Sunday, March 18, 2018

Welfare Gains from Trade in Multi-Sector Models: The Role of Aggregation and Income Elasticities


Welfare Gains from Trade in Multi-Sector Models: The Role of Aggregation and Income Elasticities


One sentence summary: In the calculation of welfare gains from trade, the income elasticity interacts with the trade elasticity that can be estimated in a bilateral trade regression as the coefficient in front of aggregated unit prices instrumented by gravity variables.


Abstract
Sectoral heterogeneity has been shown to affect country-level welfare gains from trade that can be calculated by sector-specific trade elasticities and home expenditure shares. However, empirical analyses of multi-sector models are restricted to a limited number of countries and sectors, mostly due to the lack of data on sector-specific home expenditure shares. This paper first proposes a solution to this limitation by changing the way that foreign products are aggregated at the destination country. Second, when firm-level productivity differences are controlled for in a bilateral aggregate trade equation, gravity variables are shown to be perfect instruments for aggregated unit prices that enter trade regressions for the estimation of trade elasticities. Third, when the assumption of unitary income elasticity is relaxed, the trade elasticity in the calculation of welfare gains is shown to be replaced by the newly-introduced welfare elasticity, a function of trade and income elasticities. Empirical evidence is shown for the heterogeneity of these elasticity measures across countries.


Non-technical Summary
Welfare gains from trade (measured as costs of autarky) can be expressed by two key parameters, namely the trade elasticity and home expenditure share. When the one-sector environment is extended to a multi-sector one, due to the way that sectors are aggregated at the destination country (i.e., an upper tier aggregation of individual utility across sectors), the two key parameters are required at the sector level in order to calculate welfare gains from trade. Although the trade elasticity can be estimated for pretty much any sector and any country by using the corresponding trade and price/tariff data, home expenditure data are available only for certain aggregation of sectors in certain countries. Accordingly, in order to calculate welfare gains in a multi-sector environment, several studies focusing on the estimation of both parameters (at the sectoral level) have been restricted to a limited number of countries and a limited number of sectors. Since one-sector trade elasticity measures are biased due to sectoral heterogeneity, unbiased welfare gains from trade cannot be calculated for several countries that lack sector-level home expenditure data (although they have all the necessary trade data). Moreover, when the number of countries is limited, the estimated elasticity parameters (especially those that are common across countries) simply cannot represent global trade patterns, which would lead into biased welfare calculations.

Considering a multi-sector framework in order to have unbiased welfare implications, this paper proposes a new aggregation approach that does not require any sector-level home expenditure data. Instead, the overall home expenditure share data are shown to be enough to calculate the welfare gains from trade. This is achieved by using an Armington model, where aggregation at any destination country is achieved across source countries at the upper tier and across sectors at the middle tier; the lower tier aggregates firm-level goods. Having an upper-tier aggregation across source countries corresponds to having a unique trade elasticity measure interacting with a unique home expenditure share in order to calculate welfare gains from trade for any destination country. Accordingly, when the upper-tier elasticities are estimated, welfare gains can be calculated for any country that has data for overall home expenditure share.

On top of solving the problem of not having sector-level home expenditure data, this paper also relaxes the restrictive assumption of having unitary income elasticity in the calculation of welfare gains. In an Armington (1969) framework, it is shown that the standard trade elasticity in the case of unitary income elasticity (which is one minus the elasticity of substitution) is replaced with the country-specific income elasticity minus the elasticity of substitution (that we call welfare elasticity in this paper). It is implied that income effects counteract those of substitution effects when economies open up for trade, because consumers may simply value the utility out of home products different from that of foreign products due to the heterogeneity in income elasticity measures.
The inclusion of income elasticity in welfare gains calculations is achieved by using implicitly additively separable nonhomothetic constant elasticity of substitution (CES) preferences across source countries at the upper tier of individual utility. Such an approach is essential to separately capture the income effects in the utility function, without giving away the standard features of having CES preferences, so that one can easily distinguish between income and substitution effects, even in the calculation of welfare gains.

