Multiple measures of loyalty: validity and reliability tests in two retail settings.

Pleshko, Larry P.

ABSTRACT

The author employs multiple indicators to measure retailer brand loyalty in consumers of fast-food burger outlets and health clubs. Validity tests and reliability tests are performed. The authors find that the six measures are coherent regarding construct validity, exhibiting high relationships to each other in two dimensions: preferences and usage. The findings are very similar in both retail settings, providing evidence for external validity. Additionally, regarding reliability, support is provided that the different measures can be considered equivalent alternative forms for which to measure loyalty. Also a selected loyalty indicator exhibits criterion validity, being related to other variables as might be predicted. The use of these multiple loyalty indicators should provide more confidence in the outcomes of future studies pertaining to relationships with brand loyalty.

INTRODUCTION

General business wisdom suggests a company focus at least a proportion of marketing efforts on the development, maintenance, or enhancement of customer loyalty (Dick & Basu 1994). This emphasis is important because a company with a large number of brand loyal buyers will be more secure in its markets and should have a higher market share than

other firms without this asset (Raj 1985; Robinson 1979; Smith & Basu 2002). Having more brand loyal buyers than competitors has many advantages, including a greater response to advertising (Raj 1982), larger purchase quantities per occasion (Tellis 1988), and reduced marketing costs (Rosenberg & Czepial 1983). Brand loyal buyers are important because, as markets become more mature, increases in share become more expensive and improving customer loyalty is a means of increasing and maintaining share (Gounaris & Stathakopoulos 2004).

The purpose of this study is twofold. First, the authors offer and describe six parallel forms of measurement for store loyalty. Second, the validity and reliability of these loyalty indicators are tested in samples of consumers who use fast-food burger outlets and health clubs. In particular, the study addresses concept validity, construct validity, criterion validity, external validity, and parallel forms reliability. Ideally, multiple indicators should provide more reliable and valid measures and, therefore, more confidence in the conclusions drawn from any statistical analysis.

LOYALTY

Firms must find ways to keep current customers, attract new customers, and to retain these buyers over the long term. A firm must therefore continuously battle with competitors to maintain or increase both the number of buyers and the loyalty of these customers. When a firm fails to hold a strong relative competitive position, it runs the risk of a widespread phenomenon called "Double Jeopardy" (DJ), where small-share brands attract somewhat fewer loyal consumers buying in smaller quantities than large-share brands which have a larger loyalty base purchasing larger quantities (Ehrenberg & Goodhardt 2002; Donthu 1994; Martin 1973; McPhee 1963; Michael & Smith 1999). Therefore, brand loyalty is a critical strategic issue since it directly affects firm performance and customer behaviors.

McDowell and Dick (2001) postulate that a brand's performance is driven by both the number of individuals buying a particular brand and the frequency of repeat purchases from these customers. A firm's ability to manage these two factors dictates the extent to which it maintains and sustains its customer base, as well as its market share. Indeed, Lehmann and Winner (1997) suggest that cultivating repeat business is a prerequisite for maintaining a firm's market share. Likewise, Robinson (1979) and Raj (1985) state that the larger the number of loyal customers, the more secure will be the brand's market share. With estimates regarding the number of 'truly' brand loyal buyers for certain retailers hovering around twenty-five percent, one can begin to imagine the importance of the battles for the consumer that we witness every day in consumer markets (Pleshko & Heiens 1996, 1997).

Many definitions of loyalty exist. But, it is generally described as the propensity of a buyer to purchase the same brand repeatedly. Faithfulness, consistency, and a lack of switching all might be useful in defining loyalty (c.f. Etzel et al 2004). Dick & Basu (1994) point out that loyalty is not only a behavioral phenomenon, but also an attitudinal phenomenon with preferences being important in the conceptual definition. Combining behaviors and attitudes results in four categories of loyalty: truly loyal buyers, latently loyal buyers, spuriously loyal buyers, and buyers with no loyalty towards a brand. True loyalty exists when buyers make high percentages of purchases from the preferred brand. No loyalty indicates no preferences and little or no purchasing from a brand. Spurious loyalty indicates a buyer heavily purchasing from a brand which is not the favorite. Finally, Latent loyalty is evident when a buyer has a favorite but does not purchase that brand very often. Thus, spurious and latent loyalty might be opposite sides of the same coin.

