Abstract
Selected Issues
The Impact of Exchange Rate Movements on Consumption in Uruguay1
This paper investigates the impact of exchange rate movements on private consumption in Uruguay. We show that a large share of Uruguayan households is liquidity constrained, which allows the transitory real income shocks brought about by exchange rate pass-through to have a significant impact on consumption. Moreover, exchange rate pass-through is highly heterogenous, with relative prices of durables increasing (decreasing) following a depreciation (appreciation). This creates incentives for households to engage in intertemporal substitution where they buy durables when they are relatively cheaper. The results offer a potential explanation for the often noted ‘excess volatility of consumption’ in emerging markets for the case of Uruguay.
A. Introduction and Overview
1. Anecdotal evidence suggests that the relationship between exchange rate movements and consumption is highly negative in the case of Uruguay. As a motivation, consider the below figure which shows the relationship between changes in the exchange rate (an increase represents a depreciation) and changes in private consumption. A 10 percent depreciation is associated with a reduction in consumption growth of roughly 1 percent for two quarters. Calculating equivalent elasticities for Brazil, Chile, Colombia, Mexico, and Peru we find that only in Peru is the elasticity of a similar magnitude, while it is essentially zero in both Brazil and Colombia.
2. Uruguay is a highly-dollarized economy, which makes the relationship between exchange rate movements and private consumption particularly complex. Dollarization plays an important role in several aspects of the Uruguayan economy—both in the financial system with deposit dollarization above 70 percent and credit dollarization of around 50 percent, and in the real economy, where about 14 percent of the CPI basket consists of goods which are priced in U.S. dollars.2 Goods priced in dollars are overwhelmingly tradable durables and have a much higher pass-through than others.
3. We find two main explanations for the impact of exchange rate movements on private consumption in Uruguay. The first is a negative impact of depreciations on real incomes. Nominal wages are fixed in the short-run and with an immediate exchange rate pass-through of around 15 percent, real wages temporarily fall. A high share of Uruguayan households is liquidity constrained leading them to adjust consumption in response to temporary income shocks. Second, we show that exchange rate pass-through by good is very heterogenous, and in particular it is close to 100 percent for durable goods. Depreciations thus act as a price hike on durable goods and durable consumption reacts very strongly to exchange rate movements possibly indicating inter-temporal substitution by households.
4. We do not find much evidence for exchange-rate-induced wealth effects. To look for wealth effects we start from an exercise by Lluberas and Odriozola (2015) who show that even though saving occurs overwhelmingly in USD, even large exchange rate movements have only a small impact on liquid wealth—largely because so few households have liquid savings. Nevertheless, we cannot exclude that wealth effects matter, especially since houses are priced in USD and many Uruguayans own their home. We can only note that wealth effects would go in the opposite direction of the other effects. Given that overall depreciation (appreciations) are associated with less (more) consumption, the wealth effect thus cannot be strong enough to offset them.
The Relationship between USD Exchange Rate and Private Consumption: Stylized Fact
Citation: IMF Staff Country Reports 2018, 024; 10.5089/9781484339824.002.A001
Source: IMF Staff Estimates. Note The graph shows reduced-form impulse response functions from a binary VAR in first-differences for (log) Private consumption and the de-trended (log) Exchange Rate. Results are virtually indistinguishable when we use the raw exchange rate data instead of the detrended one but detrending accounts for the fact that agents are likely to expect a continued slow depreciation of the Peso given the large inflation differential with the US. Data are quarterly and lag-length of the VAR is 2. Private consumption data from the national accounts is seasonally adjusted using the Census Bureau X-13 method. See data Appendix for details.5. We also do not find much support for a change in expectations of permanent income as a consequence of exchange rate movements. We use data from the consumer confidence survey and show that the exchange rate does not seem to be related to households’ perceptions of their own economic situation nor of the economic situation of the country. We also show that even when controlling for terms-of-trade shocks (as a proxy for real shocks which affect permanent income) exchange rate movements have a strong negative impact on consumption.
6. The negative impact of the exchange rate on consumption seems to carry over to a negative effect on overall GDP. GDP could be largely unaffected by the exchange rate driven volatility in consumption (or even increase) if nearly all the consumption movements were for imported goods.3 However, using the latest available Input-Output tables we show that a non-trivial fraction of domestically consumed durables good are produced in Uruguay so that the fall in durable consumption also affects domestically produced goods. In a simple VAR setting we show that the overall impact of exchange rate movements on GDP is negative.4
7. The remainder of this paper proceeds as follows. Section B briefly lays out the related literature and a few theoretical considerations to provide an underlying structure to the paper. Section C then uses the 2014 Uruguayan Household Finance Survey (Encuesta Financiera de los Hogares Uruguayos or EFHU-2) to lay out salient features of the household wealth, income and consumption distributions in Uruguay which relate to the theoretical considerations of section B and which will be relevant for the subsequent analysis. Section D shows that the exchange rate does indeed impact private consumption in Uruguay. This is done by estimating a simple Vector-Error Correction Model (VECM) for aggregate private consumption in Uruguay. The core of the analysis is presented in section E which evaluates in some granularity the possible channels through which exchange rate movements might be affecting consumption in Uruguay.5 Last, section F discusses macro implications of the analysis for GDP and for the volatility of aggregate consumption. The Appendix contains a list with data sources and additional material.
B. Literature and Theoretical Considerations6
8. The basis for work on private consumption remains Friedman’s permanent income hypothesis which predicts that households should smooth transitory income fluctuations and react to permanent ones.7 Many modern consumption models following Zeldes (1984) build on this but model optimal consumption under uncertainty rather than perfect foresight. These models yield the insight that uncertainty implies a concave consumption function, with a higher marginal propensity to consume (MPC) at low levels of wealth. The baseline result that transitory income shocks should be saved to a large degree, carries over to these models under uncertainty.
9. Empirical evidence has found consistent evidence that the MPC out of transitory income shocks is excessively high relative to those implied by standard models.8 Assuming a sizeable share of households which spend all of their available cash every period (so called hand-to-mouth or HtM consumers) has become a prominent way to achieve a higher MPC in theoretical models.9 Given that HtM households sit at a kink in the intertemporal budget constraint, they fully reflect even transitory income shocks in current consumption.
