Critical Access Hospitals and Retail Activity: An Empirical Analysis in Oklahoma

ORIGINAL ARTICLE

Critical Access Hospitals and Retail Activity: An EmpiricalAnalysis in OklahomaLara Brooks, MS & Brian E. Whitacre, PhD

Department of Agricultural Economics, Oklahoma State University, Stillwater, Oklahoma

For further information, contact: Lara Brooks,

MS, Department of Agricultural Economics, 526

Ag Hall, Oklahoma State University, Stillwater,

OK 74074; e-mail [email protected].

doi: 10.1111/j.1748-0361.2010.00336.x

Abstract

Purpose: This paper takes an empirical approach to determining the effectthat a critical access hospital (CAH) has on local retail activity. Previous re-search on the relationship between hospitals and economic development hasprimarily focused on single-case, multiplier-oriented analysis. However, as theefficacy of federal and state-level rural health subsidies come under increasingscrutiny, more comprehensive investigations can provide support for contin-ued funding.Methods: Data from 105 rural Oklahoma communities are used to explorewhether the presence of a CAH impacts several measures of retail activity. Themeasures are: total retail sales, total number of retail establishments, and num-ber of micro and small retail establishments. Ordinary least squares regressionis used to evaluate the impact of a CAH after controlling for a host of otherfactors influencing retail activity such as local demographics, unemploymentrates, and the presence of a Wal-Mart.Findings: The presence of a CAH has a positive and significant influenceon each measure of retail activity. The parameter estimates suggest that aCAH has a similar influence on rural retail sales as a Wal-Mart, increasingtotal retail sales by approximately 28% over towns without a CAH. Othermodel results indicate that a CAH presence significantly increases the num-ber of total retail establishments and the number of micro and small businessestablishments.Conclusions: The positive results provide additional evidence on the far-reaching economic development impacts of CAHs. The results also emphasizethe importance of continued support for these rural institutions, including fed-eral and state subsidies.

Key words Critical Access Hospital, economic impact, retail, rural health.

Health care plays a large role in the US economy.Health care expenditures accounted for 16% of thegross domestic product in 2006, with hospital care ac-counting for nearly one-third of these expenditures.1

This is a substantial increase from 1990 when healthcare expenditures accounted for 12.3% of the gross do-mestic product.2 In addition to their strong and in-creasing presence in the overall economy, hospitalsand the health care they provide play crucial roles ineconomic development due to their ability to influ-

ence employment and enhance quality of life in theircommunities.3,4

This effect becomes increasingly important in rural ar-eas. A high-quality health sector is vital for rural com-munities to attract industry, businesses, new residents,and retirees. While contributing to an improved qualityof life, the health sector also has direct impacts on theemployment and income of the local community.3 Ruralhospitals, often the face of health care in rural communi-ties, play a vital yet underappreciated role in economic

The Journal of Rural Health 27 (2011) 29–38 c© 2010 National Rural Health Association 29

Critical Access Hospitals Brooks and Whitacre

development. Oftentimes, a rural hospital follows onlythe local school system as the largest employer in thecommunity while employing some of the highest paid in-dividuals in the community.5,6

While most hospitals acquire some type of public finan-cial assistance, the economics of rural health provisionsuggest that rural hospitals require a significant amountof intervention beyond the free market. Many rural hos-pitals have utilized and are even sometimes dependentupon federal and state-level programs that subsidize theirexistence. The Balanced Budget Refinement Act (BBRA)of 1999 provided assistance to rural hospitals in remoteareas by introducing the Critical Access Hospital program(CAH).7 However, as competition for federal and statetax dollars has increased, programs receiving significantamounts of aid are often targeted for cuts or at least sub-jected to high levels of scrutiny. Some have even ques-tioned the efficacy of tax-based health programs, such aswhether nonprofit hospitals actually cover the cost of thetax subsidies they receive.8 Others have questioned theamount of help states provide in terms of rural healthcare delivery.9 To combat these efforts, rural health advo-cates should seek to quantify all relevant economic con-tributions of their hospitals. While most analytical workon this topic focuses on general income and employmentmeasures, a community’s retail activity can also be greatlyimpacted by the presence of a hospital. This paper focuseson and empirically quantifies the underanalyzed relation-ship between CAHs and community retail activity.