The model is estimated by using UN Comtrade data at the six digit Harmonized System (HS) level between 1995-2015. The estimation of the country-specific income elasticity and the country-specific elasticity of substitution is achieved by using data on bilateral trade (of imports measured at the destination country) and unit prices. Since the lower tier aggregation of individual utility is achieved across firm-level goods, firm-level productivity differences are carefully connected to the data on unit prices and the corresponding estimation; therefore, although firm-level data are not utilized, firm-level productivity differences are still taken into account at the aggregated level, without making any assumptions on their distribution.

Following the literature, the aggregation across sectors is achieved by a Cobb-Douglas aggregation. Due to the tiers of aggregation introduced earlier, this corresponds to having a weighted average of log unit destination prices in the bilateral aggregate trade estimation, where the weights are simply the expenditure weights of sectors. Nevertheless, since firm-level productivity measures are carried over to the bilateral aggregate trade estimation as residuals, aggregated log unit prices become endogenous. By using the implications of the model, bilateral trade costs measured by standard gravity variables are shown to be strong instruments for aggregated unit prices in a Two-Stage Least Squares (TSLS) estimation. In this bilateral aggregate trade estimation, the coefficient in front of aggregated unit prices represents the country-specific trade elasticity (i.e., one minus the elasticity of substitution), while the coefficient in front of log total imports (due to having non-unitary income elasticity) represents the country-specific income elasticity; these estimates are further used to construct country-specific welfare elasticity estimates.

The corresponding welfare elasticity estimates have a median value of -4.3 across countries with a range between -6.0 and -2.7 that are highly consistent with trade elasticity estimates in the literature. Although the estimates are comparable, the methodology introduced in this paper in a multi-sector framework is much easier to implement, and it allows us to calculate welfare gains in pretty much all countries that have trade data. In order to show the contribution of this paper in a clear way, the obtained welfare elasticity estimates are further compared to the common (across countries) trade elasticity estimate of about -3.8 that is obtained by the very same data set. It is shown that the difference between country-specific welfare elasticity estimates and common trade elasticity estimate is the key to understand the heterogeneity of welfare gains across countries for a given measure of home expenditure share. This heterogeneity is further connected to the trade patterns and per capita income of countries by using the implications of the model.
 






Saturday, March 17, 2018

Unifying Macro Elasticities in International Economics


Unifying Macro Elasticities in International Economics


One sentence summary: International elasticity puzzle is solved when expenditure shares are taken into account while calculating price elasticities of demand; moreover, international trade and finance literature imply the very same welfare gains from trade when international finance studies have unitary (Armington) elasticity of substitution between home and foreign products or unitary terms of trade.

The corresponding SSRN working paper by Hakan Yilmazkuday is available here.

The Dallas-Fed Globalization Institute working paper version is available here.
 

Abstract
International trade studies have higher macro (Armington) elasticity measures compared to international finance studies. This observation evokes not only mixed policy implications regarding tariffs and exchange rates but also mixed welfare gains from trade. Regarding the policy implications, this so-called international elasticity puzzle is solved in this paper by showing that the macro elasticity in international trade is a weighted average of the macro elasticity in international finance and the elasticity of substitution across products of foreign source countries. It is implied that the discrepancy between the two literatures regarding the macro elasticity can be explained by carefully following the existing aggregation definitions of foreign products at destination countries. Regarding the welfare gains from trade, the two literatures are shown to have the very same implications when international finance studies have a unitary macro elasticity of substitution between home and foreign products or unitary terms of trade. Empirical evidence further supports the theoretical solutions provided in a clear way.


Non-technical Summary
International trade studies have higher macro (Armington) elasticity measures compared to international finance studies. Since price changes are converted into quantity changes and thus real effects through these elasticities, this observation has evoked mixed policy implications regarding tariffs and exchange rates in the two literatures. Moreover, since welfare gains from trade are directly connected to these macro elasticity measures, this observation has also evoked mixed welfare gains between the two literatures. Due these mixed implications, this observation has been called the international elasticity puzzle.