10. Recent work has shown that across developed countries a large share of the population is liquidity constraint with virtually no savings to speak of and a high MPC. The usual approach to measure HtM households in the data used to be to look for individuals with no or very little net wealth. Measured this way, however, the number of such households is too low to explain a high aggregate MPC out of transitory shocks. In a recent paper, Kaplan et. al (2014) suggest that net wealth is the wrong measure to detect HtM consumers and that rather net liquid wealth should be used—since with some transaction costs, illiquid wealth is not available to smooth small to medium size transitory shocks. They present comprehensive empirical evidence that large numbers of de facto liquidity constrained consumers exist in the U.S. and other advanced economies. They also show that the majority of them are what they call “wealthy hand-to-mouth” consumers—they own illiquid assets but no or very little liquid ones and they would thus not be considered HtM when using a standard net wealth measure. Kaplan and Violante (2014) develop a theoretical model to show that consumers might optimally be wealthy HtM - they invest in illiquid but higher return assets and tolerate the higher volatility across periods to achieve a higher level of lifetime consumption.10 The authors then show that both poor and wealthy HtM consumers have much higher MPCs than non-HtM consumers.
11. The next section will show that liquidity constraint ‘hand-to-mouth’ consumers constitute a very large share of Uruguayan households. As a consequence we might expect there to be a high MPC out of transitory income shocks. This provides for a powerful channel through which the exchange rate (via pass-through to prices) can impact consumption in Uruguay.
12. For the remainder of the paper it will also be important to differentiate between durable and non-durable goods consumption. Durable goods are goods which do not immediately depreciate —the lower the depreciation rate, the more durable the good is (examples include cars, furniture and electronics).11 A key insight is that durable expenditure is more volatile than non-durable expenditure. Consumers utility is increasing in the flow of nondurables and the stock of durables which implies that when nondurable consumption and expenditure adjusts, durable consumption adjust proportionally but durable expenditure will adjust by more to immediately get to the new desired stock of durables.12 This also implies a higher elasticity of demand (price and income) for durables than non-durables.
13. Households face a type of inventory problem for durable goods. Hendel and Nevo (2013) show that for storable goods (which include durables but are broader—eg, a Coca-Cola is storable) consumers face a type of inventory problem—and they optimally buy more than they consume when prices are low and vice versa. Hendel and Nevo also find that this is rationally true for short-term price drops such as sales. In the case of Uruguay, we will show that exchange rate movements have an extremely high pass-through to durable goods prices so that appreciations could potentially be viewed as a ‘sale’ for durables while a depreciation is the opposite.
C. A Few Stylized Facts on the Distribution of Household Savings, Income, and Consumption
14. In light of the above theoretical discussion, we highlight salient features of Uruguayan households’ balance sheets. For this purpose, we use data from the 2014 household finance survey (EFHU-2) which is ideally suited for such an exercise.13 The EFHU asks participants questions on real assets and related debts, liquid assets, other debts, business ownership, income and employment history, pensions, consumption and use of electronic payments and thus allows a detailed overview of household balance sheets.14 This section is organized in terms of a list of relevant summary statistics followed by key take-aways based on those summary statistics which will be used in later parts of the paper.
Liquid assets
15. Only 18 percent of Uruguayan households have liquid assets and 11 percent of households have total liquid assets over one monthly income. As the below figure shows, savings (liquid assets or “cash-on-hand”) increase strongly with income but only after a certain income threshold—roughly 25,000 Pesos per month (ca. 850 U.S. dollars). Poorer households own very little or no liquid assets. If there are any liquid savings, on average close to 80 percent of them are in the form of bank deposits. On average about 46 percent of household savings (if there are any) are in USD. Also, larger amounts are more likely to be saved in USD so that around 70 percent of the total value of savings is in USD. A corollary of this is that richer households tend to save a somewhat higher fraction of their income in U.S. dollars—doubling income leads to an increase in the fraction of savings in USD of roughly 0.13. However, even very low income households partly save in dollars if they save at all.
Real and illiquid assets
16. Close to 60 percent of households own their home. Out of those 60 percent, roughly 60 percent bought their home, while another 20 percent built it and the remainder inherited it or gained ownership through a gift. Only less than 10 percent of all households report currently paying down mortgage related debt, implying a large share of the population owns housing equity. Roughly 50 percent of households report contributing to a private pension (AFAP) account. Importantly, many household who own real estate or other illiquid assets (such as a private pension account) are ‘cash-poor”. For example, as many as 85 percent of households who own their house have no or very little liquid assets.
17. Consumer durables ownership is very wide spread. Less than 40 percent of households own their own car or vehicle and out of those who do, roughly 30 percent bought a new one in the last year. On the other hand, the very large majority of households own the following durables: computer and cellphone (86 percent), kitchen and other household appliances (99 percent) and video and audio equipment (90 percent).
Debt
18. Thirty five percent of households report having debt of some kind, including mortgage debt. For those households with debt, the median amount owed is roughly equal to 1.7 months of income and the mean amount owed is close to 3.5 months of income. Over 30 percent of households have some debt but no savings. Defining net wealth as the sum of housing equity + liquid savings—all debt, we find that 15 percent of households have negative wealth, 27 percent have zero net wealth and the remainder have positive net wealth.
Consumption
19. Unsurprisingly, the poorer the household, the higher the share of income spent on necessities. The median household spends 62 percent of his income on food, utilities, health and education, with food by far the largest component. Recall that only for households with income above roughly 25,000 Pesos did we start to see a significant positive relationship between savings and income. These are also the households who spend less than 50 percent of their monthly income on necessities. Lastly, about 65 percent of households reported spending all their income over the past 12 months, 16 percent spent more than their income and 19 percent spent less.
