VoxEU Column International trade

Demand decomposition in trade: Quality and taste

The trade literature has long focused on firms’ productivity as an explanation for export performance. But what about the demand side? This column looks at firms’ appeal in terms of quality and consumer taste. Using Belgian firm-level data, it suggests that tastes could account for 45% of the variation in export quantities across countries.

The trade literature has long focused exclusively on firms’ productivity as an explanation for the variation in firm-product export performance. While supply characteristics are the main determinant explaining entry into foreign markets, this is less the case when explaining firm performance conditioning upon entry in a market. There is now a substantial stock of papers arguing that the supply side needs to be complemented with demand-related aspects to explain sales variation. For instance, several papers highlight the role played by product ‘quality’. Earlier contributions can be found in Feenstra (1994), Broda and Weinstein (2006) and Baldwin and Harrigan (2011).1

Quality is typically modelled as a demand shifter, which may capture very different effects. Furthermore, quality is generally an ordinal concept2 which is, therefore, hard to measure. In addition, quality is not incorporated into preferences as suggested by the literature in industrial organisation (Tirole 1988). As a consequence, it is fair to say that the word “quality” is used mainly to describe what stems from the demand side. Recently, Hottman et al. (2016) choose to call firm appeal, that is, product differentiation, what cannot be explained by economies of scope and cost differences. The question that remains is whether there is more to say about what “firm appeal” means?

In our paper (Di Comite et al. 2014), we strongly argue that there is. Using Belgian firm-level export data, we show that when accounting for quality and cost through firm-product dummies, and for destination-specific competition effects, through country-product dummies, only around 55% of the variation in export quantities at firm-product level is explained. This leaves 45% of the variation in export quantities still unexplained.[3] By comparison, the same set of explanatory variables explain between 79% and 96% of the variation in prices, depending on how the price data are reported (in weights or units). We argue that taste effects, as modelled in the Hoteling-Lancaster tradition within a love-for-variety monopolistic competition model, could well account for the remaining 45% of the variation in firm-product exported quantities across destinations.

To illustrate, consider Figure 1, which shows the quantity and price rankings of Belgian export to different markets at the firm-product level. We find that the relative exported quantities of the same firm-product across a range of very different destinations vary a lot, whereas relative prices are remarkably regular across markets and keep the same ranking. This implies that the bilateral cross-country rank correlation of exported quantities (triangles, on average 70%) is low compared to that of export prices (squares, on average 90%), suggesting that country-level variation in consumer tastes is high even if the observation of prices alone would fail to see it.

Figure 1. Correlations of export prices as well as quantities across destinations

Note: The squares dots indicate pairwise (bilateral) price rank correlations in the Belgian data of similar products across destinations. The triangles dots indicate bilateral country pairwise quantity rank correlations of similar products across destinations, ranked from high to low. The horizontal lines give average correlations.

We attribute the unexplained variation in export quantities to taste differences embedded in consumer preferences across destinations. A unique feature of our model is that quality and taste factors can operate in opposite directions in consumer demand. For example, a high-quality good may have a bad match with local tastes and as a result not sell much, despite its high quality. Empirically, this amounts to saying that while quality can be captured by a parallel demand shifter resulting in an upward shift of the residual demand curve, taste for a product is a demand slope shifter that can work in any direction affecting quantities. Together, these demand shifters jointly determine the sales of a product (next to cost and competition effects). This feature offers more flexibility than other demand models in which taste and quality mostly work in the same direction.

For example, in the case of quadratic preferences, a unique feature is that taste does not affect the price of a product, whereas quality does. This distinction between taste and quality is a distinction long-made in industrial organisation that delineates horizontal from vertical product differentiation through their distinct effect on price when quantities purchased by consumers are allowed to vary. Spatial models of product-market competition argue that horizontal product characteristics (flavour, colour, etc.) need not affect price, but that they can explain market share differences. In most other current trade models, parallel demand shifters affect the equilibrium product price but obfuscate the distinction between horizontal and vertical product features. In addition, quality alone cannot explain why the total variability explained by the same set of explanatory variables differs so much between prices and quantities.

Our model belongs to the family of quadratic utility models. However, in its standard form, the quadratic utility model does not necessarily offer a good alternative to the constant elasticity of substitution when it comes to explaining the variability in sales quantities (Melitz and Ottaviano 2008, Ottaviano et al. 2002). Absent a specific taste demand parameter, the standard quadratic utility model with heterogeneous costs and quality only explains 55% of the quantity variation observed (as argued above). Allowing for taste differences generates asymmetry in demand across countries. This offers a rational for the missing variability in quantities which is firm-product-country specific. The quadratic utility model augmented with taste, as a separate demand shifter, is currently the only model that allows for a clear distinction between quality and taste in a love-for-variety model. The incorporation of taste adds to the increasing popularity of the linear demand models in applied work in addition to features such as varying markups and the presence of competition effects (Mayer et al. 2014, Dhingra 2013, Eckel and Neary 2010, Foster et al. 2008).