Retail activity can be spurred by out-of-town hospitalusers who shop while in the area, multiplier impacts ofhospital employee paychecks, and even the creation ofretail stores that may cater to hospital patients and fam-ilies (a local card or gift store, for example). The Amer-ican Hospital Association found, by using a traditionalmultiplier-based analysis, that hospitals have an impactof $76.7 billion on retail trade throughout the UnitedStates.1 At the opposite end of the spectrum, a hospitalclosure is extremely detrimental to retail activity in smallcommunities. For example, Doeksen and associates esti-mate that Perry, a rural community in Oklahoma, couldexperience a loss of over $2 million in retail sales if thehospital were to close.3 While the closing of a hospitalis excruciating in terms of job losses and anxiety aboutthe proximity of emergency health services, the poten-tial loss of sales tax collections from lower retail activityis an additional blow to the community. Traditional mul-tiplier analysis or case studies such as these demonstratesome of the economic impacts of a hospital’s presence;however, they do not empirically estimate differences be-tween communities with and without a hospital. For ex-ample, is it the case that the presence of a rural hospitalhas a larger impact on retail activity than, say, the pres-

ence of a Wal-Mart? This paper uses Oklahoma data toaddress that shortcoming and further quantify some ofthe economic relationships associated with rural hospi-tals. At an aggregate level, documenting these impactscan provide significant support for policies with a ruralhospital focus.

Background

Previous research has shown that income received byhealth care employees has a positive impact, both directand secondary, on retail sales.10 While numerous studiesreadily estimate potential retail sales impacts from the lo-cal health care sector on a case-by-case basis,11-13 there isa lack of empirical research demonstrating that the pres-ence of a CAH has a statistically significant impact onactual retail activity. In other words, no studies that weare aware of sought to estimate whether rural communi-ties with hospitals generate higher retail sales (or have ahigher number of retail establishments) than those with-out. The value-added of the current study is its extensionof the literature into a statistical realm that might impactfuture policy decisions.

Most studies on the economic impact of hospitalsuse multiplier-oriented analysis, where historical dataand input-output software programs such as IMPLAN(Minnesota IMPLAN Group, Inc., Hudson, WI) are usedto estimate the linkages between economic sectors.14

These studies can tell us, for instance, that each $1.00 inhospital payroll creates an additional $0.48 throughoutthe rest of the community economy. This technique canbe applied to the total hospital payroll to create an aggre-gate income impact, and historical estimates of the per-centage of income spent on retail sales can then be usedto estimate the hospital’s impact on retail. These meth-ods generated retail impacts of $2 million and $6.4 mil-lion, respectively, from hospitals in the rural communi-ties of Perry and Atoka, Oklahoma.3,10 Similarly, suchimpact studies could be performed for other industries,such as evaluating the economic impact of a manufactur-ing plant. Is it the case, then, that areas with hospitalshave an advantage over areas with other types of infras-tructure? This paper attempts to answer those questions.

This paper varies from previous research on the rela-tionship between CAHs and retail sales by analyzing a fullcomplement of rural communities (some with hospitals,some without) to uncover variables that influence retailactivity. Of particular interest is whether the presence ofa CAH statistically influences various types of retail activ-ity, and if so, by how much. As previously noted, mostavailable research utilizes input-output analysis to deter-mine the impact of a CAH on a local economy. This is

30 The Journal of Rural Health 27 (2011) 29–38 c© 2010 National Rural Health Association

Brooks and Whitacre Critical Access Hospitals

important to quantify the estimated dollar impact a par-ticular CAH has on an economy and provides useful po-litical talking points; however, at an aggregate level, suchstudies provide little more than anecdotal evidence.

Although anecdotal evidence is useful at times, somestudies have questioned the efficacy of the entire CAHprogram. While the BBRA legislation was designed to of-fer financial support to rural hospitals, several CAHs uti-lize a local subsidy in addition to the benefits receivedfrom a CAH designation. In fact, CAHs in Kansas expe-rienced an increase in local subsidies of 38.4% between1994 and 2001.7 This is the opposite effect that was ex-pected from the CAH program. Additionally, the CAHhospitals in Kansas that received the greatest amountof subsidies were the ones with fewer beds and feweradmissions—raising the question of whether supportingthese hospitals is adversely impacting the overall level ofcare in rural parts of the state.7 These situations are mostlikely mirrored in Oklahoma, where data for this studywere collected. In 2007, Oklahoma municipal and countygovernments contributed nearly $833 million to supporthospitals, an amount higher than the $620 million ingovernment contributions for hospitals in Kansas.15 Theavailable data do not indicate how much of this amountwas allocated for rural hospitals or even CAHs. However,Zimmerman and McAdams found in 2001 that CAHs inKansas received $227,130 on average per hospital com-pared to an average of $85,985 per non-CAH hospital.7

Defending this type of financial support requires large-scale analysis of the communities receiving the aid. Evi-dence that the presence of a CAH significantly increasesretail activity in a community would be a useful tool inthis defense.

There is not an abundance of statistical, empirical anal-ysis on this topic. Probst and colleagues16 used economet-ric modeling to comparatively determine the effects of ahospital closure. They found that counties where a hospi-tal had previously closed experienced lower incomes andslower employment growth than counties without a hos-pital closure. This research provides insight into incomeand employment loss, but it does not focus specificallyon the impact these hospitals have on retail sales. Gener-ally, research to date focuses on the potential economicimpact of the health sector, or of losing a hospital. Thispaper will expand the literature to more closely examinethe relationship between CAHs and retail activity.