Implications
20. The above observations highlight several important features of consumers in Uruguay:
There exists a large number of HtM households in Uruguay. Furthermore, a large fraction of HtM households are ‘wealthy hand-to-mouth’—they own a sizeable amount of illiquid assets (eg housing or retirement accounts) but no liquid savings. Kaplan et al. (2014) define a “poor hand-to-mouth” household as one who has less than half a month of income of liquid assets and no illiquid assets whereas they define a household with the same liquid position but some illiquid assets as “wealthy hand-to-mouth”. Using different potential measure for Uruguay we calculate that between 40 and 87 percent of households in Uruguay are HtM, with our preferred estimate-closer to the top of the range.15 This compares to numbers around 30 percent for advanced economies.16 Furthermore, we find that depending on which illiquid assets we count (housing equity or private pension accounts or both), between one half and four fifths of all HtM households are wealthy HtM.17
All households who save do so partly in dollars. The share of dollar saving in total savings increases with household income.
Access to credit appears very limited with credit constraints likely binding for many households. Note that average real interest rates on consumer loans tend to exceed 40 percent in Uruguay.
Households spend a large fraction of their income on necessities. Cars seems to be unaffordable to a significant portion of households but household appliances and other smaller durables are consumed by virtually everybody.
Household wealth and income in Uruguay
(in Pesos)
Citation: IMF Staff Country Reports 2018, 024; 10.5089/9781484339824.002.A001
Source: IMF Staff calculations based on EFHU-2. Note: Graph shows scatter plot with observations clustered into twenty equal sized bins based on monthly household income where each dot represents mean monthly household income and mean total savings within one of the twenty bins.Household income and spending on necessities’ in Uruguay
Citation: IMF Staff Country Reports 2018, 024; 10.5089/9781484339824.002.A001
Source: IMF Staff calculations based on EFHU-2. Note: Graph shows scatter plot with observations clustered into twenty equal sized bins based on monthly household income where each dot represents mean monthly household income and mean share of income spent on food, utilities, health and education within one of the twenty bins.Dollar saving and household income
Citation: IMF Staff Country Reports 2018, 024; 10.5089/9781484339824.002.A001
Source: IMF Staff calculations based on EFHU-2. Note: Graph shows scatter plot with observations clustered into twenty equal sized bins based on (log) monthly household income where each dot represents mean (log) monthly household income and mean share of savings in USD within one of the twenty bins.D. A Simple VECM for Private Consumption
21. To investigate the aggregate relationship between the exchange rate and aggregate private consumption we estimate a simple vector error correction model (VECM). While we are primarily interested in verifying the existence of a short-run link between the exchange rate, real income and private consumption we estimate a VECM here to take into account the well-established long-run relationship between real income and consumption.18 In later sections of the paper we will estimate VARs in first differences to focus exclusively on the short-run dynamics. In practice, results from the VECM and first-difference VARs are qualitatively the same, except that the VECM forces a certain degree of persistence in the effect of shocks due to the included long-run relationship. Note that we use the bilateral U.S. Dollar exchange rate rather than the nominal effective one, since the pass-through from the US dollar one is higher. This might be explained by wide-spread USD pricing even for imports from neighboring countries (see recent work by Boz et al, 2017).
22. Stationarity tests show that the series are I(1) and co-integrated with rank 1. We use augmented dickey-fuller tests to test for non-stationarity and cannot reject the null hypothesis for any of the series.19 Johanson likelihood-ratio tests for co-integration indicate the presence of one co-integrating vector. Standard information criteria suggest 2 to be the optimal lag length. The VECM is identified using a Cholesky decomposition, with consumption being the most endogenous variable. The preferred ordering is: et → yt → ct but results are virtually invariant to the ordering.20
23. Private consumption increases as a reaction to a positive shock to real income and decreases following a positive shock (depreciation) to the exchange rate. The below figure plots the orthogonalized impulse-response functions for one-standard deviation shocks to real income and the exchange rate. Both shocks have the expected sign.21 The results indicate a causal impact of the exchange rate on consumption both via real wages and directly.
Impact of Shocks on Consumption
(from VECM)
Citation: IMF Staff Country Reports 2018, 024; 10.5089/9781484339824.002.A001
Source: IMF Staff estimates. Note: Graphs show orthogonalized impulse response functions from a trivariate VECM (log private consumption, log real income index and log exchange rate).E. The Impact of Exchange Rate Movements on Private Consumption: Exploring the Channels
Real income effects
24. The short-run pass-through of exchange rates to prices is estimated at around 15 percent. To calculate pass-through of exchange rate movements we estimate a simple infinite distributed lag model
where π is CPI inflation and we control for U.S. inflation and movements in international oil prices in Ot. All variables are specified as logarithms. Short-run pass-through is given by β while the long-run pass-through is given
25. Wages do not generally react within the same or even the next quarter, leading to a temporary loss of real income in case of a depreciation. This is true even though there used to be a high degree of backwards indexation of wages in Uruguay (less so after the 2015–16 wage negotiations). Re-estimating equation (I) with real wages as the dependent variable, we find a short-run elasticity of −13 percent, indicating minimal immediate reaction of nominal wages. Exchange rate movements thus impact real wages in Uruguay. And in a setting with a large fraction of HtM consumers such a transitory income shock impacts consumption.
26. Anecdotal evidence suggests that households often operate in USD savings targets, potentially leading them to increase their savings rate following a depreciation even if not liquidity constraint. If indeed households have dollar savings targets, then changes in the exchange rate impact the required savings rate (out of Peso salaries) to be able to reach the savings target. This could potentially be an additional channel which would lead non-liquidity constraint households to adjust consumption. However, it is not clear how much of an impact on required savings rates a temporary income shock would have.
Uncertainty and expectations of permanent income
27. We look at whether exchange rate movements might impact consumers’ expectations of uncertainty or permanent income. There are several plausible hypothesis which link exchange rate movements to permanent income and uncertainty given that many exchange rate movements are driven by real sector shocks such as terms-of-trade shocks. For example, thinking about the recent (permanent) ToT shock which hit much of Latin America and which led to large depreciations of local currencies, agents could associate depreciations with a reduction in permanent income (and vice versa for appreciations). Similarly, Uruguay’s history of crisis, which were accompanied by rapid depreciations, might lead households to associate appreciations with good times and depreciations with bad times. Expectations of lower permanent income or perceptions of higher uncertainty would then lead consumers to lower consumption today in the standard macro-consumption framework.
