchange Rate System. Policy Analyses in International Economics 5. Washington: Peterson Institute for International Economics.
1. In a single factor (labor) Ricardian framework, Balassa (1964) posits that the rich country has higher absolute productivity than the poor country in both the traded good and the non-traded good (or service), but that the “advantage is greater in regard to traded goods …. The relative price of the non-traded commodity will thus be higher in the [rich] country with higher productivity levels than in the other [poor country]. Since the prices of traded goods are equalized in the two countries through international exchange, …the currency of the country with the higher productivity levels will appear to be overvalued in terms of purchasing power parity.” Balassa (1964, pp. 585–86).
2. Their minimum ratio of actual to PPP exchange rate occurs at relative PPP per capita income with the logarithm -3.25 (p. 9); e-3.25 = 0.039. The authors cite Bangladesh as an example. Note further that for reasons that are not clear, the curve becomes an inverted U if the sample is limited to developed countries.
3. See Cline 2014, note 32.
4. Using the SMIM model (Cline 2008, 2016) trade weights for 34 key economies (with the euro area treated as a single economy) and deflating by consumer prices.
5. In the SMIM model, the impact parameter for China is a reduction of 0.239 percent of GDP for each percentage point rise in the real effective exchange rate. The lower end of the range stated takes the 2 percent surplus in 2012–13 as the base; the higher end considers a base of 0.5 percent surplus.