In order to have a better idea about the magnitude of this puzzle, we would like to consider the elasticity measures used in the literature. Although elasticity measures differ across studies, international finance studies mostly use a macro elasticity value of about 1.5, while international trade studies mostly use a macro elasticity value of about 5. It is implied that if we directly employ these numbers in a policy analysis, say, in order to investigate the effects of a foreign price change due to tariffs or exchange rates, international trade studies imply quantity changes that are at least three times the international finance studies. Similarly, if we use the standard formula for the welfare gains from trade, which is a function of home expenditure share and the macro elasticity, international finance studies imply welfare gains (in percentage terms) that are about eight times the international trade studies (for any given home expenditure share).

This paper unifies these macro elasticities that lead into mixed results in the two literatures. Since the upper-tier aggregation is achieved across source countries (including home country) in international trade studies, and it is achieved across home and foreign products in international finance studies, the two literatures are connected to each other by an additional tier of aggregation across different foreign countries in international finance. Although such an additional tier is missing in international finance studies, it is well understood to exist in the background.

Within this framework, we show that there is no connection between the macro elasticities of the two literatures when expenditure shares are negligible in the calculation of price elasticities. The tables turn when such expenditure shares are taken into account in this paper, which results in having price elasticities of demand connected to elasticities of substitution through such expenditure shares. We show that such a strategy helps us understand several differences across studies in the literature regarding the elasticity measures such as the differences due to simulating versus estimating, differences due to the level of disaggregation (e.g., having different digits of data), and differences between long-run and short-run elasticity measures. More importantly, such a strategy also allows us connect macro elasticities in the two literatures through the elasticity of substitution across different foreign countries in international finance that is newly introduced in this paper. In particular, when expenditure shares are considered, we show that the macro elasticity in international trade is a weighted average of the macro elasticity in international finance and the elasticity of substitution across products of foreign countries. It is implied that one can always find an elasticity of substitution across foreign countries that would be consistent with different macro elasticities in the two literatures; therefore, the puzzle is solved theoretically. Then, how can one make sense of the mixed policy implications and mixed welfare gains from trade implied by the two literatures? We focus on this question next.

From a researcher’s or a policy maker’s perspective, since policy implications for any individual good coming from an individual foreign country should not depend on how foreign goods are artificially aggregated at the destination country, we equalize the micro price elasticities of demand (depending on expenditure shares) between the two literatures to show theoretically and confirm empirically that the elasticity of substitution across different foreign countries plays an important role in the determination of the puzzle. Therefore, the policy implications at the individual foreign good level are automatically equalized, although the artificially created upper-level elasticities are allowed to change between the two literatures; this inductive approach is different from the deductive approach followed by the existing literature, where the upper-level variables (e.g., utility) are taken as given, while lower-level variables are allowed to change. Since policy implications are equalized between the two literatures for each and every foreign good, the international elasticity puzzle disappears at the disaggregated micro level. It is implied that the policy analysis should be first achieved at the disaggregated micro level and then aggregated up to obtain macro implications.

After solving the puzzle due to its policy implications at the micro level, we continue with focusing on different implications by the two literatures regarding the welfare gains from trade at the macro level (which also correspond to the macro-level effects of a foreign shock in home country). Rather than using the standard formula for the calculation of welfare gains, we consider the implications of each literature (measured by the costs of autarky) in order to have a comparison by searching for aggregate price indices that would have to adjust to keep the consumer utility the same between the current openness to trade and a hypothetical autarky, for any given expenditure. We show that the two literatures have the very same welfare gains from trade when there is unitary macro elasticity of substitution between home and foreign products or when there is unitary terms of trade in international finance. Hence, there is no international elasticity puzzle from the perspective of welfare gains from trade as long as there is unitary macro elasticity or unitary terms of trade in international finance and as long as we do not make any simplifying assumptions regarding the implications of our models for welfare gains from trade; we should rather derive our model-specific formulas.

In sum, this paper proposes solutions to two dimensions of the international elasticity puzzle, one regarding mixed policy implications, the other regarding the implications for welfare gains from trade. We empirically test both of these solutions by estimating the corresponding elasticity measures. Such an empirical investigation requires data on both foreign/international and domestic/internal trade. Since international trade data sources do not record domestic trade, we shift our focus to interstate trade data from the U.S., where data are available for both interstate (foreign) and intrastate (domestic) trade. The empirical results show evidence for unitary macro elasticity of substitution between home and foreign products (in the upper-tier aggregation), which implies that the two literatures already imply the very same welfare gains from trade, and thus international elasticity puzzle is solved when the definition of foreign basket of goods is achieved carefully. The empirical results also confirm that the macro elasticity in international trade is a weighted average of the macro elasticity in international finance and the elasticity of substitution across products of different foreign source countries, where weights can also be explained by the implications of the theory. Overall, the two solutions proposed to the international elasticity puzzle in this paper are supported empirically in a clear way.

