Openness and Inflation: Theory and Evidence

Romer’s starting point is Kydland and Prescott’s (1977) demonstration that absent policy precommitment, monetary policy-makers face an incentive to generate surprise inflation to raise output above its (suboptimally low, due to imperfect competition or distortionary taxation) natural rate, producing inefficiently high equilibrium inflation with no output gain. Nearly fifteen years after that paper, Romer notes, “there is virtually no empirical evidence on the question of whether the mechanism identified by Kydland and Prescott is important to actual inflations,” with the debate polarized between Barro and Gordon (1983a), who treat the model as broadly descriptive, and Taylor (1983), who argues institutions largely overcome the problem (Sec. I). Romer’s contribution is to extract a testable cross-sectional prediction: following Rogoff’s (1985b) observation that surprise monetary expansion depreciates the real exchange rate – which is why coordinated expansion by two countries is more inflationary than unilateral expansion in Rogoff’s model, since joint expansion cannot move the exchange rate between them – it follows that a more open economy, which imports more of its consumption basket, suffers a larger welfare cost from the same real depreciation, and therefore faces a smaller incentive to generate surprise inflation in the first place.

The model extends existing closed-economy sticky-price, monopolistic-competition frameworks (citing Blanchard and Kiyotaki 1987 and Ball and Romer 1990) to an open economy where a fraction a of consumption is imported. CPI inflation x is a weighted average of imported-goods inflation and domestically-produced-goods inflation (eq. 1); demand for domestic output requires a fall in its relative price – real depreciation – to expand (eq. 2); a fraction f of domestic prices are flexible and set according to a markup-type rule in output (eq. 4), with the rest sticky at pre-set levels; money demand is standard (eq. 5); and an analogous system describes the single foreign country. Solving for the effects of an ex post (unanticipated) monetary expansion with prices already set (eqs. 10a-d), Romer shows that because the parameter governing flexible-price responsiveness does not depend on openness a, the output effect of a monetary surprise is smaller, and both domestic-price and CPI-inflation effects are larger, in a more open economy – so the output-inflation tradeoff (dy/dp or dy/dx) is strictly less favorable the more open the economy is, even though the impact on the real exchange rate itself is shown to be independent of openness under the model’s specific assumptions.

Equilibrium requires (i) rational expectations, which (given no uncertainty in the model) pins output at its flexible-price level, and (ii) that the monetary authority’s ex post optimal money growth condition (eq. 16) holds exactly at that flexible-price equilibrium – i.e., expected (and hence pre-set) inflation adjusts to the point where the monetary authority no longer wishes to create additional surprise inflation. Substituting the comparative statics from Q2 into this condition (eq. 16), Romer shows the left-hand side (marginal benefit of a money-supply surprise) is strictly decreasing in openness a, while the marginal cost term (dx/dm, via the convex inflation-cost function c(.)) is strictly increasing in a – so maintaining the equality as a rises requires x (equilibrium inflation) to fall: “increased openness reduces equilibrium inflation” is established formally, not merely illustrated (Sec. II, end). Romer also shows that endogenizing the fraction of flexible-price firms f (allowing costly ex post price adjustment that varies by firm) strengthens rather than weakens this result, since the incentive to retain price flexibility is itself greater in a more open economy under natural assumptions.

The sample is the 114 non-centrally-planned economies listed by Summers and Heston (1988) with available data on openness and inflation (seven more are excluded for insufficient data), covering the post-1973 floating-rate period (Sec. III) – chosen because Bretton Woods limited independent monetary policy before 1973. Inflation is the change in the log GDP/GNP deflator (or CPI where unavailable); openness is the average import share of GDP/GNP since 1973. The basic specification regresses the log of average inflation (to reduce the influence of a handful of extreme-inflation countries) on a constant and openness, with further specifications adding dummies for alternative data sources, log real per capita income, and regional/OECD dummies.

In the basic regression (Table 1, Column 1), the t-statistic on openness is -3.8, and the point estimates imply average inflation of 18 percent for a closed economy, 14 percent at a 25 percent import share, 11 percent at 50 percent, and 8 percent at 75 percent – openness alone accounts for over ten percent of cross-country variation in average inflation. Adding data-source dummies, log real income per capita, and regional/OECD dummies (Columns 2-4) leaves the coefficient “little changed” (the Column 4 coefficient is about 80 percent of the Column 1 value, with t-statistic still above 3), even though the regional dummies themselves are jointly highly significant (South America’s dummy coefficient is 1.17; Africa and Central America 0.20 and 0.30; OECD and oil producers only 0.09 and 0.06).