Methods

Variables

One hundred and five Oklahoma towns were selected forthe study by virtue of meeting certain specifications: all

towns included had to collect sales tax during the yearsof 2000 to 2005; the average population (taken from theUS Census) from 2000 to 2005 had to be within the rangeof 1,000 to 5,000; and they had to be considered “rural.”For the purpose of this study, rural is defined by usingRural-Urban Commuting Area (RUCA) codes from theUSDA, effectively eliminating smaller towns with a highdependency on nearby urban areas. Therefore, ZIP codeswith a RUCA code of 1-3 were removed. Communitieswith a RUCA code of 4-10 all remained in the sample.We restricted our analysis to towns with populations be-tween 1,000 and 5,000 for several reasons. First, there isa high incidence of hospitals in towns with populationsover 5,000. Of the 63 towns in Oklahoma with popula-tions between 5,000 and 10,000, 43 (68%) have a hospi-tal, which limits the variation in our data. Further, centralplace theory suggests that the level of retail activity willincrease with population, and we want to control for thiscorrelation using some population cutoff point. To checkthe robustness of our results, we also analyze towns withpopulations up to 10,000.

Hospitals with a Critical Access designation were uti-lized for this study. CAHs were selected since they arelocated in more remote areas, are by definition smallerfacilities, and are more typical for communities between1,000 and 5,000. CAHs are generally at least 35 milesfrom another hospital and have a maximum of 25 beds.17

We focus on CAHs as opposed to small Prospective Pay-ment System (PPS) hospitals for the following reasons:(1) CAHs generally receive more financial support thanPPS hospitals, and (2) the PPS hospitals that fall withinour eligibility guidelines of population and rurality arenot overly small. The 9 PPS hospitals in communities withpopulations between 1,000 and 5,000 and with RUCAcodes between 4 and 10 have an average size of 43beds, which is more than a 70% increase in size froma CAH. These larger hospitals may increase economic ac-tivity simply due to their size (higher employment base,more business-to-business purchases), and we want toevaluate the impact of hospitals that are more commonin rural Oklahoma. Thus, we restrict our analysis to CAHswithin our study criteria. The 2009 Oklahoma State De-partment of Health’s Medical Facility Directory and theOklahoma State Office of Rural Health indicated that 25of the 105 towns selected were homes to CAHs duringthe time of the analysis.18 While Oklahoma contains 33CAHs, 8 were not included in our dataset since theirtown’s population did not fit our criteria (between 1,000and 5,000) or had RUCA codes between 1 and 3. Figure 1displays the location of the selected towns and also in-dicates which of these towns have a CAH. As Figure 1shows, the included towns are geographically dispersedacross the state.



Figure 1 Location of Towns Used in Study.

Retail sales estimates were obtained by using sales taxcollections at the town/city level from the Oklahoma TaxCommission. To account for yearly fluctuations associ-ated with traditional business cycles, average retail salesfrom 2000 to 2005 were used. A total of 4 models areutilized to determine the influence that CAHs have onvarious measures of retail activity. These models use (1)total retail sales, (2) the total number of retail establish-ments, (3) the number of micro (1-4 employees) retail es-tablishments, and (4) the number of small (5-19 employ-ees) retail establishments as dependent variables. Thesevariables present a picture of the type of retail activityCAHs may be influencing in addition to more traditionalsummary-level measurements. The numbers of total re-tail, micro retail, and small retail establishments were de-rived from the US Census Bureau through the ZIP-CodeBusiness Patterns. Again, an average from 2000 to 2005was used to smooth potential data fluctuations.

The economic literature suggests that any number ofvariables can influence retail activity. Retail sales are verylikely correlated with the number of total retail estab-lishments in a community.19 County seats are often cen-ters of activity for nonmetropolitan areas, so a dummyvariable controlling for their presence is included to helpobtain a precise estimate on the impact of a CAH. Unem-ployment rates are also included, as high levels of unem-ployment are likely to reduce retail activity.

Dependency on a particular sector, such as farming ormanufacturing, can also affect retail activity. Microbusi-nesses were found to be more prevalent and contributedmore to local sales in farming/rural areas.20 Dummy vari-ables for county-level dependencies on farming, manu-facturing, and government sectors were taken from theUSDA and applied to the current dataset.