28. We purge’ exchange rate movements from real sector shocks by controlling for terms-of-trade movements and find that the residual variation in exchange rates still matters for private consumption. To do so we estimate a trivariate VAR which includes terms of trade movements, exchange rate movements and private consumption growth, with the terms-of-trade ordered first and the exchange rate second in the Cholesky decomposition. The forecast-error decomposition shows that shocks to the exchange rate not driven by terms-of-trade movements still account for 22 percent of the variation in private consumption growth. We also find that terms of trade shocks only account for 15 percent of the variation in exchange rate movements. Taken together this offers some suggestive evidence that exchange rates movements are not simply a proxy for real sector shocks.
29. We also use data from the consumer confidence survey and find no indication for an impact of the exchange rate on proxies for uncertainty or expected permanent income once other factors are controlled for. We use survey data on households’ assessment and outlook for their personal situation and their outlook for the country’s situation to proxy for expectations of uncertainty or permanent income.23 We find that current real wages are the key determinant of households’ perceptions of their personal economic situation. The interest rate on the other hand influences perceptions of the country’s economic situation. The exchange rate, however, does not influence perceptions of either the personal or the country’s economic situation.24
Determinants of consumer confidence components
Determinants of consumer confidence components
| (1) | (2) | (3) | |
|---|---|---|---|
| VARIABLES | Personal Situation | Country’s Situation | Intention to buy Durables |
| Employment (SA, %) | 1.088 (0.763) |
0.650 (1.311) |
1.769 (1.080) |
| Real Wage Index | 1.140*** (0.362) |
0.418 (0.555) |
−0.163 (0.445) |
| Peso/USD Exchange Rate | −0.0824 (0.250) |
0.395 (0.426) |
−2.195*** (0.387) |
| Interest Rate (12 months) | −0.133 (0.230) |
−0.810** (0.301) |
−1.182*** (0.304) |
| Constant | 117.2* (61.73) |
111.6 (107.4) |
−38.48 (82.37) |
| Observations | 39 | 39 | 40 |
| Linear Time Trend | Yes | Yes | Yes |
| R-squared | 0.543 | 0.409 | 0.864 |
Determinants of consumer confidence components
| (1) | (2) | (3) | |
|---|---|---|---|
| VARIABLES | Personal Situation | Country’s Situation | Intention to buy Durables |
| Employment (SA, %) | 1.088 (0.763) |
0.650 (1.311) |
1.769 (1.080) |
| Real Wage Index | 1.140*** (0.362) |
0.418 (0.555) |
−0.163 (0.445) |
| Peso/USD Exchange Rate | −0.0824 (0.250) |
0.395 (0.426) |
−2.195*** (0.387) |
| Interest Rate (12 months) | −0.133 (0.230) |
−0.810** (0.301) |
−1.182*** (0.304) |
| Constant | 117.2* (61.73) |
111.6 (107.4) |
−38.48 (82.37) |
| Observations | 39 | 39 | 40 |
| Linear Time Trend | Yes | Yes | Yes |
| R-squared | 0.543 | 0.409 | 0.864 |
Relative price effects and the consumption of durables
30. Intentions to buy durable goods, the third element of the consumer confidence survey, are very strongly related to the exchange rate. In this section we analyze the link between durable consumption and the exchange rate in more detail and argue that the relative increase (decrease) in durables prices brought about by a depreciation (appreciation) is key for understanding the impact of exchange rate movements on consumption.
31. As a first step it is worth noting that about 14 percent of the CPI basket are quoted in USD in Uruguay and that the overwhelming part of these goods are durables with a very high exchange rate pass-through. To see the high pass-through consider the below figure which shows the distribution of estimated short-run pass-through coefficients for all goods in the CPI basket. The majority of items have a very low pass-through, indeed often not significantly different from zero. On the other hand, there is a small mass of items which have a pass-through close to 1—precisely the durables mentioned previously (households appliances, cars, etc.).25 When we estimate pass-through by group of item (tradable and durable, tradable and non-durable and non-tradable) we find that the estimated pass-through for both the latter two groups is not statistically significantly different from zero.
Exchange rate pass-through by item
Citation: IMF Staff Country Reports 2018, 024; 10.5089/9781484339824.002.A001
Source: IMF Staff Estimates. Note: Figure shows estimated pass-through coefficients for all goods in the CPI basket.32. Exchange rate movements thus have important implications for relative prices. Clearly, relative prices between durables and the rest of the consumption basket change with the exchange rate. Additionally, as long as exchange rate movements are perceived to be temporary (we control for a linear time trend, so that exchange rate movements can be seen as deviations from trend), there is a change in relative durables prices today versus durables prices in the future.
33. Indeed, durable purchases in Uruguay are highly influenced by exchange rate (and thus price) movements. Plotting the two shows just how closely sales of cars are aligned with exchange rate movements on a quarterly basis. Car sales track the exchange rate in virtually every quarter of the sample.
34. It is not possible to formally assign weights to the real wage effect and the intertemporal substitution effect in terms of their importance for private consumption movements. While we find tentative evidence for both, it is hard to rule out that it is ‘only’ the real income effect which causes changes in durable (and overall) consumption or that ‘only’ intertemporal substitution of durables is responsible. The key to be able to comment further would be more disaggregated data, notably more disaggregated consumption data.
The Exchange Rate and Car Sales
Citation: IMF Staff Country Reports 2018, 024; 10.5089/9781484339824.002.A001
Source: IMF Staff calculations. Note: Figure shows co-movements between the exchange rate (de-trended) and car sales.Wealth effects
35. Exchange rate movements also have direct wealth effects given the high dollarization of assets. This effect is of the opposite sign relative to the other effects discussed previously—depreciations increase the value of dollar assets. For households with a significant fraction of savings in USD this could potentially provide a substantial boost to real wealth.26
36. It is not possible to fully explore this channel, but at least the impact on liquid wealth appears small. As discussed previously, few households have liquid savings so that liquid wealth effects will only impact a small segment of the population—what is more, it would impact only the wealthiest segment of the population which is likely to have the lowest MPC. Indeed, Lluberas and Odriozola (2015) simulate the wealth implications of exchange rate movements based on data from the 2014 EFHU-2 and find that even very large movements of over 50 percent have only small implications for the distribution of wealth. Taking into account illiquid assets complicates the analysis, given that a much larger share of the population own illiquid assets and houses are generally valued in USD. Unfortunately, without more data we are not able to dig any deeper—we can only observe that if wealth effects exist, they are not strong enough to offset the income and substitution effects.