The corresponding working paper by Hakan Yilmazkuday is available here.


Friday, March 16, 2018

Domestic vs. International Welfare Gains from Trade


Domestic vs. International Welfare Gains from Trade


One sentence summary: Domestic trade contributes about 91 percent to overall welfare gains from trade.

Abstract
Using varieties of a rich model that considers sectoral heterogeneity and input-output linkages, this paper shows that the overall welfare gains of a region within a country can be decomposed into domestic versus international welfare gains from trade. Empirical results based on state-level data from the U.S. suggest that about 91 percent of the overall welfare gains of a state are due to domestic trade with other states, on average across alternative model specifications, with a range between 72 percent and 99 percent across states. When national-level data are used for the U.S., international welfare gains are shown to be almost identical to the those obtained by the aggregation of state-level results, suggesting that one can use the implications of a region-level analysis to have national-level results based on welfare gains from trade. We use this implication to propose an approximation to measure the domestic welfare gains from trade when domestic trade data are not available. Accordingly, using the implications of the model introduced, a Dispersion of Economic Activity Index (DEAI) is introduced that depends on internal distance and elasticity measures. It is empirically shown that DEAI can capture domestic welfare gains from trade within the U.S. when standard internal distance and elasticity measures in the literature are employed. Important policy suggestions follow.


Non-technical Summary
Domestic trade of a typical state in the U.S. is about five times its international trade, where about three quarters of this domestic trade is achieved with other states. It is implied that a typical state is about 20 percent open to international trade, while it is about 60 percent open to domestic trade. Since welfare gains from trade are known to be directly connected to such openness measures for a vast variety of models, the greater part of the welfare gains are implied to be through domestic trade. Nevertheless, since domestic trade data are not available for the majority of the countries, the existing literature has mostly focused on international welfare gains from trade that represent only a small portion of overall welfare gains.

Within this picture, this paper introduces a rich model considering sectoral heterogeneity as well as input-output linkages, where the unit of investigation is set as regions representing U.S. states. As standard in the literature, the corresponding welfare gains from trade are shown to be a function of expenditure shares and model parameters, where changes in expenditure shares are used to capture the changes in welfare in case of a hypothetical change in trade costs. The corresponding literature has focused on the hypothetical case of an autarky in the context of international trade. This paper follows this literature by having the same definition of international autarky while calculating the international welfare gains from trade.

The main contribution of this paper is achieved by considering an additional/alternative hypothetical case of autarky, namely domestic autarky, which is useful to calculate the domestic welfare gains from trade. In particular, domestic autarky is defined as the case in which a region still imports products internationally, but the domestic trade with other regions of the same country is shut down in this hypothetical case. It is shown that the overall percentage welfare gains from trade is the summation of domestic and international welfare gains from trade.


Based on the significant difference between international and domestic openness measures of states in the U.S., the corresponding welfare analysis shows that about 91 percent of the overall percentage welfare gains of a state are due to domestic trade with other states, on average across alternative model specifications, with a range between 72 percent and 99 percent across states.


When the same investigation is replicated at the U.S. level, it is shown that the international welfare gains from trade measures are almost identical to the measures obtained by the state-level analysis. Therefore, one can use the implications of a state-level analysis to have U.S. level results based on welfare gains from trade. We combine this result with other implications of the model in order to propose an approximation for domestic welfare gains from trade when domestic trade data are not available. Accordingly, we introduce a Dispersion of Economic Activity Index (DEAI) that can be calculated by using domestic distance measures (that can easily be obtained within a country) as well as other parameters such as trade elasticity and distance elasticity of trade costs. It is empirically shown that the proposed DEAI can capture the effects of domestic welfare gains from trade within the U.S. when great circle distance measures (that are calculated using latitudes and longitudes of states) are employed together with the elasticity measures borrowed from the literature.