The presence of Wal-Mart in a rural community notonly presents mixed emotions, but also mixed economicresults. Goetz and Rupasingha found that the presence ofWal-Mart decreases social capital and can potentially re-

duce economic growth for communities.21 Artz and Stonefound that the presence of a Wal-Mart Supercenter in anonmetropolitan community can decrease local grocerystore sales by nearly 17% within the first 2 years of open-ing.22 Irwin and Clark found similar results of Wal-Martbeing a detriment to small retailers, but they do state thatthe opening of a new Wal-Mart has the opportunity tostimulate retail sales for the community by attracting out-side shoppers.23

Geography can also influence retail activity. Oklahomais roughly split in half by Interstate 35, so a locationdummy variable for a location east of Interstate 35 is in-cluded. The eastern part of the state has experienced sig-nificantly more growth since 1990. In fact, counties in theeastern half saw their population increase by 3% over the2000-2008 period, while the western half experienced apopulation decline of 1%. Population trends and expec-tations can impact business location decisions, so the lo-cation dummy variable attempts to account for this aspectof a rural community’s economy. In addition, a continu-ous variable for distance from an interstate was includedto further dissect the importance of location. While a CAHis by definition at least 35 miles from another hospital,this does not necessarily account for a community’s dis-tance from other neighboring locations or sources of cus-tomers. A CAH community may benefit from being lo-cated directly off of a major interstate, which can pos-itively impact retail activity due to higher amounts oftraffic, while a similar-sized community located far awayfrom an interstate would not have this advantage. There-fore, the distance variable identifies the relative isolationof the community.

Demographic characteristics of community residentswere also considered in variable selection. Three cate-gorical age variables are included (with the proportionover age 65 excluded as a default) to determine if agecomposition significantly impacts retail activity. Incomecan be assumed to have a positive impact on retail sales,



Table 1 Independent and Dependent Variables

Variable Type Description Data Source

Independent Variables

Population Continuous Community-level population (2000-2005) Census

CAH 0/1 Critical Access Hospital Oklahoma Department

of Health

Wal-Mart 0/1 Wal-Mart present in town/city Walmart.com

County Seat 0/1 County seat OK County Data

I35 0/1 Location variable for east of I-35 Oklahoma Dept. of

Transportation

Farm Dep 0/1 Farming-dependent county USDA ERS

Manufacturing Dep 0/1 Manufacturing-dependent county USDA ERS

Govt Dep 0/1 Government-dependent county USDA ERS

Household Income Continuous Household income (2000) Census

Household Size Continuous Household size (2000) Census

Age under 19 Continuous Percentage of population under 19 (2000) Census

Age 20-44 Continuous Percentage of population under 20-44 (2000) Census

Age 45-64 Continuous Percentage of population 44-64 (2000) Census

Unemployment Continuous Average unemployment rate (2000-2005) BLS

Distance Continuous Distance (miles) from interstate GIS

Retail Est. Continuous Average number of retail establishments (2000-2005) Census

Dependent Variables

Retail Sales Continuous Average total retail sales (2000-2005) Oklahoma Tax

Commission

Retail Est. Continuous Average number of retail establishments (2000-2005) Census

Micro Business Continuous Average number of micro retail est. (1-4 employees) (2000-2005) Census

Small Business Continuous Average number of small retail est. (5-19 employees) (2000-2005) Census

so average household income is included as an indepen-dent variable. Household size can also play a role in re-tail activity, but a potentially negative one since largerhouseholds can share the purchase of some items. Alldata with the exception of the farm-, manufacturing-,and government-dependent counties, and the unemploy-ment rate, are at the community or ZIP code level.Industry-dependent data and unemployment rates wereonly available at the county level. While some variablesare available on an annual basis, others are restrictedto decennial census measures. Table 1 lists the variablesused in the econometric models, along with their type(binary or continuous), year, and data source.

Descriptive Statistics

When comparing communities with a CAH to those with-out, t tests on variable means uncovered several signifi-cant differences (Table 2). Average population was sig-nificantly higher in communities with a CAH, but weattempted to minimize the importance of this variableby restricting our analysis to towns between 1,000 and5,000. Furthermore, there were significant discrepanciesin age, with larger proportions of younger groups (under19 and 20-44) present in communities without a CAH.

This is intuitive, since proximate health care is increas-ingly important for older (45+) age categories. Residentsof communities with a CAH have longer to travel to getto an interstate, as evidenced by the significant distanceterm. In terms of retail activity, significant differences ex-ist for all 4 measures, as communities that have a CAHdisplay higher levels of retail establishments, retail sales,and micro and small retail businesses. This suggests thatthe availability of a CAH may in fact influence differentmeasures of retail. We turn to our econometric modelsto uncover the distinct impact of CAH presence on retailactivity.