F. Macro-Implications: GDP, Volatility of Consumption and Concluding Thoughts
The implications for GDP
37. Thinking more broadly about the link between exchange rate movements and GDP, the baseline view in much of macro is that exchange rate depreciations tend to be expansionary. The standard channel refers to expenditure switching effects, whereby exchange rate movements stimulate a move away from imports and towards domestically produced goods, while at the same time stimulating exports.
38. Following Krugman and Taylor (1978), among others, the possibility of contractionary depreciations has also been acknowledged. The mechanisms which can lead to a contractionary effect include a consumption channel such as the one we investigate in this paper—redistribution of income to agents with a low marginal propensity to consume, for example a rise in profits at the cost of real wages—or negative balance sheet effects in the presence of FX liabilities (see, among others Moreno (1999), Kandil et al (2007) and IMF (2009)). The consumption channels we highlighted in this paper will negatively affect GDP unless the overall fall in consumption is fully offset by a substitution away from imports.
39. Data from Input-Output tables show that Uruguay produces a non-trivial amount of the tradable, durable goods it consumes, opening the door to contractionary depreciations. Recall that tradable durables are the goods with by far the biggest relative price changes following exchange rate movements and they are likely to be driving most of the fall in total consumption. If all these goods were imported, then the impact on GDP might be modest at most. The below table uses the latest available (2013) Input-Output table in constant 2005 Pesos to highlight that while the majority of durable goods are imported, the domestic share is non-trivial. For furniture, it is as high as 40 percent, for example.
Durable and Storable Goods in Input-Output Tables
Durable and Storable Goods in Input-Output Tables
| Data from 2013 Partial Input-Output Table Uruguay | Share of sector in total household consumption | Share of Supply which is domestic | Share of Household consumption in final demand |
|---|---|---|---|
| Metals, machines and electronic products such as TVs and Radios | 4% | 19% | 18% |
| Cars and other vehicles | 4% | 15% | 40% |
| Clothing | 4% | 38% | 91% |
| Furniture | 2% | 43% | 54% |
Durable and Storable Goods in Input-Output Tables
| Data from 2013 Partial Input-Output Table Uruguay | Share of sector in total household consumption | Share of Supply which is domestic | Share of Household consumption in final demand |
|---|---|---|---|
| Metals, machines and electronic products such as TVs and Radios | 4% | 19% | 18% |
| Cars and other vehicles | 4% | 15% | 40% |
| Clothing | 4% | 38% | 91% |
| Furniture | 2% | 43% | 54% |
40. Intratemporal substitution between imported and domestically-produced goods is non-neglible which should limit the adverse effect on demand. The below figure shows that there is at least some substitution, however, as would be expected. The ratio of imported consumer goods to total consumption expenditures increases when the exchange rate appreciates and falls when it depreciates27. At the very least, the effect of lower total consumption is thus partially attenuated by lower imports.
41. When including GDP in the VAR analysis we find a negative association between depreciations and GDP growth. Specifically, we add the growth rate of GDP to the trivariate VAR used in section E (which included the growth rate of private consumption, terms of trade and de trended exchange rate) and recover the orthogonalized impulse response functions shown in the bottom figure. The results show that GDP is negatively impacted by exchange rate movements but quantitatively somewhat less so than private consumption. Both GDP and private consumption are negatively affected in the contemporaneous quarter and the next but after that the impact is not statistically significantly different from zero. We do not attempt to fully explore the channels through which exchange rate movements impact GDP—for example, we have not commented at all on the reaction of exports—but finding a negative aggregate relationship highlights the importance of the consumption channel which we explored in this paper. Finding a smaller quantitative impact for GDP than private consumption additionally suggests that there is some offset—import substitution is a likely candidate as a countervailing force.
Intratemporal expenditure switching
Citation: IMF Staff Country Reports 2018, 024; 10.5089/9781484339824.002.A001
Source: IMF Staff Calculations. Note: Figure shows co-movements between the exchange rate (de-trended) and the ratio of the consum ption goods im port index and total private consumption (in constant 2005 pesos). The level of the ratio cannot easily be interpreted since the units of the nominator and denominator are unrelated.Impact of Exchange Rate Shock GDP
Citation: IMF Staff Country Reports 2018, 024; 10.5089/9781484339824.002.A001
Source: IMF Staff estimates. Note: The underlying VAR includes the change in (log) GDP, (log) private consumption, (log) terms of trade and de-trended (log) exchange rate. Shocks are identified using a Cholesky decom position with terms of trade ordered first, followed by the exchange rate, private consumption and GDP last.Excess volatility of consumption in emerging markets—a potential explanation for Uruguay
42. It is a well-known fact that volatility of consumption in emerging markets is higher than in advanced economies (Alvarez-Parra et al., 2013). In addition, consumption expenditure volatility in emerging markets tends to exceed that of income—the so-called “excess volatility of consumption puzzle” (Aguiar and Gopinath, 2007). However, once durable consumption is excludedthis is not true anymore (Alvarez-Parra et al., 2013).28 For Uruguay, the ratio of the standard deviation of quarterly consumption to the standard deviation of quarterly GDP (calculated following the methodology29 of Alvarez-Parra et al. (2013) is 1.28. This is towards the higher end of the emerging markets sample and above the average of 1.14 calculated in their paper.
43. Alvarez et al. (2013) suggest that counter-cyclical interest rates which drive durables consumption are key to explaining excess volatility of consumption in emerging markets. In a standard RBC setup frequent persistent shocks to productivity are needed to explain excess volatility of consumption (Aguiar and Gopinath, 2007). But Alvarez et al. (2013) show that introducing durable consumption combined with financial frictions into a RBC model reduces the need for frequent productivity trend shocks. Their mechanism runs from a counter-cyclical level of interest rates (financial frictions lead to counter-cyclical risk premia) to pro-cyclical durables consumption (taking advantage of low rates during expansions) and finally to counter-cyclical trade balances (another salient feature of emerging markets).
