Income-induced expenditure switching: Supermarket scanner data evidence from the 2008-09 Crisis in Latvia

Rudolfs Bems, Julian di Giovanni

08 September 2014



Conventional macroeconomic models posit that the relative price is the sole driver of expenditure switching – a country’s imports fall and its exports rise in proportion to its relative price change vis-à-vis that of the rest of the world (Corcos et al. 2012, Nitsch and Pisu 2008). In Bems and di Giovanni (2014), we revisit external adjustment and relative price changes during a sudden stop episode empirically, and show that an income effect can also play an important role in expenditure switching.

We examine the dramatic 2008-09 sudden stop crisis in Latvia, which stands out in terms of its size – GDP in 2008-09 fell by 25% – as well as the unconventional policies pursued in response to the crisis. Specifically, the country chose to keep currency pegged to the euro at its pre-crisis rate. Still, within two years, a 20% of GDP trade deficit was reduced to balanced trade and GDP growth resumed (see Figure 1). Latvia’s experience is of interest because it is one of the few examples where large external adjustment was achieved without a nominal devaluation – to the surprise of many economists – and faster than expected. Latvia’s crisis episode can also potentially offer new insights about the painful process of rebalancing within a monetary union.

We focus on the drastic adjustment in imports, which accounted for the bulk of the external adjustment (see Figure 1). How much did changes in relative prices contribute to the adjustment? Can it be explained with income compression? To answer these questions, we study consumers’ supermarket purchases of domestic and imported items using a unique scanner-level dataset covering 2006Q2-2011Q1. These data provide both prices and quantities at the individual item level, and crucially identify the country origin of each item. The key advantage of the dataset is that it allows for a direct measurement of expenditure switching and relative prices with internally consistent data. We find the dataset to be representative of aggregate expenditures on food, which constitutes more than a quarter of household expenditures in Latvia.

Figure 1. Latvia’s sudden stop episode

To understand the behaviour of imports, we first compute the contribution of expenditure switching from imported to domestic goods to the total fall in imports. Second, we examine the margins – ‘across’ or ‘within’ product groups – at which expenditure switching and relative price changes took place. Finally, we ask whether the observed relative price changes explain the observed expenditure switching through the lens of conventional macro models, or whether other channels are required to match the data.

Empirical findings

We establish several empirical findings about the adjustment:

  • There was expenditure switching from imported to domestic goods during the crisis.

Expenditure switching accounted for one third of the total 26% contraction of imports. The other two thirds were due to a proportional fall in domestic and imported goods in response to the crisis-induced fall in aggregate income. The observed expenditure switching is summarised in Panel (a) of Figure 2, where the solid line depicts year-on-year (y-o-y) changes in the total import expenditure share. In 2009, relative to 2008, 3% of food expenditures were reallocated from imports to domestic goods.

  • Expenditure switching took place mostly within narrowly defined product groups.

To convey this finding, Panel (a) in Figure 2 further decomposes total expenditure switching into contributions from substitution:

  • Between items within narrowly defined product groups
  • Across product groups (with different import intensities)

The bulk of expenditure switching is explained by substitution between items within product groups, as consumers substituted from imported items (e.g. chocolate candies, beer, and herb teas) to similar domestic products.

Figure 2. Expenditure switching and relative price adjustment: Total and within/across components

  • There was no systematic change in the relative price of imports within product groups.

A corresponding decomposition of the relative price of imported food (relative to food CPI), reported in panel (b) of Figure 2, reveals that the change in the relative price of imports (a modest 4.4% y-o-y appreciation in 2009) was driven entirely by changes in prices across broad product groups.

From the standpoint of the conventional macro model, the within/across product group asymmetry between expenditure switching (i.e., panel (a) in Figure 2) and import price changes (i.e., panel (b) in Figure 2) presents a puzzle. Why did consumers switch to domestic substitute items within product groups, if such items did not become less expensive than their imported counterparts? Our proposed answer focuses on the shift in consumed item mix within product groups.

  • Consumers substituted from expensive imported items to cheaper domestic alternatives.