Econometric Models

Four different dependent variables are used in our anal-ysis to see whether CAH status impacts various measuresof retail activity. These variables include: (1) community-level retail sales (RS), (2) number of retail establishments(Retail Est.), (3) number of micro retail establishments(Micro Business), and (4) number of small retail estab-lishments (Small Business). Given the continuous natureof the dependent variables, ordinary least squares (OLS)modeling was used to determine the impacts that the se-lected independent variables have on the various retail



Table 2 Descriptive Statistics

Observations with CAH Observations without CAH

Mean SD Mean SD

CAH 1 0 0 0

Population 2,672 1,075 2,048 1,028∗∗

Wal-Mart 0.160 0.374 0.113 0.318

County Seat 0.440 0.507 0.338 0.476

I35 0.480 0.510 0.550 0.501

Farm Dep 0.280 0.458 0.175 0.382

Manufacturing Dep 0.120 0.332 0.163 0.371

Govt Dep 0.120 0.332 0.125 0.371

Household Income 25,785 4,970 24,823 4,829

Household Size 2.378 0.077 2.439 0.143∗∗

Age under 19 0.281 0.019 0.294 0.042∗∗

Age 20-44 0.303 0.039 0.322 0.060∗∗

Age 45-64 0.212 0.016 0.207 0.025∗∗

Age over 65 0.204 0.023 0.177 0.045∗∗

Unemployment 4.539 1.093 4.780 1.043

Distance 36.994 32.113 23.848 24.599∗

Retail Est. 20.833 8.619 14.842 10.826∗∗

Retail Sales 22,344,898 14,534,037 14,296,629 13,371,420∗∗

Micro Business 10.180 4.060 7.392 5.395∗∗

Small Business 8.767 3.850 6.196 4.407∗∗

Observations n = 25 n = 80

∗Means are statistically different at the P = .10 level.∗∗Means are statistically different at the P = .05 level.

measures. Models 2-4 use count data for their depen-dent variables. Rather than dealing with the difficult-to-interpret coefficients from a Poisson regression (which issometimes used with count data), we took square roots ofthe dependent variables in these models. This addressesthe normality requirement for using OLS and also findssupport from numerous studies on the topic.24-26 The ba-sic OLS model takes the form:

yi = B ′Xi + εi

Where yi depicts the dependent variables that will be uti-lized for community i, Xi is a vector of independent vari-ables that are hypothesized to impact the retail activity ofa community, B′ symbolizes the parameter estimates forthe vector Xi, and εi is the associated error term. For ex-ample, in Model (1), yi represents average retail sales incommunity i, which is melded based on characteristics ofcommunity i (Xi).

The dependent variables for yi are the 4 previouslymentioned measures of retail activity. The explanatoryvariables represented in vector X are very similar for eachof the 4 models. Dummy variables for CAH, Wal-Mart,and County Seat are present in all 4 models. CAH andCounty Seat were expected to have a positive impactin all 4 models, while Wal-Mart was expected to havea positive impact on retail sales but not on retail estab-

lishments. There are 3 county-level industry-specific vari-ables: farm dependent (Farm Dep), manufacturing (Man-ufacturing Dep), and government dependent (Govt Dep).There are 2 location variables: I35 is a dummy variablefor if the community is located east of Interstate 35, andDistance represents the actual number of miles the com-munity is located from an interstate. Distance was ex-pected to have a negative impact in each of the 4 modelssince distance from a major roadway lessens a commu-nity’s ability to attract shoppers.

Three continuous variables pertaining to local eco-nomic and demographic data are included in each of themodels. Household Income refers to the median house-hold income on the community level. This variable wasexpected to have a positive impact on retail activity sincehigher-income levels suggest greater ability to spend onretail items. Given the dramatically different units forboth this variable and retail sales, both are converted tologarithmic form so that the OLS normality assumptionholds. Household Size represents the average householdsize on the community level. This value was expected tobe negative since the household per capita income hasthe possibility of being lower with a higher number ofoccupants. Unemployment represents the average (2000-2005) unemployment rate, which likely has a negativerelationship with retail activity.



All 4 models also include the proportion of residents inthese age categories: Age Under 19, Age 20-44, and Age45-64. The age group of over 65 is used as the default cat-egory in this sample. It was expected that the age groupsof 20-44 and 45-64 would have a positive impact on re-tail activity since these 2 groups account for the majorityof the workforce. Finally, the number of retail establish-ments is included in the model for retail sales, since thenumber of stores will likely impact total sales.

Results

The 4 models listed above attempt to identify the impact(if any) that CAHs have on retail activity through actualretail sales and the number of total, micro, and smallretail establishments per town. The existence of het-eroskedasticity was tested using the Breusch-Pagan Test,and the only model for which it was an issue (Model 1:test statistic of 29.1, P value = .015) consequently in-corporated heteroskedastic-consisted standard errors tocorrect the problem. Each dependent variable was alsotested for normality using a skewness-kurtosis test, withno problematic results. Table 3 displays the results for the4 models.