Additionally, there are other relevant aspects of loyalty. Brand insistence and brand preference might be considered varying degrees of true loyalty (c.f. Murphy & Enis 1986). Insistence is when a buyer has a favorite and will not purchase another brand, even if the favorite is unavailable, as is the case sometimes with automobiles or music or health clubs. Brand preference is when a buyer has a favorite but will switch if the favorite is unavailable, as is the case sometimes with soda or fast-food outlets. Brand avoidance occurs when a buyer goes out of the way not to purchase a specific brand, maybe to avoid an inferior product or a socially unacceptable brand

Two problem areas in studying loyalty are the difficulty with measurement and the lack of definitional clarity for brand loyalty itself (Badinger & Rubinson 1997). Single indicators of loyalty are used in most studies or only a single dimension is investigated, leaving questions of reliability and validity. The Dick and Basu (1994) scheme is useful, but there are obstacles associated with implementing this typology due to voluminous data collection requirements (Pleshko & Heiens 1996, 1997). Therefore, until more distinct definitions are employed and better measurement is utilized, studies involving loyalty will have reliability and validity concerns. A more specific focus on the definition and measurement of loyalty should improve future studies using this concept.

DATA COLLECTION

The data for the current study are gathered in a large university town in the southeastern USA. The sampling frame is comprised of undergraduate business students, a group of consumers who are frequent users of fast-food hamburger outlets or health clubs, the two retailer types which are the focus point for data collection in this study. Information is accepted only from users of the retailer type, thus eliminating nonusers from the study. The data are from self-administered questionnaires collected from three variations of the same research instrument for each type of retailer. The purpose of the instrument variations is to minimize any ordering effects in the collection of the data. Six classes, three for health clubs and three for fast-food outlets, are selected for inclusion in the study from the offering at the university by random selection of core course offerings in the business school. This process results in ninety-eight usable responses for fast-food outlets and eighty-one useable responses for health clubs. As the nonusers were screened during the sampling process and all those selected responded, few unusable surveys were found.

Multiple brands are included in the study for each type of retailer. For the fast-food retail outlets, there are six chains selected for inclusion in the study. Each of these fast-food retail brands has multiple outlets in the accessible geographical market area of the sampling frame. Also, each of the retailer brands has market shares above five percent. For the health clubs, there are sixteen clubs identified in the relevant geographical market area. Eleven of these clubs are eliminated from study after data collection due to small numbers of users or small market shares. This resulted in five health clubs with market shares above five percent that are included in this study.

PROPOSED INDICATORS OF LOYALTY AND MEASUREMENT

Validity and reliability concerns must be considered when collecting data for any construct. Regarding validity, researchers must derive accurate measures of the concept under study. This is normally addressed in the initial stages by considering content validity and construct validity (Cooper & Schindler 2001). To show content validity, the loyalty measures should cover the range and dimensions considered relevant to the construct. Therefore, based on the extent literature for loyalty, both attitudes and behaviors should be included and also a specific type of loyalty outlined, if desired. The author proposes the use of five behavioral indicators (#1,#3,#4,#5,#6) and one attitudinal measure (#2). The six indicators are [1] Loyalty-% of total use for each store (L%OFTOT), [2] Loyalty-% regarding self report category (LSLFR%), [3] Loyalty-% use of outlet (LUSED%), [4] Loyalty-% most-used store (LMOST%), [5] Loyalty-% 2nd most used store (L2MST%), and [6] Loyalty-% last purchase (LLAST%). These are described in paragraphs to follow.

