Abstract
Rural retail trade industries can serve an export function for and contribute to the economic growth of a rural place. The importance of a retail trade industry to the economy of a rural place depends on its market area. Since those market areas often extend into a place’s hinterlands, retail trade can effectively have an export component for the place itself. Yet empirical determination of the export role of a rural retail trade industry has been historically confounded by data limitations which limit testing for and identification of an industry’s market area. Using geographic information system tools to define variables for differing driving distances (as contrasted with circles defined by Euclidean distance) from a place, this paper tests for the market areas of seven retail trade industries in rural places and for the impact of retail trade in hinterlands on those market areas. This paper contributes to the relevant literature in three ways. First, it tests formally for a retail industry’s “market area driving time.” Prior articles which have incorporated distance into their analyses have not generally tested for the actual travel times more appropriate for such analyses. Second, it finds that not undertaking such testing might produce misleading results regarding the relationship between a place and the hinterlands in its market areas. Third, the regression results suggest subtlety in the export nature of rural retail trade, with some industries losing their export nature when competing businesses appear in hinterlands (Weak Export Retail industries) and other industries retaining their export nature under those circumstances (Strong Export Retail industries).
Similar content being viewed by others
Notes
Articles which have addressed simultaneity between two variables in which one dependent variable is a count variable have generally assumed that the second variable is either binary or that it has a linear reduced form (e.g., Windmeijer and Santos Silva 1997; Wooldridge 2010: 742–747). Those models are not relevant because we have a count-data regressor which is possibly endogenous, whose reduced form will not be linear.
The vector z will include the elements of x \(_\mathbf{1}\) which are exogenous and the instruments for \({y}_{n}\) (the endogenous variable).
All of the \(42\, \hbox {R}^{2}\)’s exceeded 0.49, with thirty-one of the \(\hbox {R}^{2}\)’s exceeding 0.60.
Data on the exact sizes of establishments and their distribution, which would preferable, are not readily available. There are data available on size classes of establishments, the number of establishments that exist in a particular range of employment levels. However, using those categories would lead to problematic discontinuities in empirical analyses.
The states were Colorado, Idaho, Montana, Utah, Iowa, Kansas, Minnesota, Missouri, Nebraska, North Dakota and South Dakota.
Microsoft MapPoint calculations are based on posted speeds and average travel time. The default settings were used to construct our radii and were cross-checked with sample data in ArcGIS as well as Google Maps.
The dependent variable and the regressors in the \(P\) test regression vary according to the hypothesis chosen to be null.
Both the 20- and 30-min driving time are rejected when they are paired, and the 30- and 40-min intervals are accepted when they are paired. The 50-min interval is superior when tested against the 40-min interval and is no different than the 60-min interval. When the 20-min distance is tested against the 50-min interval, both intervals are rejected. The same is true when the 20-min interval is tested against the 60-min interval.
It would be tempting to search all of the regression results for explanations of the differences between the results here and those in Mushinski and Weiler (2002). Because their definition of neighboring areas as the remainder of the county in which a place was located allowed for market areas of differing shape and size, such a search was not meaningfully possible.
References
Beavon KSO, Hay AM (1978) Long run average cost, price and Christaller’s concepts of range: an explanatory note. Geography 63(2):98–100
Berry BJL (1967) Geography of market centers and retail distribution. Prentice-Hall Inc, Englewood Cliffs
Berry BJL, Garrison WL (1958) A note on central place theory and the range of a good. Econ Geogr 34(4):304–311
Cho S-H et al (2011) Relationship between value of open space and distance from housing locations within a community. J Geogr Syst 12:393–414
Christaller W (1933) Die zentralen Orte in Süddeutschland: Eine ökonomisch geographische Untersuchung über die Gesetzmäßigkeit der Verbreitung and Entwicklung der Siedlungen mit städtischen Funktionen. Doctoral dissertation (Gustav Fischer Verlag, Jena). Translated by Baskin C (1966) Central places in Southern Germany. Prentice Hall, Englewood Cliffs
Davidson R, MacKinnon JG (2004) Econometric theory and methods. Oxford University Press, Cambridge
Davidson R, MacKinnon JG (1981) Several tests for model specification in the presence of alternative hypotheses. Econometrica 49(3):781–793
Deller SC, Harris TR (1993) Estimation of minimum market thresholds for rural commercial sectors using stochastic frontier estimators. Reg Sci Perspect 23(1):3–17
Eaton BC, Lipsey RG (1982) An economic theory of central places. Econ J 92:56–72
Golledge RG, Rushton G, Clark WAV (1966) Some spatial characteristics of Iowa’s dispersed farm population and their implications for the grouping of central place functions. Econ Geogr 42(3):261–272
Hanjoul P, Beguin H, Thill J-C (1989) Advances in the theory of market areas. Geogr Anal 21(3):185–196
Henderson JW, Kelly TM, Taylor BA (2000) The impact of agglomeration economies on estimated demand thresholds: an extension of Wensley and Stabler. J Reg Sci 40(4):719–733
Isard W (1956) Location and space-economy. Wiley, New York
Kaiser MS, Cressie N (1997) Modeling Poisson variables with positive spatial dependence. Stat Probab Lett 35:423–432
Launhardt W (1885) Mathematische Begrundung der Volkswirtschaftslehre. W. Englemann, Leipzig
Lösch A (1954) The economics of location. Yale University Press, New Haven
Mullahy J (1997) Instrumental-variable estimation of count data models: applications to models of cigarette smoking behavior. Rev Econ Stat 79(4):586–593
Mulligan GF (1984) Agglomeration and central place theory: a review of the literature. Int Reg Sci Rev 9(1):1–42
Mulligan GF, Fik TJ (1994) Price and location conjectures in medium- and long-run spatial competition models. J Reg Sci 34(2):179–198