44. Volatile nominal exchange rates and thus volatile durable consumption (even without counter-cyclical risk premia) could offer a related but complementary explanation. The evidence in the present paper suggests that in the case of Uruguay, the exchange rate could be an alternative mechanism for explaining volatility of expenditures on durables. Access to household credit is very limited, reducing the impact the interest rate can have in Uruguay. On the other hand, the evidence presented in this paper suggests that volatile exchange rates could be a key driver for consumption volatility.30
References
Alvarez-Parra, F. L. Brandao-Marquez and M. Toledo (2013), “Durable goods, financial frictions, and business cycles in emerging economies”, Journal of Monetary Economics, Vol. 60: pp. 720–736.
Aguiar, M. and G. Gopinath, (2007), “Emerging Market Business Cycles: The Cycle Is the Trend”, Journal of Political Economy, Vol.115: pp. 69–102.
Boz, E., G. Gopinath and M. Plagborg-Moller (2017), “Global Trade and the Dollar”, mimeo.
Carroll, C.D. (2001), “A Theory of the Consumption Function, With and Without Liquidity Constraints”, NBER Working Paper No. 8387.
Carroll, C.D. (2013) “Representing consumption and saving without a representative consumer”, CFS Working Paper No. 464.
Friedman, M.A. (1957), A Theory of the Consumption Function, Princeton University Press.
Gali, J., J.D. Lopez-Salido and J. Valles (2007), “Understanding the Effects of Government Spending on Consumption”, Journal of the European Economic Association, Vol. 5: pp. 227–270.
Hsieh, C. “Do Consumers React to Anticipated Income Changes? Evidence from the Alsaka Permanent Fund”, American Economic Review, Vol. 93: pp. 397–405.
Inter-American Development Bank (2015), “Understanding the Determinants of Household Savings: Micro Evidence for Latin America”, Technical Note Nr IDB-TN-843.
International Monetary Fund (2009), “Crisis and Recovery”, World Economic Outlook April 2009, Washington, DC: International Monetary Fund.
International Monetary Fund (2016) “Managing Transitions and Risks”, WHD Regional Economic Outlook, Washington, DC: International Monetary Fund.
Jappelli, T. and L. Pistaferri (2010), “The Consumption response to Income Changes”, NBER Working Paper No. 15739.
Johnson, D., J. Parker and N. Souleles (2006), “Household Expenditure and the Income Tax Rebates of 2001”, American Economic Review, Vol. 96: pp. 1589–1610.
Kandil, M., H. Berument and N.N. Dincer (2007), “The effects of exchange rate fluctuations on economic activity in Turkey”, Journal of Asian Economics, Vol. 18: pp. 466–489.
Kaplan, G. and G. L. Violante (2014) “A Model of the Consumption Response to Fiscal Stimulus Payments”, Econometrica, Vol. 82: pp. 1199–1239.
Kaplan, G., G.L. Violante and J. Weidner (2014), “The Wealthy Hand-to-Mouth”, Brookings Papers on Economic Activity, Spring 2014: pp. 77–153.
Krugman, P. and L. Taylor (1978), “Contractionary effects of devaluation”, Journal of International Economics, Vol. 8: pp445–456.
Matheson, T. and C. Goes (2017), “Consumption in Brazil: Where to Next?”, IMF Working Paper WP/17/75. Washington, DC: International Monetary Fund.
Mian, A., K. Rao and A. Sufi (2013), “Household Balance Sheets, Consumption, and the Economic Slump”, Quarterly Journal of Economics, Vol. 128: pp. 1687–1726.
Moreno, R. (1999), “Depreciation and recessions in East Asia”, Economic Review - Federal Reserve Bank of San Francisco, Vol. 3: pp. 27–40.
Lluberas, R. and J. Odriozola (2015), “Inflation, currency depreciation and households balance sheet in Uruguay”, Banco Central del Uruguay, Documento de Trabaja No. 2015009.
Obstfeld, M. and K. Rogoff (1996), Foundations of International Macroeconomics, Cambridge, MA: MIT Press.
Toscani, F. (2017) “Balance Sheet Analysis in Uruguay – High Dollarization but Limited Risks”, IMF Selected Issues Paper. Washington, DC: International Monetary Fund.
Universidad de la Republica (2016), “Encuesta Financiera de los Hogares Uruguayos (EFHA-2): Metodologia y Guia para el Usuario”.
Zeldes, S.P. (1984), “Optimal Consumption with Stochastic Income”, Pd.D. thesis, Massachusetts Institue of Technology.
Appendix I. Additional Material and Data Sources
Macroeconomic data
We use data from Q1 2005 to Q1 2017 in the analysis.
Real Private Consumption: Quarterly data are from the national accounts provided by the Uruguayan Central Bank (BCU). We seasonally adjust the data using the Census Bureau X-13 method.
Seasonally adjusted real GDP: Quarterly data are from the BCU.
Exchange Rate: The quarterly bllateral Peso-US Dollar exchange rate is taken from Haver. For much of the analysis we detrend the series by regressing it on a time trend and working with the residuals of that regression.
Real and nominal wages indices: Monthly data are taken from the national statistics institute (INE). We average the data by quarter to obtain quarterly data.
CPI index and price indices by good: Monthly data taken from INE. We average the data by quarter to obtain quarterly data.
Employment: Quarterly employment rate provided by INE. The data are seasonally adjusted by Haver.
Consumer confidence index and sub-indices: Monthly data from Universidad Catholica de Uruguay. We average the data by quarter to obtain quarterly data.
12-month interest rate: Data taken from BCU.
Import volume index and sub-indices: Data taken from BCU.
Car Sales: Monthly data taken from Asociación del Comercio Automotor del Uruguay and then seasonally adjusted and summed over the quarter.
Household Finance Survey (EFHU-2)
We only work with the raw non-imputed data for the 3490 households which participated in the survey. The number of responses varies by households. Generally, when a household did not reply to a specific question we simply exclude it from that part of the analysis. In the following paragraph we outline key data manipulations.
Total liquid savings: To calculate summary statistics on liquid savings we used the responses to the question which asks households to give an estimate of their total savings. If a household reported in a separate question that that they do have some savings but didn’t give an estimate, that household is excluded. If a household reported not having any savings, we impute a 0 for the estimate of their total liquid savings.