In Model (1), the dependent variable is the log of av-erage retail sales. CAHs have a positive and significantimpact on this variable, which was anticipated since mostcase studies assert the importance of CAHs to retail sales.Given the log-linear nature of model (1), the interpreta-tion of a dummy variable is not completely straightfor-ward. Here the percentage impact on retail sales is givenby the formula 100 ∗ [exp(0.2434) − 1] = 27.6%.27 Thus,retail sales are 27.6% higher in communities with CAHs,which is similar to the 26.5% increase shown for com-munities with a Wal-Mart.

The only other parameters that are significant arethe expected positive values on household income andthe number of retail establishments (Retail Est.). Sincehousehold income is also taken as a logarithmic vari-able, its interpretation is as an elasticity: a 1% increasein the household income level would lead to a 0.73%increase in retail sales collected. This is again fairly in-tuitive, as part of the increased income would likely bespent in retail. The parameter on the number of retail es-tablishments suggests that another retail store would in-crease retail sales by 6.4%, again a reasonable value giventhe average number of stores per community (16) in oursample.

Models (2), (3), and (4) produce strikingly similar re-sults, which is not surprising given that all dependentvariables are related to general retail activity. In each case,the presence of a CAH is positive and statistically signifi-

cant at the 5% level, which supports our general hypoth-esis that the presence of these hospitals plays a substantialrole in viability of the retail sector. In addition, the pres-ence of a Wal-Mart is also positive and highly significant.This impact was somewhat unexpected since the major-ity of research on the topic has indicated that Wal-Martcan hurt small businesses; however, other research hassuggested that Wal-Mart may actually encourage nichebusinesses.22,23 The age categories are nearly all signifi-cant and negatively related to the number of retail estab-lishments, suggesting that higher proportions of youngerresidents would lead to a reduction in retail establish-ments. By default, however, this implies that larger pro-portions of older residents would lead to more establish-ments. Aside from age, the presence of a CAH, and thepresence of a Wal-Mart, no other variables are significantin any of the establishment models. This suggests that thefactors discussed above are the dominant ones impactingthe number of retail establishments.

As previously noted, to test the robustness of our re-sults, we enlarged our sample to include towns with pop-ulations between 1,000 and 10,000. The pattern of resultsheld, with the CAH and Wal-Mart variables continuing toshow positive and highly significant influence on each ofthe 4 retail activity measures.

Discussion and Conclusion

Generally, the econometric models confirm that CAHs dohave statistically significant and positive impacts on re-tail activity, including the amount of retail sales and thenumber of retail establishments. In all 4 models, CAH andWal-Mart were the only variables that were consistentlystatistically significant and positive. This suggests that,even after other factors are accounted for (such as thepresence of a county seat, low unemployment levels, ora strong manufacturing sector), having a CAH in a com-munity leads to higher levels of retail activity. The empir-ical results strengthen previous economic impact studiestypically performed on single communities using input-output analysis.

This research incorporates data that somewhat limit thelevel of analysis. In particular, data on patient (or re-tail customer) origin are not available. Therefore, whileCAH communities show increased retail sales, it is un-certain from exactly where they are attracting customers.While retail sales, local sales tax collections, and retailestablishments were utilized in this study, the type ofsales or retail establishments was not evaluated. Futureresearch could examine more in-depth measures of salestax collections by including Standard Industrial Classifi-cation codes or types of retail establishments by North



Table

3OrdinaryLeastS

qua

resReg

ressionResults

Mod

el(1)

DV:Log

ofAverage

RetailSales

Mod

el(2)

DV:TotalRetail

Establishm

ents(SqRt)

Mod

el(3)

DV:M

icro

BusinessRetail

Establishm

ents(SqRt)

Mod

el(4)

DV:SmallB

usinessRetail

Establishm

ents(SqRt)

Param

eter

SEPValue

Param

eter

SEPValue

Param

eter

SEPValue

Param

eter

SEPValue

CAH

0.24

340.10

11.003

1∗∗

0.67

480.24

74.007

7∗∗

0.50

000.18

90.009

6∗∗

0.42

060.16

53.012

6∗∗

Wal-M

art

0.23

550.15

34.065

2∗1.66

960.32

00<

.000

1∗∗

1.03

270.24

45<

.000

1∗∗

0.99

600.21

38<

.000

1∗∗

Cou

ntySe

at0.05

500.08

59.507

60.07

760.21

65.720

90.03

160.16

54.848

80.06

460.14

46.656

1

I35

0.19

020.12

35.148

50.38

590.31

00.216

50.35

180.23

68.140

90.21

320.20

71.306

0

Farm

Dep

−0.033

90.13

93.758

1−0

.476

40.34

77.174

1−0

.149

10.26

56.575

9−0

.381

00.23

23.104

5

Man

ufacturing

Dep

−0.138

00.12

63.201

00.39

840.31

50.209

10.23

560.24

06.330

10.31

050.21

04.143

5

Gov

tDep

−0.046

00.13

91.713

7−0

.307

80.35

04.382

0−0

.374

20.26

77.165

6−0

.123

60.23

41.598

7

Log(Hou

seho

ldInco

me)