Mulligan GF, Fik TJ (1995) Consistent price and location conjectures in spatial competition models. Ann Reg Sci 29:91–109
Mulligan GF, Partridge MD, Carruthers JI (2012) Central place theory and its reemergence in regional science. Ann Reg Sci 48:405–431
Mushinski D, Weiler S (2002) A note on the geographic interdependencies of retail market areas. J Reg Sci 42(1):75–86
O’Kelly ME, Miller H (1989) A synthesis of some market area delimitation models. Growth Change 20(3):14–33
Rietveld P et al (1999) On the relationship between travel time and travel distance of commuters. Ann Reg Sci 33:269–287
Rushton G, Golledge RG, Clark WAV (1967) Formulation and test of a normative model for the spatial allocation of grocery expenditures by a dispersed population. Ann Assoc Am Geogr 57(2):389–400
Shields M, Kures M (2006) Black out of the blue light: an analysis of Kmart store closing decisions. J Retail Consum Serv 14:259–268
Shonkwiler JS, Harris TR (1993) A non-Gaussian time series analysis of rural retail business counts. J Reg Sci 33:37–48
Shonkwiler JS, Harris TR (1996) Rural retail business thresholds and interdependencies. J Reg Sci 36: 617–630
Shonkwiler JS, Harris TR (1997) Interdependence of retail business. Growth Change 27:520–533
Thill J-C (1992) Spatial competition and market interdependence. Pap Reg Sci 71(3):259–275
Thill J-C (1986) A note on multipurpose and multistop shopping, sales, and market areas of firms. J Reg Sci 26:775–778
US Department of Commerce (2002) Census of retail trade: geographic area series. Bureau of the Census, Washington, DC
Windmeijer F (2002) ExpEnd, a GAUSS programme for non-linear GMM estimation of exponential models with endogenous regressors for cross-section and panel data. Centre for Microdata Methods and Practice. Institute for Fiscal Studies, London
Windmeijer FAG, Santos Silva JMC (1997) Endogeneity in count data models: an application to demand for health care. J Appl Econom 12:281–294
Wooldridge JM (2010) Econometric analysis of cross section and panel data, 2nd edn. The MIT Press, Cambridge
Yu T-H et al (2012) Assessing the residential property tax revenue impact of a shopping center. J Real Estate Financ Econ 45:604–621
Acknowledgments
The authors thank Wade Hudson for his research assistance, Eric Thompson for an insightful comment, participants in a seminar at the University of Colorado-Denver, and the Centre For Microdata Methods and Practice at the Institute For Fiscal Studies and the Department of Economics at the University College London for providing the GAUSS programming code used in our estimation.
Author information
Authors and Affiliations
Corresponding author
Appendix: Description of the hypothesis testing analysis
Appendix: Description of the hypothesis testing analysis
The details of the hypothesis testing analysis are provided using Table 3, which presents \(P\) test results for three industries: Gasoline Stations (4471), Health and Personal Care Stores (4461) and Building Material and Supplies Dealers (4441). The analysis for the Gasoline Stations industry is presented first, since a clear testing result is obtained. Table 3 indicates that when the 20-min driving distance is tested against the 30-min distance, the 20-min driving distance is accepted when it is the null hypothesis and the 30-min distance is rejected when it is the null hypothesis. These results imply that the 20-min distance is preferred when compared with the 30-min distance. Table 3 also indicates that when the 30-min driving distance is tested against the 40-min driving distance, the 30-min driving distance is accepted when it is the null hypothesis and the 40-min driving distance is accepted when it is the null hypothesis. The testing results are inconclusive; both driving distances are equally acceptable as a possible driving distance. When this result is combined with the testing result for the 20- versus 30-min driving times, it is concluded that the 20-min driving time is superior over both the 30- and 40-min driving times. Test results for the 40- and 50-min driving times indicate that the former is more likely when compared with the latter; the 40-min driving time is accepted when it is the null hypothesis and the 50-min driving time is rejected. Combining this testing result with the earlier test result, one may conclude that the 20-min driving time is the most likely. Since the test of the 50- and 60-min driving times are inconclusive, we ultimately concluded that the market area is households within a 20-min driving time.
Health and Personal Care establishments are examined next, as testing results for adjacent driving times yield two preferred driving times with one of those times rejected when the two driving times are tested against each other. Testing results for adjacent driving times support 30 min as being more likely when tested against the 20- and 40-min driving times, and they support the 50-min driving time when tested against the 20- and 60-min driving times. Since testing for adjacent driving times is inconclusive with regard to the 30- and 50-min driving times, the 30-min driving time was tested against the 50-min driving time. A test of the 30-min driving time against the 50-min driving time indicates that the 30-min driving time is preferred. Thus, the industry’s market area includes households within a 30-min driving time.
Finally, the test results for the Building Materials Establishments industry are considered since they provide an example of an inconclusive analysis. The pairwise tests indicated only that the 20-min driving distance is not superior when tested against the 30-min driving distance. Otherwise, no conclusions are possible for driving distances of 30 min or more. In cases such as this, we concluded that the more likely market area is the shortest driving time among the times for which one can reach no conclusions. Because inclusion of the incremental area does not affect the relationship, it does not matter to the relationship. In a situation where all other things are equal, we chose shorter driving times over longer ones. Thus, we concluded that the most likely driving time is 30 min. Table 4 identifies test results for the remaining industries.
Rights and permissions
About this article
Cite this article
Mushinski, D., Weiler, S. & Widner, B. The impact of retail establishments in hinterlands on the export role of retail establishments in rural places. Ann Reg Sci 52, 469–487 (2014). https://doi.org/10.1007/s00168-014-0595-3
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00168-014-0595-3