Overall, this paper contributes to the existing literature by showing that (i) domestic welfare gains are much higher than international welfare gains from trade and (ii) when domestic trade data are not available, domestic welfare gains from trade can be approximated by a dispersion of economic activity index (DEAI). Regarding policy implications, the calculated DEAI measures can be compared with the standard measures of international welfare gains from trade in order to evaluate policies toward integrating the regions of a country with the rest of the world.




Thursday, March 15, 2018

Spatial Dispersion of Retail Margins: Evidence from Turkish Agricultural Prices


Spatial Dispersion of Retail Margins: Evidence from Turkish Agricultural Prices


One sentence summary: The farmer share of Turkish grocery prices is only about 16 percent, corresponding to about 84 percent of a distribution share (77 percent of retail margins and 7 percent of transportation costs).

The corresponding paper by Hakan Yilmazkuday has been accepted for publication at Agricultural Economics.

The working paper version is available here.

Abstract
The farmer share of retail prices is shown to be about 16 percent, corresponding to about 84 percent of a distribution share, on average across agricultural products and regions within Turkey. The share of transportation costs in retail prices is only about 7 percent, while the share of retail margins is about 77 percent of retail prices. The dispersion of retail prices across regions is shown to be mostly due to local wages and variable markups, while the contribution of traded-input prices is relatively small. Accordingly, the high dispersion of farmer prices across locations is not reflected in the dispersion of retail prices due to the high contribution of retail margins. These retail margins are also shown to account for about one third of the consumer welfare dispersion across regions and more than half of the consumer welfare dispersion across products.


Non-technical Summary
The portion of agricultural retail prices received by farmers, the so-called farmer share, is about 15 percent across countries. Accordingly, the distribution share consisting of transportation costs and retail margins constitute the bigger portion of retail prices. The decomposition of this distribution share into its components is important to understand the welfare and policy implications for both consumers and farmers. For example, if the distribution share is high due to transportation costs, the optimal policy to improve welfare would be to reduce them through investments on infrastructure or subsidies on transportation-related costs, while if the distribution share is high due to retail margins, they can be subject to effective price regulations that can increase consumer welfare due to lower prices and increase farmer welfare due to higher sales. Since retail margins increase with the market share, the effects of retail margins (and thus the corresponding price regulation) may even be higher in regions with higher market power. Accordingly, it is essential to have a decomposition of the distribution share both an average and across regions to achieve optimal policies.

This paper achieves such decomposition by using retail- and farm-level micro price data on agricultural products across Turkish regions. In order to have an empirical motivation and implications for welfare, a simple model is introduced, where the economic environment consists of regions inhabited by individuals who consume local retail goods and retailers who purchase traded-inputs from producers subject to transportation costs. Since we empirically focus on individual products of green groceries, producers correspond to farmers who produce homogenous goods. Accordingly, each retailer searches for the minimum price across farmers at the product level, subject to transportation costs. Once transportation costs and source farms are identified through estimations based on price data obtained from both farmers and retailers, traded-input prices are determined for retailers. By further introducing a structure on local retail costs through the model, the retail prices are decomposed into farmer prices and distribution costs (consisting of transportation costs and retail margins).

The empirical results show that the farm share is about 16 percent of retail prices on average across agricultural goods and regions, corresponding to about 84 percent of a distribution share. The share of transportation costs in retail prices is only about 7 percent, while retail margins (defined as the ratio of retail to traded-input prices including transportation costs) are about 4.47, implying that about 77 percent of retail prices are accounted for by the retail sector. This result corresponds to slightly higher transportation costs within Turkey compared to similar costs in the U.S. of about 4 percent for food products. This may be surprising, because the U.S. is a much more spatially dispersed economy (due its land size) and thus one may expect lower transportation costs within Turkey. Nevertheless, since transportation costs between farmers and retailer highly depend on fuel prices, the difference in transportation costs of Turkey and the U.S. can easily be attributed to ratio of fuel prices in Turkey to those in the U.S. which has an average of about 2.6 between 1994 and 2011 according to World Development Indicators.