0.73

090.28

64.012

4∗∗

0.35

760.72

28.622

0−0

.031

00.55

22.955

30.55

170.48

29.256

2

Hou

seho

ldSize

−0.364

20.42

91.516

40.09

581.08

29.929

70.21

000.82

72.800

2−0

.232

00.72

34.749

1

Age

under

191.84

932.07

23.500

1−1

0.47

205.12

16.043

8∗∗

−6.137

63.91

23.120

2−6

.884

23.42

14.047

2∗∗

Age

20-44

0.16

671.29

01.922

4−1

0.20

553.14

01.001

6∗∗

−6.633

42.39

87.006

9∗∗

−6.363

820

.976

9.002

1∗∗

Age

45-64

−2.739

42.71

06.350

4−2

1.97

326.59

07.001

2∗∗

−12.14

335.03

46.017

9∗∗

−16.83

244.40

28.000

2∗∗

Une

mploym

ent

−0.039

80.05

91.469

50.11

260.14

84.450

00.09

670.11

34.396

10.05

780.09

91.564

9

Distanc

e−0

.002

00.00

19.250

40.00

050.00

48.924

6−0

.000

30.00

36.927

50.00

070.00

32.815

6

RetailEst.

0.06

400.00

55<

.000

1∗∗

Intercep

t8.74

342.88

37.003

2∗∗

9.73

507.20

49.180

08.03

525.50

37.147

84.44

374.81

31.358

3

R2

0.81

950.45

110.40

820.42

10

∗ Significan

tatthe

P=.10level.

∗∗Sign

ificant

attheP

=.05level.



American Industrial Classification System codes. Further-more, while this study has demonstrated the impact thatCAHs have on total retail activity, more in-depth researchcould explore what types of “niche” markets complementCAHs and thus make good additions to those communi-ties with a hospital. Finally, our analysis is limited to thesingle state of Oklahoma and utilizes a relatively smallsample size, and future efforts may attempt to replicateour findings on a regional or national scale.

A few important policy implications arise from this re-search. It is important for a CAH to stay open in a ruralcommunity for health care reasons. However, other pos-itive externalities that occur from the presence of a CAHare often overlooked. This research concludes that townswith a CAH have statistically higher levels of retail salesand more retail establishments (including those with lessthan 20 employees). These results suggest that a CAH notonly attracts patients, but it also attracts shoppers whomake purchases at local retail establishments. From a pol-icy perspective, subsidizing CAHs is beneficial to the localretail sector as well as the more commonly recognizedhealth aspect.

An analysis of exactly how much support is too muchfor a local CAH is beyond the scope of this paper. How-ever, our research demonstrates that the presence of aCAH increases retail sales and boosts the number of re-tail establishments in rural communities. Providing ade-quate levels of funding support for CAHs can therefore beviewed as a component of promoting economic activity inthe retail sector. Previous research indicates that educa-tion, engagement, and awareness of the local health sec-tor can all increase community support, including higherutilization of local facilities and continued local financialsupport.28 Higher-level financial support, such as that forthe CAH program itself, requires assessment of the pro-gram impact on an aggregate scale. This paper does, how-ever, add to the body of evidence on CAH benefits, and itmakes an economic argument for continued aid to ruralhospitals.

References

1. American Hospital Association. Beyond Health Care: The

Economic Contribution of Hospitals. Washington, DC:

Trendwatch; April 2008.

2. Department of Health and Human Services, Centers for

Medicare and Medicaid Services, National health

expenditure data page. Available at: www.cms.hhs.gov/

NationalHealthExpendData./ Accessed September 25,

2009.

3. Doeksen GA, Johnson T, Biard-Holmes D, Schott V. A

healthy health sector is crucial for community economic

development. J Rural Health. 1998;14(1):66-72.

4. Avery S. Rural hospitals’ contributions to health care and

local economies. CHA Special Report. Sacramento, CA:

California Healthcare Association; July 2002.

5. Novack NL. Bridging the gap in rural healthcare. Main

Street Econ. Federal Reserve Bank of Kansas City.

September 2003.

6. Doeksen GA, Cordes S, Shaffer R. Health Care’s

Contribution to Rural Economic Development. Rockville, MD:

Office of Rural Health Policy; 1992.

7. Zimmerman MK, McAdams R. Public support for rural

health care: federal programs and local subsidies. In:

Kronenfield JJ, ed. Chronic Care, Health Care Systems and

Services Integration: Research in the Sociology of Health Care.

