From the standpoint of workhorse trade models featuring the assumption of a representative consumer, there is an uncomfortable degree of heterogeneity in the views of the general population towards free trade. According to a Pew Research Center survey conducted in 2006, about 55% of households in the US with annual incomes over $150,000 felt helped by trade, whereas the same was true for only 35% of those with annual incomes below $40,000 (Pew Research Center 2006). What is the reason for such large disparities?
One obvious answer is trade-induced wage effects. Indeed, trade liberalisation may lead to heterogeneous wage and employment effects via several channels, including those related to firms where workers are employed (Egger and Kreickermeier 2009, Helpman et al. 2010), workers’ skills (Burstein and Vogel 2016) and/or workers’ local labour markets (Autor et al. 2013).
However, even if wage and employment effects are absent, trade has heterogeneous effects on consumers via differences in the price effects. For example, according to the same 2006 Pew survey, 32%, 30%, and 23% of consumers in the US believed that trade agreements increase, decrease, or have no effect on prices, respectively. These large differences may be due to the fact that people consume different baskets of goods and that the prices of those baskets respond differently to trade liberalisation. In a new paper, I use a model of trade with heterogeneous consumers and non-homothetic preferences to explore this channel and examine the magnitude of the errors in the estimates of welfare gains from trade introduced by the assumption of a representative consumer (Nigai 2016).
Between-country and within-country heterogeneity
Developing and developed countries differ inter alia in terms of the relative consumption of food, manufacturing goods, and services. This has to be taken into account when evaluating how opening up to trade would affect consumers in different countries. Recent papers (e.g. Fieler 2011, Markusen 2013, Tombe 2015) correct for these differences across countries by employing non-homothetic preferences and emphasising the role of the average per-capita income. However, within-country consumption heterogeneity sometimes may be larger than between-country heterogeneity, especially in countries with low average income and high income inequality. In such countries, the poor spend a large share of their total income on food, whereas this share is much lower for the rich. This can be seen in Figure 1, where I plot the predictions of my calibrated model for the share of income spent on food across the different deciles of the population (d = 1, …, 10, ordered according to the income level with d = 1 being the poorest) across 92 countries against normalised average per-capita income.
The reason for the heterogeneity is twofold. First, using detailed income data I calibrate each consumer group as having a certain share of the aggregate income such that the population is heterogeneous in their nominal incomes. Second, consumers differ in terms of their marginal propensity to consume manufacturing goods and services relative to agricultural goods. This is modelled as a stochastic, but positive, relationship between consumer’s initial income and preferences for non-food goods. On the one hand, this ensures that the rich consume relatively more manufacturing goods and services. On the other hand, because the relationship is not deterministic there is some heterogeneity in tastes even within the same income decile.
Figure 1. Expenditure shares on food versus average real per-capita income
In many low- and middle-income countries, consumers spend most of their incomes on food such that the trade-off between food and everything else is central. In that case, the differences in the welfare gains are determined by how much food prices go down in response to falling trade barriers relative to prices of other goods and services. This is in contrast to the recent results by Fajgelbaum and Khandelwal (2016), who argue that trade has pro-poor effects. They examine data on fewer and generally richer economies and emphasise the difference between less traded services and everything else.
How far off are predictions under the assumption of a representative consumer?
The vast majority of studies that target policymakers and, potentially, trade negotiators are based on the assumption of a representative consumer and homothetic preferences. To assess the accuracy of the predictions in such models, it is important to check how close the assumption of a representative consumer approximates the gains of different consumer types. For that, I develop two benchmarks. The first benchmark is based on the assumption of a representative consumer with homothetic preferences. The second benchmark is also based on the assumption of a representative consumer, but features non-homothetic preferences such that the representative consumer values consumption of manufacturing goods and services relatively more as the income grows. The two benchmarks are first used to evaluate the gains from a 15% reduction in trade barriers across the globe. The results plotted in Figure 2 suggest that welfare measures based on the assumption of a representative consumer would predict positive welfare gains for all countries. Naturally, driven by the data and the model, the gains differ across countries in the sample.
Figure 2. Gains from trade under the assumption of a representative consumer with homothetic and non-homothetic preferences
I next calculate the errors from using the assumption of a representative consumer defined as the difference between predictions under the assumption of a representative consumer and predictions under consumer heterogeneity across different income deciles. The results are plotted in Figure 3 and they suggest that the errors under the assumption of a representative consumer (both in the homothetic and non-homothetic case) are substantial. In both cases, the tendency is to overestimate the gains for the poor and underestimate them for the rich. The magnitude of the errors is substantial relative to the gains reported in Figure 1.
Figure 3. Errors under the assumption of a representative consumer with homothetic and non-homothetic preferences
The size of the errors in Figure 3 is mainly driven by developing countries with low average income and high income inequality. For example, people in the poorest decile in Ethiopia would gain 5.7%, whereas those in the richest decile would get a 10.9% increase in their welfare. Such differences are less acute for developed countries – for example, the difference in the gains from trade between the poorest and the richest deciles in the US would only be about 0.5 percentage points. Overall, even though in this experiment all consumers benefit from falling trade barriers, the rich tend to gain disproportionately more, and especially so in developing countries. This is because given an equal fall in trade costs, the prices of manufacturing goods fall more relative to agricultural goods. Since richer consumers spend a larger part of their income on manufacturing goods and services, they benefit relatively more.
But why do the prices of manufacturing goods fall more relative to food given the same reduction in trade costs? It turns out that the answer to this question lies in how different types of technology are dispersed across the globe. My structural estimates of the technology dispersion parameters strongly suggest that the dispersion of technologies across countries is much higher in manufacturing than in agriculture. Intuitively, this means that across countries the productivity differences in producing t-shirts are higher than, say, in growing tomatoes. Hence, the price differences between domestically produced and potentially-imported goods are larger in the former sector. Since the rich spend relatively more on clothing than food, they will benefit relatively more.
While evaluating the gains from trade for a representative consumer may remain a useful exercise for aggregate analysis, it is clear that consumers’ benefits from free trade agreements vary substantially across different income groups. Hence, it is unsurprising that certain population groups often fail to relate themselves to the predictions and numbers crunched for an average consumer. From that perspective, it seems important to evaluate the gains from potential free trade agreements along the entire income distribution.
Autor, D H, D Dorn and G H Hanson (2013) “The China Syndrome: Local labor market effects of import competition in the United States”, American Economic Review, 103(6): 2121-68.
Burstein, A and J Vogel (2016) “International trade, technology, and the skill premium”, Journal of Political Economy, forthcoming.
Egger, H and U Kreickemeier (2009) “Firm heterogeneity and the labor market effects of trade liberalization”, International Economic Review, 50(1): 187- 216.
Fajgelbaum P D and A K Khandelwal (2016) “Measuring the unequal gains from trade”, Quarterly Journal of Economics, 131(3): 1113-1180.
Fieler, A C (2011) “Non-homotheticity and bilateral trade: Evidence and a quantitative explanation”, Econometrica, 79(4): 1069-1101.
Helpman, E, O Itskhoki and S Redding (2010) “Inequality and unemployment in a global economy”, Econometrica, 78(4): 1239-1283.
Markusen, J R (2013) “Putting per-capita income back into trade theory”, Journal of International Economics, 90(2): 255 -265.
Nigai, S (2016) “On measuring the welfare gains from trade under consumer heterogeneity”, Economic Journal, 126: 1193-1237.
Pew Research Center (2006) “Free trade agreements get a mixed review”.
Tombe, T (2015) “The missing food problem”, American Economic Journal: Macroeconomics, 7(3): 226-258.