We find that within narrow product groups imported items are, on average, 30% more expensive than comparable domestic items, and that consumers substituted towards cheaper similar items during the crisis. This substitution generated expenditure switching from imported to domestic items without the need for any adjustment in relative prices. Figure 3 depicts this result by plotting an index that measures switching between high and low unit values within product groups. Negative (positive) values imply that consumers are switching towards cheaper (more expensive) items. We find that there was an across-the-board switching towards cheaper items during the crisis, including from imported to domestic items. This finding is consistent with the flight-from-quality hypothesis put forth in Burstein et al. (2005), who present facts on consumer shopping patterns during the 2001 Argentinean crisis.

Figure 3. Switching between cheap/expensive items

Quantifying the drivers of expenditure switching

Motivated by the empirical findings, we next:

Quantify contributions of relative prices and substitution towards cheaper items to the observed expenditure switching, and;
Link the consumer’s substitution behaviour to the observed fall in aggregate income.

We model an expenditure allocation problem of a representative consumer. Given item-level prices and the crisis-induced fall in income, a consumer decides how to allocate expenditures across and within product groups. The consumer’s choice depends on relative prices as well as an income-driven demand for quality. The latter channel introduces a non-homotheticity, modelled following Hallak (2006): A negative income shock, such as the one faced by consumers in Latvia, generates substitution from expensive to cheap items, irrespective of relative price changes. With quality considerations switched off, the model is a conventional CES demand system.

The model provides us with a simple structure, which we apply in a panel regression setting to item-level data. We estimate model parameters and construct a predicted aggregate measure of expenditure switching over the sample period. Results, presented in Figure 4, are quite striking.

  • First, the CES model with only the relative price channel operative – i.e., the workhorse model in international macroeconomics – performs poorly.  The model’s predicted expenditure switching does not account for any of the switching observed in the data, particularly during the crisis.
  • Second, the non-homothetic model with income-induced expenditure switching operative fares much better. It captures 80% of the expenditure switching in the data.
  • Further, we show that it is the income channel, not relative prices, which drives expenditure switching in the non-homothetic model.

Figure 4. Import expenditure share: Data and model predictions

Conclusions and policy implications

These results offer two key takeaways. We find that:

  • The conventional relative price channel did not contribute significantly to expenditure switching on the import side.
  • Expenditure switching was instead a result of income-induced expenditure switching, as consumption of within-group item mix shifted towards cheaper domestic substitutes.

Our findings raise several interesting research/policy questions. First, is our documented income-induced expenditure switching for supermarket items in Latvia a more general phenomenon extending across sectors and countries? Second, our findings suggest that, when induced by income rather than relative prices, expenditure switching (and resulting external adjustment) does not necessarily constitute rebalancing. As income increases back to the pre-crisis levels, expenditure switching is reversed and external imbalances can re-emerge. Next, there is need for a better understanding of what drove changes (or lack thereof) in the relative price of domestic and foreign goods following Latvia’s internal devaluation. Finally, future work should examine welfare implications of the observed expenditure switching. It would be interesting to investigate how costly the substitution to lower quality goods was in welfare terms, given the dramatic fall in income during the crisis.


Bems, R and J di Giovanni (2014), “Income-Induced Expenditure Switching”, CEPR DP# 9887, March.

Burstein, A, M Eichenbaum, and S Rebelo (2005), “Large Devaluations and the Real Exchange Rate”, Journal of Political Economy, 113(4): 742-784.

Corcos, G, M Del Gatto, G Mion, G IP Ottaviano (2012), “Productivity and firm selection: Quantifying the ‘new’ gains from trade”,, 10 July

Hallak, J C (2006), “Product Quality and the Direction of Trade.” Journal of International Economics, 68(1): 238-265.

Nitsch, V and M Pisu (2008), “The euro and trade: New evidence”,, 15 July



Topics:  Europe's nations and regions Macroeconomic policy

Tags:  expenditure switching, income effect

Economist, Research Department, IMF

ICREA Research Professor in the Department of Economics and Business, Universitat Pompeu Fabra