Vol. 22. New York, NY: Elsevier; 2004:22-45.

8. Morrisey MA, Wedig GJ, Hassan M. Do nonprofit

hospitals pay their way? Health Aff . 1996;15(4):

132-144.

9. Slifkin R. State laws and programs that affect rural health

delivery. In: Ricketts TC III, ed. Rural Health in the United

States. New York, NY: Oxford Press; 1999:84-94.

10. Doeksen GA, Schott V. Economic importance of the

health-care sector in a rural community. Rural Remote

Health. 2003; 3(online):135. Available at:

www.rrh.org.au/articles/subviewnew.asp?ArticleID=135.

Accessed October 23, 2009.

11. Brooks L, Whitacre B, Ralstin S, Woodruff A, Kaiser C,

Schott V. The Economic Impact of the Health Sector on the

Okeene Municipal Hospital Medical Service Area. Stillwater,

OK: Oklahoma Cooperative Extension Service; January

2009. Available at: www.okruralhealthworks.org/

PDFWEB/AE-09005.pdf. Accessed October 4, 2009.

12. Office of Rural Health and Primary Care. Economic Impact

of the Johnson Memorial Health Services and Related Health

Sectors of Lac qui Parle County. St Paul, MN: ORHPC; June

2008.Available at: www.health.state.mn.us/divs/orhpc/

flex/studies/rhwlacquiparle.pdf. Accessed October 4,

2009.

13. Kentucky Rural Health Works. The Economic impact of the

local health care system on the Casey County economy.

Lexington, KY: KRHW, Kentucky Cooperative Extension

Service; February 2003. Available at: www.ca.uky.edu/

KRHW/impact.html. Accessed October 2009.

14. Olson D, Lindall S. IMPLAN Professional Software, Analysis,

and Data Guide. Stillwater, MN: Minnesota IMPLAN

Group; 2004.

15. US Census Bureau. State and local government finance.

Available at: http://www.census.gov/govs/estimate/

index.html. Accessed November 18, 2009.

16. Probst JC, Samuels ME, Hussey JR, Berry DE, Ricketts,

TC. Economic impact of hospital closure on small rural

counties, 1984 to 1988: demonstration of a comparative

analysis approach. J Rural Health. 1999;15(4):375-390.

17. Lawler MK, Doeksen GA, Schott V. Impact of conversion

to critical access hospital status for Oklahoma’s rural

hospitals. J Rural Health. 2003;19(2):135-138.



18. Oklahoma State Department of Health. Medical facilities

directory. Oklahoma City, OK: Oklahoma State

Department of Health; January 2010. Available at:

www.ok.gov/health/documents/MF%20MedicalFacilities.

Directory.eff.01.11.10.pdf. Accessed February 12, 2010.

19. Ferber R. Variations in retail sales between cities.

J Marketing. 1958;22(3):295-303.

20. Muske G, Woods M. Micro businesses as an economic

development tool: what they bring and what they need.

J Comm Dev Society. 2004;35(1):97-116.

21. Goetz SJ, Rupasingha A. Wal-Mart and social capital. Am

J Agri Econ. 2006;88(5):1304-1310.

22. Artz GM, Stone KE. Analyzing the impact of Wal-Mart

Supercenters on local food store sales. Am J Agri Econ.

2006;88(5):1296-1303.

23. Irwin E, Clark J. The local costs and benefits of Wal-Mart.

Department of Agricultural, Environmental and

Developmental Economics. Ohio State University.

February 2006. Available at: www-agecon.ag.ohio-

state.edu/programs/ComRegEcon/walmart/Irwin

Clark-WalMart Final 03 06.pdf. Accessed August 7,

2009.

24. Anscombe F. The transformation of Poisson, binomial

and negative-binomial data. Biometrika. 1948;35:

246-254.

25. Bartlett M. The square root transformation in analysis of

variance. Suppl J Royal Stat Soc. 1936;3:68-78.

26. Gregiore T, Lin Q, Boudreau J, Nelson N. Regression

estimation following the square-root transformation of

the response. Forest Sci. 2008;54(6):597-606.

27. Halvorsen R, Palmquist R. The interpretation of dummy

variables in semilogarithmic equations. Am Econ Rev.

1980;70(3):474-475.

28. Zimmerman MK, McAdams R, Halpert, BP. Funding

health services in the rural United States: federal policies

and local solutions. In: Glasgow N, Morton LW, Johnson

NE, eds. Critical Issues in Rural Health. Ames, IA: Blackwell;

2004:211-224.


Critical Access Hospitals and Retail Activity: An Empirical Analysis in Oklahoma

Documents

Transcript of Critical Access Hospitals and Retail Activity: An Empirical Analysis in Oklahoma