USD and Peso liquid savings: To calculate the value of Peso and Dollar savings, we take the reported answers to the questions which ask what fraction of savings are in USD and Peso, respectively, and multiply it with the total liquid savings number.
Consumption: To calculate the share of income spent on essentials we take sum estimated monthly spending on food, utilities, health and education and divide it by reported total household income.
List of USD quoted prices
List of USD quoted prices
| Item | Weight in CPI Basket | Item | Weight in CPI Basket |
|---|---|---|---|
| Air conditioning | 0.0% | Home insurance | 0.1% |
| Armchair | 0.1% | Hotel | 0.4% |
| Bed | 0.1% | Jeans | 0.3% |
| Bed Sheets | 0.2% | Kitchen | 0.1% |
| Bicycle | 0.0% | Kitchen furniture | 0.1% |
| Car | 1.2% | Lawyer’s services | 0.2% |
| Car battery | 0.2% | Lighting appiances | 0.0% |
| Car cavor | 0.2% | Living room set | 0.1% |
| Car insurance | 1.5% | Mattress | 0.3% |
| Car rental | 0.1% | Motorcycle | 0.3% |
| CD player | 0.1% | Motorcycle accessories | 0.0% |
| CDs | 0.1% | Pan | 0.1% |
| Cellphones | 0.3% | Post | 0.1% |
| Childrens’ bicycles | 0.1% | Postgraduate education | 0.1% |
| Compuer desk | 0.1% | Printer cartridges | 0.0% |
| Computer | 0.2% | Printing paper | 0.0% |
| Digital photcamera | 0.0% | Quilts | 0.1% |
| Dining room set | 0.1% | Rent of primary residence | 3.7% |
| Drill | 0.1% | Television | 0.1% |
| DVD player | 0.1% | Toys | 0.1% |
| Electric lamps | 0.1% | Vehicle repair | 1.1% |
| Excursions | 0.4% | Wardrobe | 0.1% |
| Flight tickets | 0.4% | Washing machine | 0.1% |
| Fridge | 0.2% | Water heater | 0.1% |
| Glasses | 0.3% | 13.7% |
List of USD quoted prices
| Item | Weight in CPI Basket | Item | Weight in CPI Basket |
|---|---|---|---|
| Air conditioning | 0.0% | Home insurance | 0.1% |
| Armchair | 0.1% | Hotel | 0.4% |
| Bed | 0.1% | Jeans | 0.3% |
| Bed Sheets | 0.2% | Kitchen | 0.1% |
| Bicycle | 0.0% | Kitchen furniture | 0.1% |
| Car | 1.2% | Lawyer’s services | 0.2% |
| Car battery | 0.2% | Lighting appiances | 0.0% |
| Car cavor | 0.2% | Living room set | 0.1% |
| Car insurance | 1.5% | Mattress | 0.3% |
| Car rental | 0.1% | Motorcycle | 0.3% |
| CD player | 0.1% | Motorcycle accessories | 0.0% |
| CDs | 0.1% | Pan | 0.1% |
| Cellphones | 0.3% | Post | 0.1% |
| Childrens’ bicycles | 0.1% | Postgraduate education | 0.1% |
| Compuer desk | 0.1% | Printer cartridges | 0.0% |
| Computer | 0.2% | Printing paper | 0.0% |
| Digital photcamera | 0.0% | Quilts | 0.1% |
| Dining room set | 0.1% | Rent of primary residence | 3.7% |
| Drill | 0.1% | Television | 0.1% |
| DVD player | 0.1% | Toys | 0.1% |
| Electric lamps | 0.1% | Vehicle repair | 1.1% |
| Excursions | 0.4% | Wardrobe | 0.1% |
| Flight tickets | 0.4% | Washing machine | 0.1% |
| Fridge | 0.2% | Water heater | 0.1% |
| Glasses | 0.3% | 13.7% |
List of goods and services with highest exchange rate pass-through
List of goods and services with highest exchange rate pass-through
| Item | Pass-Through Estimate | USD Quoted? | Durable? | Tradable? |
|---|---|---|---|---|
| DVD Player | 0.99 | Yes | Yes | Yes |
| Motorcycles | 0.97 | Yes | Yes | Yes |
| Home insurance | 0.93 | Yes | Yes | No |
| Flight tickets | 0.92 | Yes | No | No |
| Equipo fotográfico, cinematográfico e instrumentos ópticos | 0.91 | Yes | Yes | Yes |
| Equipos para fotografía y video | 0.91 | Yes | Yes | Yes |
| Digital camera | 0.91 | Yes | Yes | Yes |
| Car | 0.91 | Yes | Yes | Yes |
| Fridge | 0.91 | Yes | Yes | Yes |
| Washing machine | 0.89 | Yes | Yes | Yes |
| Aire acondicionado | 0.89 | Yes | Yes | Yes |
| Computadoras personales | 0.89 | Yes | Yes | Yes |
| Kitchen | 0.86 | Yes | Yes | Yes |
| Computer | 0.85 | Yes | Yes | Yes |
| Flank steak | 0.83 | No | No | Yes |
| Excursions | 0.82 | Yes | No | No |
| Hotel | 0.80 | Yes | No | No |
| Car Cover | 0.80 | Yes | Yes | Yes |
| CD Player | 0.79 | Yes | Yes | Yes |
| Television | 0.77 | Yes | Yes | Yes |
List of goods and services with highest exchange rate pass-through
| Item | Pass-Through Estimate | USD Quoted? | Durable? | Tradable? |
|---|---|---|---|---|
| DVD Player | 0.99 | Yes | Yes | Yes |
| Motorcycles | 0.97 | Yes | Yes | Yes |
| Home insurance | 0.93 | Yes | Yes | No |
| Flight tickets | 0.92 | Yes | No | No |
| Equipo fotográfico, cinematográfico e instrumentos ópticos | 0.91 | Yes | Yes | Yes |
| Equipos para fotografía y video | 0.91 | Yes | Yes | Yes |
| Digital camera | 0.91 | Yes | Yes | Yes |
| Car | 0.91 | Yes | Yes | Yes |
| Fridge | 0.91 | Yes | Yes | Yes |
| Washing machine | 0.89 | Yes | Yes | Yes |
| Aire acondicionado | 0.89 | Yes | Yes | Yes |
| Computadoras personales | 0.89 | Yes | Yes | Yes |
| Kitchen | 0.86 | Yes | Yes | Yes |
| Computer | 0.85 | Yes | Yes | Yes |
| Flank steak | 0.83 | No | No | Yes |
| Excursions | 0.82 | Yes | No | No |
| Hotel | 0.80 | Yes | No | No |
| Car Cover | 0.80 | Yes | Yes | Yes |
| CD Player | 0.79 | Yes | Yes | Yes |
| Television | 0.77 | Yes | Yes | Yes |
Prepared by Frederik Toscani.
