Emerging Market Business Cycles: The Cycle is the Trend

Mark Aguiar and Gita Gopinath, Journal of Political Economy 2007. Read here in the NBER Working Paper 10734 version (August 2004). No talk recording exists; written from the paper alone.

Neumeyer and Perri, one entry back, explained emerging-market cycles by adding machinery — working capital, country risk, a transmission belt from spreads to labor demand. Aguiar and Gopinath’s counter is a study in subtraction: keep the plain-vanilla frictionless small-open-economy RBC model, no GHH preferences required, no financial frictions at all, and change one assumption about the shocks. In developed economies, productivity fluctuates transitorily around a stable trend. In emerging markets, they propose, the trend itself is the thing that moves — hence the title, which is one of the field’s great five-word theses. The motivation is not statistical but political: emerging markets live through “dramatic reversals in fiscal, monetary and trade policies” — regime switches — and a regime switch is not a blip around trend, it is a new trend. Their exhibits: Venezuelan petroleum productivity fell 50 percent within five years of the 1975 nationalization; Brazilian iron-ore productivity roughly doubled after the 1991 privatization.

Why the nature of the shock is everything

The facts to be explained, from their 26-country panel (13 emerging, 13 developed, all small, 1980–2003): emerging markets have twice the output volatility; consumption about 40 percent more volatile than income (developed: slightly less than one); net exports strongly countercyclical (about three times the developed correlation); and — the discipline on any proposed explanation — filtered output shows the same autocorrelation in both groups, so you cannot get there just by cranking up shock persistence.

The permanent-income logic does the rest. Income follows ln y = trend + transitory, with the trend growth rate gₜ itself stochastic. A positive transitory shock is a windfall: consumers smooth, saving rises, the current account improves — the standard model’s acyclical-to-procyclical trade balance. A positive trend shock is a promotion: today’s income is up but future income is up by more, so consumption jumps more than output, saving falls, investment surges, and the country runs a deficit in good times — a countercyclical current account, and consumption more volatile than income, from a frictionless model with ordinary Cobb-Douglas preferences. When you learn your raise is permanent, you borrow against it; a country whose good news is chronically about the trend behaves like a household that keeps getting promoted and demoted.

The quantitative discipline is GMM estimation of the income process on Mexico (the archetypal emerging market) versus Canada (the archetypal developed SOE). The estimated ratio of trend-shock to transitory-shock volatility: over 2 for Mexico, about 0.5 for Canada. Feed each process into the same model: Mexico’s version produces a trade-balance–output correlation of −0.6 (data: −0.7) and consumption 10–15 percent more volatile than income (data: ~25 percent); Canada’s version delivers a near-acyclical current account (−0.2 in the data) and consumption 20 percent smoother than income. Same model, two shock mixes, both business cycles.

The Tequila test and the VAR

The showpiece is out-of-sample in spirit: Kalman-filter Mexico’s observed Solow residuals into trend and transitory components, feed them through the model, and watch 1994–95. The model produces a sudden stop — an abrupt trade-balance reversal of 8.5 percentage points of GDP between 1994Q4 and 1995Q2, against 8.7 in the data, with the accompanying collapse of output, consumption and investment.

 versus the model fed the decomposed Solow residuals (solid). The model reversal is 8.5 percentage points; the data’’s is 8.7.’)

What makes the fit informative is the mechanism: it is not just that the shock was big, but that the crisis loaded onto the trend component — agents read 1994 as bad news about where Mexico was going, not where it was. A complementary King–Plosser–Stock–Watson variance decomposition, resting only on balanced-growth cointegration assumptions, closes the case: permanent shocks account for about 50 percent of business-cycle-frequency output variance in Canada (on par with U.S. estimates) and 82 percent in Mexico. For emerging markets, the cycle is, measurably, the trend.

The debate it started

The paper is refreshingly candid about what it leaves open: “the important question of what drives the trend remains, and the answer may involve frictions.” That sentence is the hinge of a decade-long argument. The rival reading — pressed by García-Cicco, Pancrazi and Uribe, and implicit in Neumeyer–Perri — is that trend shocks are the reduced form of financial frictions: an interest-rate spike plus working capital looks, through the lens of a frictionless model, exactly like bad news about trend growth. The two papers in this block are thus not complements so much as competing sufficient statistics for the same pathology, and later work (Chang–Fernández and others) found the data prefer a blend, with the frictions doing more than this paper’s frictionless benchmark would suggest. This reading list has already met both faces of the synthesis: Restrepo-Echavarría’s stagnation paper (entry 16) uses exactly this paper’s trend-growth shocks inside a limited-commitment friction, and Tomz–Wright (entry 13) showed that the default model’s fit to history improves precisely when income shocks are to the trend. And the syllabus’s closing entry asks the question this paper’s decomposition invites: if what we call “the trend” in emerging markets is partly measurement — activity sloshing between formal and informal sectors — how much of the excess volatility is real at all?