Table 2 shows the coefficient survives excluding high-inflation countries (>30 percent), excluding countries with closely tied monetary policies (Francophone African unions, EMS members, dollarized economies), restricting to the 63 countries with the most reliable data, and splitting by region, with no evidence the coefficient differs significantly across regions (Sec. III). The relationship is absent, however, among the roughly 15-20 wealthiest countries in the sample (the United States, Canada, Norway, and similar): among these, there is “virtually no link between openness and inflation,” and – more importantly – their average inflation is uniformly low (8.7 percent), with the full-sample regression underpredicting inflation for 17 of these 18 countries. Excluding these 18, the remaining 96 countries show an average inflation rate of 18.9 percent and a stronger openness-inflation link (t-statistic -4.6; inflation falls from 21 percent to 8 percent as openness rises from 0 to 75 percent at median income). Romer interprets this as evidence that a small group of highly developed countries has found “some means of solving the problem of dynamic inconsistency of optimal monetary policy,” while the rest of the world has not – and notes the data reject the idea that this group reflects a smooth function of development: within the non-elite sample, the openness-inflation link if anything strengthens, not weakens, with income (footnote 13).

Because the import share can itself reflect protectionist policy choices correlated with other inflationary policies (e.g., large budget deficits financed partly by seignorage), Romer reestimates by instrumental variables using log land area as an instrument for openness (land area predicts openness strongly – a regression of import share on log land area alone has a t-statistic of -9.5 and R-squared of 0.45 – and is plausibly uncorrelated with current policy choices). The IV estimates of the openness-inflation link are, if anything, larger than the OLS estimates (e.g., -1.32 vs. -1.01 in the simplest specification), remain highly significant, and Hausman tests fail to reject equality of the IV and OLS estimates (Table 3) – providing no evidence that endogenous openness drives the result.

The alternative hypothesis holds that smaller/more open economies rely more on tariff revenue (so need less seignorage) and/or face a more elastic money demand (so set a lower optimal inflation tax), generating a spurious negative openness-inflation correlation unrelated to the precommitment mechanism (Sec. III, “Alternative Explanations”). Romer offers three counter-arguments: (1) if the budgetary channel were driving the result, the openness-inflation link should weaken as per-capita income rises (since tariff/seignorage revenue shares fall sharply with income – from 32 percent of government revenue for the poorest countries to 3 percent for the 18 richest, and seignorage from 8 percent to 0 percent) – but the interaction term between openness and log income actually enters negatively (-0.37 to -0.72 depending on sample), the opposite sign the budgetary story predicts; (2) the magnitude of tariff reliance is far too small: a drop in import share from 60 to 20 percent of GDP is estimated to lower the tariff-revenue share of government income by only about 13 percentage points, far short of the roughly 50 percent increase in average inflation the main regressions associate with that same change in openness; (3) estimates of money-demand elasticity from Fair (1987) and Driscoll and Lahiri (1983), regressed on openness, show a small and statistically insignificant relationship, with the implied elasticity of money demand under 0.1 even in quite open economies – too small for seignorage-revenue loss to be a central driver of average inflation (Table 4), a conclusion Romer buttresses by noting Cagan’s (1956) classic hyperinflation evidence likewise implies money-demand inflation elasticities below 0.1 for inflation rates under 20 percent.

Romer concludes the findings “suggest that models in which the absence of precommitment in monetary policy leads to inefficiently high average levels of inflation are essential to understanding inflation in most of the world” (Sec. IV). He draws out a specific policy caution: “increased economic cooperation and integration, particularly outside the most highly developed countries, should be viewed with caution: in the absence of some mechanism to overcome the dynamic inconsistency problem, increased integration may lead to large increases in average inflation” – i.e., since the paper’s own mechanism implies that effectively shrinking an economy’s “openness” (by integrating it into a larger bloc) removes exactly the real-exchange-rate discipline the theory identifies, deeper economic integration among non-elite economies could raise, not lower, average inflation absent other precommitment devices.