 


When a comparison is achieved across regions, the dispersion of retail prices is mostly due to local wages and variable markups (95 percent), while the contribution of traded-input prices is relatively small (5 percent). When we further investigate the dispersion of traded-input prices across locations, the contribution of transportation costs dominate by 95 percent, while that of source prices is only 5 percent. It is implied that the high dispersion of farmer prices across locations is not reflected in the dispersion of retail prices due to factors such as local input costs, variable markups, and transportation costs. It is also shown that retail margins are dispersed across regions at the product level. The implications of the model suggest that the dispersion of retail margins (across regions) is explained 52 percent by traded-input prices, and 48 percent by local wages and variable markups. Since the dispersion of traded-input prices is mostly due to transportation costs, it is implied that final consumers face different retail margins across locations due to all of transportation costs, local wages and variable markups.

Finally, using the implications of the model for consumer welfare, on average across individual products, about 30 percent of the consumer welfare dispersion is explained by retail margins across Turkish locations, while another 70 percent is explained by differences in either real economic sizes of regions or traded-input prices. On the other hand, within the same location, retail margins contribute by about 60 percent to the consumer welfare dispersion across products. Hence, the retail margin (that can be mostly explained by local wages and variable markups) is one of the key variables in understanding the dispersion of consumer welfare across regions.

The corresponding working paper by Hakan Yilmazkuday is available here.

Wednesday, March 14, 2018

Redistributive Effects of Gasoline Prices


Redistributive Effects of Gasoline Prices


One sentence summary: There are significant redistributive effects of gasoline price changes among U.S. consumers, where the main determinant is shown to be the consumer income.

The corresponding paper by Demet Yilmazkuday and Hakan Yilmazkuday has been accepted for publication at Networks and Spatial Economics.


Abstract
Consumers face significantly different gasoline prices across gas stations. Using gasoline price data obtained from 98,753 gas stations within the U.S., it is shown that such differences can be explained by a model utilizing the gasoline demand of consumers depending on their income and commuting distance/time, where the pricing strategies of both gas stations and refiners are taken into account. The corresponding welfare analysis shows that there are significant redistributive effects of gasoline price changes among consumers where the main determinant is shown to be the consumer income; e.g., welfare costs of an increase in gasoline prices are found to be higher for lower income consumers.


Non-technical Summary
Gasoline prices have significant effects on an economy, because higher energy prices can slow economic growth and affect individual welfare in many ways. As one example, gasoline prices have increased before any historical economic downturn in the U.S. As another example, consider the survey reported by Bankrate.com in May 2012, which depicts that, from the end of December 2011 through mid-April 2012, the price of regular gas rose from a national average of $3.30 per gallon to $3.94 (an increase about 19%), and, as a result, 59% of consumers cut back on nonessential spending on things such as vacations and dining out, only because of gasoline price changes. These macroeconomic examples provide an average picture of the gasoline price effects, but is the magnitude of these effects the same across consumers? The answer to this question is essential to understand the redistributive effects of gasoline prices, especially when gasoline prices differ across consumers.

Consider the following figure where each circle represents the location of a gas station. The colors of the circles represent the prices in U.S. dollars per gallon. Price intervals represent the intervals corresponding to the first, second, third, fourth and fifth 20th percentile of average of daily gasoline prices obtained from 98,753 gas stations between September 8th and September 14th, 2014. As is evident, while the gasoline prices are more expensive in the Northeast and the West (including Alaska and Hawaii), they are relatively cheaper in the Southeast.





To better understand the magnitude of gasoline price differences across consumers, consider a typical day (of September 14th, 2014) when the retail-level gasoline price difference between any two gas stations within the U.S. was as high as $2.28 per gallon of regular gas. If you think that this price dispersion was due to differences in state-taxes per gallon, which ranged between 42.75 cents (for New York) and 8 cents (for Georgia) in 2014, you are only partially right, because, for a typical day (of September 14th, 2014), the price difference between any two gas stations within any given state of the U.S. was as high as $1.68 (for the state of Massachusetts) followed by $0.99 (for the state of New York). Therefore, a detailed analysis is required to understand gasoline price dispersion at the gas-station level, which is the key to the investigation of the redistributive effects of gasoline price changes.