See Toscani (2017).
Recall that we find that overall consumption falls as a consequence of depreciations—expenditure switching from imports to domestic goods does happen but as long as overall consumption falls this.
In a recent publication on external adjustment in Latin America and the Caribbean (WHD April 2017 REO) the IMF showed that while a negative income effect dominated in most countries in the region, a positive expenditure switching effect has become more important over the past years.
Ideally, one would like to be able to have household-level panel data which capture microeconomic heterogeneity in expenditures, income, assets and debt to be able to isolate and estimate each channel. Given the very substantial data constraints, this paper will analyze the link between exchange rate movements and consumption through several complementary but partial approaches to shed more light on the question.
Much of the discussion in this section is based on Carroll (2001) and Kaplan et al. (2014) (as well as Mark Aguiar’s and Karen Pence’s comments on Kaplan et al.). Also see Carroll (2013) for an overview of the current state of the consumption literature and the importance of acknowledging the fact that aggregate consumption behavior cannot be well captured by a representative household but that heterogeneity in income, assets and potentially preferences are crucial.
Indeed, empirical evidence suggests that this works reasonably well for large fluctuations (see, for example, Hsieh, 2003).
See Johnson et al. (2006) for a prominent example and Jappelli and Pistaferri (2010) for a literature review.
See for example saver-spender models where impatient spenders borrow from patient savers and consume all their income every period (Gali et al., 2006).
Apart from optimal portfolio allocation, wealthy HtM consumers can also arise following a large negative shock or extreme impatience. Hyperbolic discounters might be wealthy HtM consumers to protect themselves against future excessive consumption, for example.
A good which does not depreciate at all is equivalent to wealth.
Consumers attempt to smooth the service flow of durable services rather than expenditure on durables. Of course, when households are liquidity constrained they cannot smooth as desired (see above discussion on HtM consumers).
The survey was conducted by the Economics department of the Universidad de la Republica. The design is based on the Bank of Spain’s “Encuesta Financiera de las Familias espanolas” as well as the Bank of Italy’s “Survey of Household Income and Wealth” and the Chilean Central Bank’s “Encuesta Financiera de Hogares”. The baseline sample is derived from the Uruguayen Statistic Institute’s 2012 version of the continuous household survey (ECH- 2012). Given the high concentration of wealth in Uruguay—as in other countries—the EFHU-2 survey over-samples very high income and wealth households. Using the appropriate survey weights, the data is representative of households in urban areas in Uruguay. See methodological guide (“Metodologia y guia para el usario EFHU-2) published in August 2016 for further details.
See Lluberas and Odriozola (2015) for descriptive statistics on wealth and household balance sheets using EFHU-2. For household saving behavior in Latin America more broadly see IDB (2015).
We calculate HtM status in two preferred ways. (i) based on the ratio of liquid assets to income and (ii) based on the response to a survey question which asks households whether they spent more, as much or less than their income over the past year. For this measure, HtM households are defined as those which consume as much or more than their income. For measure (i), note that we do not subtract any debt from liquid assets since we do not have a measure of credit card debt which is what Kaplan et al. (2014) use. In that sense our measure is a lower bound for the share of HtM households. We get to the 40 percent number when we only consider households with 0 or negative net total (rather than liquid) wealth as HtM.
Kaplan et al. (2014) calculate the share of HtM households and the prevalence of wealth HtM households using survey data from the US, UK, Canada, Australia, Germany, France, Italy and Spain. Our numbers are not directly comparable to theirs given a somewhat less granular analysis in the present study but it is nevertheless clear that while HtM households make up a significant portion of the population in advanced economies, the fraction is much higher still in Uruguay.
For the U.S. the split is about 2/3—1/3.
Note that we do not have data on real disposable income but only a real wage index. This blurs the interpretation of results to some degree.
Tests are done for the series with and without a linear trend.
et ≡ exchange rate, yt ≡ real income index, ct ≡ consumption. We also estimate a larger model which includes 12-month interest rates and employment. Results become more sensitive to the ordering but the basic result on the impact on real income and the exchange rate on consumption remains. See Matheson and Goes (2017) for a similar VECM for consumption in Brazil.
Note that the persistence of the impact of shocks is somewhat built-in to the analysis through the long-run relationship of the VECM where both real wages and the exchange rate enter significantly.
As we will see below there is huge heterogeneity in terms of pass-through by good.
We include a time trend in the regressions so that the exchange rate coefficient can be interpreted as the reaction to a deviation from the expected trend. As noted before, this is relevant given that the Peso tends to slowly trend-depreciate against the USD and agents expect it to do so.
When we exclude real wages from the regression in column 1, the coefficient on the exchange rate becomes significant, again indicating that the income channel of exchange rate movements is indeed relevant.
The Appendix shows a table with pass-through by item for those items in the right tale of the pass-through distribution. The appendix also lists all USD-quoted goods and services which are part of the CPI.
Hedging against exchange rate and inflation shocks is of course one of the key reasons for households to hold dollar assets.
Note that the magnitude of the ratio cannot be easily interpreted since the nominator is an index of consumer goods imports while the denominator is private consumption in constant 2005 Pesos.
Also see Obstfeld and Rogoff (1996) for the basic point that durable goods increase current account volatility.
Variables are in logs and de-trended using an HP filter. The ratio of volatilities for the de-trended series in Uruguay is 1.14.
The two mechanisms are potentially observationally equivalent if exchange rate movements lead to interest rate movements to satisfy interest rate parity. In the mechanism here proposed the nominal exchange rate would need to be an exogenous parameter. Anecdotal observations for Uruguay does not make it seem unlikely though that global factors are a key driver of the Peso.