Summing up, when the interest of the researcher lies in the identification of quality and taste effects of a given firm or firm-product, our approach offers an attractive alternative to several other models in the literature.

While the model has been developed to estimate demand parameters at the firm-product level, it can also be used at higher levels of aggregation. Researchers with access to firm-level data (Amadeus, Orbis) or product-level data (Comext, Comtrade) can thus also estimate demand parameters stemming from the model. For policy purposes, the model has recently been used in this context. While the model in principle allows for the construction of both quality and taste indices, the empirical implementation of a quality index, purged from taste effects is easiest to implement. A quality index cleaned from taste effects was developed for each of the 10,000 products exported to the European market by different exporting countries (Vandenbussche 2014, Di Comite 2012). In Figure 2, we show how the distribution of relative quality of the exported products of one particular country, Latvia, to the EU market evolved over the period 2003-2013.

Figure 2. Distribution of the quality index for products exported by Latvia with respect to the EU28 average

Note: When the quality index is equal to 1, the quality of the CN8 Latvian product has the same level as the EU28 average. Higher values indicate better qualities.

The distribution of quality was normalised with respect to the EU average in the same period. The plot clearly suggests a shift of the quality distribution of exported products to the right, which corresponds to an improvement of the quality distribution for Latvia in that period. The evolution of the quality index for Latvia is consistent with and offers a potential explanation for the evolution of Latvia’s export market shares, which have been rising over the same period (see Figure 3).

Figure 3. Market share of Latvian exports in total EU28 imports

This simple illustration highlights the usefulness of this new quality metric for both academic and policy-related work and warrants further investigation. The construction of taste indices is more demanding in terms of data and thus still awaits empirical implementation.


Baldwin, R. and J. Harrigan 2011. "Zeros, Quality, and Space: Trade Theory and Trade Evidence", American Economic Journal: Microeconomics, 3(2): 60-88

Broda, C. and Weinstein, D. E. (2006) "Globalization and the gains from variety", The Quarterly Journal of Economics 121: 541-585.

Dhingra, S. (2013) "Trading away wide brands for cheap brands", American Economic Review 103: 2554-84.

Di Comite, F., Thisse, J.-F. and H. Vandenbussche (2014) "Verti-zontal differentiation in export markets", Journal of International Economics 93: 50-66.

Di Comite, F. (2012) "Measuring quality and non-cost competitiveness at a country-product level. European Economy", Economic Papers 467, Directorate General Economic and Financial Affairs (DG ECFIN), European Commission.

Eckel, C., Iacovone, L., Javorcik, B. and J. P. Neary (2015) "Multi-product firms at home and away: Cost- versus quality-based competence", Journal of International Economics 95: 216-32.

Eckel, C. and J. P. Neary (2010) "Multi-product firms and flexible manufacturing in the global economy", Review of Economic Studies 77: 188-217.

Feenstra, R. C., 1994. "New Product Varieties and the Measurement of International Prices," American Economic Review 84(1): 157-77.

Foster, L., Haltiwanger, J. and C. Syverson (2008) "Reallocation, firm turnover, and efficiency: Selection on productivity or profitability?", American Economic Review 98: 394-425.

Hottman, C., Redding, S.J. and D.E. Weinstein (2016) "Quantifying the sources of firm heterogeneity", Quarterly Journal of Economics, forthcoming.

Khandelwal A. (2010) "The long and short (of) quality ladders", Review of Economic Studies 77: 1450-76.

Mayer, T., Melitz, M.J. and G.I.P. Ottaviano (2014) "Market size, competition, and the product mix of exporters", American Economic Review 104: 495-536.

Melitz, M.J. and G.I.P. Ottaviano (2008) "Market size, trade, and productivity", Review of Economic Studies 75: 295-316.

Ottaviano G.I.P., Tabuchi, T. and J.-F. Thisse (2002) "Agglomeration and trade revisited", International Economic Review 43: 409-36.

Roberts, M. J., Xu, D. Y., Fan, X. and S. Zhang (2012) "A structural model of demand, cost, and export market selection for Chinese footwear producers", NBER Working Papers, 17725.

Tirole, J. (1988). The Theory of Industrial Organization. Cambridge, Mass., MIT Press.

Vandenbussche, H. (2014) "Quality in Exports", European Economy, Economic Papers 528, Directorate General Economic and Monetary Affairs (DG ECFIN), European Commission.

Verhoogen, E.A. (2008) "Trade, quality upgrading, and wage inequality in the Mexican manufacturing sector", Quarterly Journal of Economics 128: 489-530.


[1] The papers cited above use a constant-elasticity-of-substitution (CES) demand. Other papers using different demand preferences include Verhoogen (2008), Khandelwal (2010), Roberts et al. (2012), Dhingra (2013), as well as Eckel et al. (2015).

[2] It is difficult to compare the quality of any two varieties of the same product, unless data on costs or input prices are available, like in Verhoogen (2008).

[3] Hottman et al. (2016) find that 50% to 75% of the variance in US firm size can be attributed to differences in firm appeal; 23% to 30% to differences in product scope; and less than 20% to average marginal cost differences.

1,470 Reads