This paper achieves such an investigation by modeling the gasoline consumption of individuals and the pricing strategy of gas stations and refiners. The optimization in the model results in the gasoline demand of consumers depending on their income and commuting requirements as well as the price of gasoline. Gas stations take this demand into consideration while maximizing their profits, which results in a linear gasoline price expression due to having Leontief production functions. Refiners take into account the demand coming from gas stations to maximize their own profits. When the behavior of all agents in the model are combined, a final expression for gasoline prices is obtained at the gas station level, which depends on the income and commuting behavior of consumers as well as refiner-related costs.

Using data on gas-station level gasoline prices, zip-code level income and zip-code level commuting within the U.S., the implications of the model are estimated. The results show that most of the variation of gasoline prices (across gas stations) is explained by the proposed model. As a supplementary result, the average (across gas stations) markup per gallon is estimated about 16 cents, which is consistent with the surveys achieved by independent organizations.

After showing that the implications of the proposed model are consistent with gasoline price data, together with other supplementary data, we move to the welfare analysis to investigate the redistributive effects of gasoline price changes across consumers within the U.S.. The implications of the model combined with the results coming from the empirical investigation suggest that 1 percent of an increase in gasoline prices can lead to a reduction in consumer utility ranging between 0.08 percent and 2.76 percent (with an average of 0.82 percent) within the U.S.. Therefore, there are in fact significant redistributive effects of gasoline price changes. When the sources of these redistributive effects are further investigated, it is shown that consumer income is the main determinant; i.e., welfare costs related to a gasoline price increase are higher for lower-income consumers. It is implied that, in order to minimize the redistributive welfare effects of gasoline price changes, special policies should be conducted for lower-income consumers, especially when gasoline prices increase significantly.

Although gasoline prices can be affected by income, commuting distance/time, oil prices, and refiner costs according to the proposed model, they can also be affected by local or national taxes that have not been modeled here (nevertheless, they have been controlled for in the empirical investigation). Therefore, a change in any of these variables would change gasoline prices, and, thus, any policy conducted on such variables would result in redistributive welfare effects among consumers according to the analysis, above. Accordingly, one policy suggestion would be to provide gasoline tax cuts for neighborhoods with lower-income consumers. Providing tax reimbursements for lower-income consumers depending on their gasoline consumption and/or the gasoline (or oil) price changes over the preceding year can also be considered. Another one would be to promote/subsidize fuel-efficient cars for lower-income consumers that would effectively reduce the share of gasoline in their expenditure. Even though the formal investigation of such suggestions is out of the scope of this paper, future research can focus on the public policy implications of a more local analysis based on the insights of this study.

The working paper version is available here.




Tuesday, December 20, 2016

Understanding Long-run Price Dispersion


Understanding Long-run Price Dispersion


One sentence summary: At the PPP level, almost all of price dispersion is attributed to unskilled wage dispersion, while at the LOP level, borders and distance contribute about equally to price dispersion that is rising in the distribution share.

The corresponding paper by Mario J. Crucini and Hakan Yilmazkuday has been published at Journal of Monetary Economics.


Abstract
We use a unique panel of retail prices spanning 123 cities in 79 countries from 1990 to 2005, to uncover six novel properties of long-run international price dispersion. First, at the PPP level, virtually all (91.6%) of price dispersion is attributed to service-sector wages, consistent with a dominant role of the retail distribution margin. Second, at the level of individual goods and services, the average contribution of service-sector wages is significantly reduced, one-third as large (31.9%). This reflects the fact that good-specific sources of price dispersion, such as trade costs and good-specific markups, tend to average out across goods. Third, at the LOP level, borders and distance contribute about equally to price dispersion with distance elasticities consistent with the existing trade gravity literature which links trade volumes (rather than relative prices) to borders and distance. Fourth, in the cross-section, price dispersion is rising in the distribution share consistent with the notion that baby-sitting services and haircuts embody local wages to a far greater extent than highly traded manufactured goods. Fifth, we provide the first estimates of distribution margins at the micro-level and show them to be very different across goods and substantial in the aggregate, where they account for about 55% of consumption expenditure. Sixth, these estimates are broadly consistent with more aggregated U.S. NIPA measures currently used in the literature.


