Geographical Dispersion of Consumer Search Behavior
One sentence summary: Gas stations in the U.S. achieve
higher profits if they are located in zip codes with smaller areas, lower
population densities, higher income and/or higher commuting times.
The corresponding paper by Hakan Yilmazkuday has been published at Applied Economics.
The working paper version is available here.
The corresponding paper by Hakan Yilmazkuday has been published at Applied Economics.
The working paper version is available here.
Abstract
This paper investigates whether consumer search behavior differs across zip codes within the U.S.. As an application, daily gasoline price data covering virtually all gas stations within the U.S. are employed to estimate the distribution of search costs in each zip code. The results show that there are significant differences across zip codes regarding the expected number of searches achieved before consumers purchase gasoline. In order to have a systematic explanation, such differences are further connected to geographic, demographic and economic conditions of the zip codes in a secondary analysis. The corresponding results imply several strategies for gas stations in order to maximize profits/markups; suggestions follow for policy makers and regulators to reduce redistributive effects of information barriers across locations.
Non-technical Summary
Prices for the very same (homogeneous) good can be different across retailers. This is most apparent in the gasoline market where gas stations post alternative prices even within the same zip code in the U.S. For example, consider the following figure where the gasoline price spread has a median value of 14 cents with a range between 0 and 98 cents.
Since these retail prices are already controlled for
gas-station and time fixed effects at the zip code level, they are independent
of any gas station characteristics such as their location, brand, competition
level, having a car wash or a convenient store as well as time-varying supply
or demand shocks. One potential explanation to these price spreads is then the
lack of information that consumers have, which has been connected to search
costs in the literature. In particular, if consumers do not search for lower
prices, retailers may easily charge higher markups or get involved in collusive
behavior. Accordingly, policy makers have considered this lack of information
as a potential problem reducing consumer welfare due to information frictions. Within
this picture, we investigate the search behavior across consumers in different
five-digit zip codes within the U.S., where gasoline purchases account for
approximately 5% of consumer spending.
First, we investigate whether the search behavior of
consumers differs across zip codes; we are particularly interested in the
expected number of searches achieved by consumers before making a purchase.
Accordingly, by using retail level gasoline price data obtained from virtually
all gas stations with the U.S. as an application, we first estimate the
expected number of searches and the corresponding search cost distributions at
the zip code level. This is achieved by considering the implications of a
non-sequential consumer search model with heterogeneous search costs. The
results show that the expected number of searches have a median of 1.66 across
zip codes, which implies that consumers do not search much on average before
purchasing gasoline. However, the estimates for the expected number of searches
range between 0.17 and 12.94 across zip codes; hence, there are significant
differences in the search behavior of consumers across zip codes.
Understanding the reasons behind this heterogeneity in the
search behavior of consumers across regions is the key to reduce the
redistributive effects of information frictions. In particular, if information
frictions are systematically higher in certain regions, policy makers can
reduce them only by achieving region-specific policies. Instead, if a common
multi-region policy is conducted, although it would reduce information frictions
in all regions, it would not necessarily reduce redistributive effects of
information frictions across regions.
Accordingly, in a secondary analysis, we investigate whether
the heterogeneity in the estimated consumer search behavior can be explained
systematically across zip codes. In particular, we attempt to connect the
estimated expected number of searches to geographic, demographic and economic
conditions of zip codes. The results show that geographical factors increasing
the price dispersion across gas stations such as the average distance between
them, overall area of the zip code or population density all contribute
positively (across zip codes) to the expected number of searches achieved by
consumers before making a purchase. On the other hand, income and commuting
time are shown to be negatively related to the expected number of searches
across zip codes, potentially capturing the opportunity cost of time for making
a search.
Consumers in zip codes with individuals working in different
industries are also shown to having different search behavior, with industries
such as retail trade and public administration contributing most to the
expected number of searches. We also show that consumers in zip codes with
higher percentage of Black or African American people search more compared to
those with white or Asian people. Finally, consumers in zip codes with higher
percentage of females are shown to search more compared to other zip codes.
Retailers can charge higher markups if consumers do not
search for lower prices, which is one of the implications of the model used in
this paper. Combining this information with the fact that gasoline is a
relatively inelastic product (according to the U.S. Energy Information
Administration), it is implied by the results of this paper that gas stations
can achieve higher profit margins if they would be located in zip codes in
which gas stations are closer to each other; this partly explains why we
observe gas stations located very close to each other in certain zip codes.
Similarly, higher profit margins can be achieved in zip codes with smaller
areas, lower population densities, higher income and/or higher commuting times;
e.g., gas station profits would be maximized in zip codes with individuals
having annual income levels above $35K. On the other hand, such profits would
be lower in zip codes with higher percentage of Black or African American
individuals, followed by those with higher percentage of white and Asian
individuals. The profits would be lower also in zip codes with higher
percentage of individuals working in industries such as retail trade and public
administration. Finally, zip codes with a higher percentage of male population
are also good locations to have a gas station in order to maximize profits.
Policy suggestions directly correspond to the duality of the results based on gas-station markups across zip codes. In particular, if the main objective is to reduce the redistributive effects of information frictions across locations, the corresponding suggestion is that the policy makers should consider the heterogeneity of consumer search behavior across markets (where the heterogeneity has been shown to depend on geographic, demographic and economic conditions) by conducting local policies rather than a common multi-location policy.