Non-technical Summary
The Law-of-One-Price (LOP) is the theoretical proposition that, absent official and natural barriers to trade, international prices are equated in common currency units, and a laborer's purchasing power (i.e., real wage) is determined only by their labor productivity. A stark empirical implication of this proposition is that the cross-country correlation between price levels and wage levels is zero. As is well known, this implication of goods market integration is grossly at odds with the data. The Penn Effect, in recognition of the ambitious work of Heston, Kravis, Lipsey, who developed the Penn World Tables, shows a strong positive correlation between international price levels and per capita income.

The following figure shows the microeconomic counterpart of this fact using the panel data of our study. Microeconomic in this context means the prices of individual goods and services across cities of the world, as opposed to aggregate price levels at the national level. Specifically, each point in the scattterplot is the price of an individual good or service in a particular city plotted against the hourly wage of domestic cleaning help in that particular city. Prices and wages have been averaged over the period 1990 to 2005 to eliminate transitory deviations associated with business cycles and exchange rate fluctuations.


Specifically, there are 300 goods and services (up to missing observations) for each city and there are 123 cities in total. The prices and wages used to construct these time-averages are from the Economist Intelligence Unit (EIU) World Cost of Living Survey which spans 79 countries. As far as we know, this is the first study to use time-averaged data to study long-run deviations from the LOP and Purchasing Power Parity (PPP). The points labeled with an asterisk are price levels computed as expenditure-weighted averages of the individual prices.

In the figure, the estimated line through the scatter of price levels has a slope of 0.52 and an R-squared value of 0.37. The estimation is by geometric mean regression to consider for possible measurement errors in both the price and wage data. A common set of consumption expenditure weights are used for all cities. These consumption expenditure weights are taken from the PWT, averaged across all OECD nations. 

In words: a doubling of wages is associated with a 52 percent higher price level. This finding is typically associated with the seminal works of Harrod (1933), Balassa (1964), and Samuelson (1964); however, the HBS theory assumes that LOP holds for traded goods but not for non-traded goods. According to this view, called the classical dichotomy, there should be a horizontal line traced out by traded goods for which the LOP holds and a line with a slope of unity for non-traded goods. The trivial example is the hourly wage of domestic help itself, which produces a slope of one by recognizing that the market price of this non-traded service is, in fact, the hourly wage for unskilled labor. The figure above, obviously, is not much more sympathetic to the classical dichotomy than it is to complete market integration.

To help resolve this puzzle, this paper estimates distribution and trade cost wedges using a trade model augmented with a retail distribution sector (developed in Crucini and Yilmazkuday, 2009). We have two sets of results, one for relative price levels (PPP) and the other at the level of individual goods (LOP). 

Regarding PPP, the variance of price levels for international city pairs is found to be almost entirely explained by international wage differences, 92% by our estimate. Both the absolute amount of price dispersion and the relative importance of wage differences falls when the sample is restricted to cities in countries at similar stages of development while the role of retail productivity increases. The contribution of cross-city wage differences falls to 8% when the sample is restricted to city pairs within the same country. It is important to keep in mind that the amount of price level dispersion across cities that are located in the same country is a trivial 3-5%; as such, a modest amount of wage or retail productivity variance goes a long way in terms of accounting for the lion's share of the variance. The thrust of the PPP analysis is that when long run price level differences are consequential, the differences are attributable to the level of economic development, not traditional trade frictions.

The table turns dramatically in favor of borders and trade costs and away from wages and retail productivity, as explanatory factors, when the focus is LOP deviations. Pooling all international city pairs, the explanatory power of the HBS theory (wage dispersion) falls by a factor of three, to about 32%. Traditional theories of trade that emphasis distance and borders now account for the lion's share of price disperison, about 41%. City effects account for almost none of the international LOP variation. Essentially, this is because international LOP deviations are both large and idiosyncratic to the good once we condition on the wage level. The remainder is a residual term, which may reflect good and location-specific markups as well as other variables omitted from the model. The following figure shows how this decomposition changes across goods, where the vertical axis shows the deviations from LOP, while the horizontal axis shows the distribution share of goods.