Note that the indicators represent two dimensions related to loyalty: (a) a behavioral dimension, called usage and (b) an attitudinal dimension, called preference. The single indicator related to preference should be expanded in future studies. The inclusion of a preference ranking or rating question(s) would give us multiple indicators of the preference dimension. Thus, two additional indicators are recommended: [7] Loyalty-% most preferred and [8] Loyalty-% 2nd most preferred. However, these two indictors are not included in the study, but should be included in future studies.

The overall indicators for the six loyalty items included in the study are derived by summing across multiple variable components and across respondents to arrive at a sample total for each indicator, which are all percentages. Table 1 and Table 2 show the specific numbers for each of the constructs in the two retail settings. The definitions and measures for each of the constructs are identical in the two retailer types and are explained in the following paragraphs.

Loyalty-% of total use each store (L%OFTOT) is defined as the percentage out of his total times the respondent uses each brand, if they are users of that brand. Respondents are asked to write the number of 'times' they use each brand over a specified period of time. Note that the buyer may not use brand A most often, but the information for brand A is still included. It is calculated for each respondent for each brand used and then the overall sample average percentage is calculated. Thus, for respondent X who uses brands A and B: L%A = (times A)/(times A + times B ... + times N) and L%B = (times B)/(times A + times B ... + times N).

Loyalty-self report category (LSLFR%) is defined as the percentage of respondents classified as loyal to a particular retail brand out of the total users of that brand. First, respondents are assigned to the brand which they use the most often. This is derived from the 'times' question as with L%OFTOT. Additionally, a two-item interval scale is used to determine if the respondent considers himself to be loyal and/or have a favorite brand. From the scale, respondents are assigned to a high loyalty category based on high scores on the scale. Thus, respondents are classified as loyal to the brand they use most often only if they also score high on the scale. Otherwise, they are classified as not loyal. This indicator might be closest to Dick and Basu's (1994) typology.

Loyalty-% use of outlet (LUSED%) is defined as the percentage of respondents who use each brand. This information is also derived from the 'times' question, as with L%OFTOT, however only use/not use data is taken. It is calculated for each brand as the total users of the brand in relation to the total sample. Thus, LuseA = (sum of users of A)/(total sample size). It is included because we use brands which we prefer.

Loyalty-% of most-used store (LMOST%) is defined as the percentage of times that each brand is used as the primary brand in the category for a respondent. This is also derived from the 'times' question. A specific brand is the most-used for a respondent when the largest number of 'times' is indicated. Thus, for respondent X, if times A>times B, times C ... , times N, then brand A is assigned to respondent X as most used store. The overall indicator is a summation for each brand of those respondents most using each store. This is included because it is logical for users to buy most often from their favorite brand.

Loyalty-% 2nd most used store (L2MST%) is defined similarly to LMOST%, except we are interested in the brand with the second highest 'times used'. This is included as it is possible for buyers to be loyal to more than one store.

Loyalty-% last purchase (LLAST%) is defined as the brand from which the respondent last purchased. Respondents are asked to check a box next to the brands. This is included because good predictors of future behaviors are recent past behaviors. The indicator is a percentage calculated by summing, for each brand, all the respondents who indicated that they bought from a specific brand last. Then, this summation is divided by the total respondents to get the indicator. Thus, LLast%A = (total purchased A last)/ (total respondents).

ANALYSIS

To address construct validity, factor analysis with reliability is normally used or, alternatively, inter-indicator correlations are analyzed. Since the indicators are sample summaries, and not specific to individual respondents, it is not possible to test for validity and reliability using factor analysis and Pearson correlations. The authors use Spearman's rho, the rank-order test to investigate construct validity and reliability. The six 'parallel forms' of the same construct should ideally provide similar indications of loyalty towards each of the specific retail brands. This procedure will provide an estimate of reliability referred to as equivalent measures or parallel forms. In other words, the six loyalty indicators should result in a rank for each brand, regarding loyalty, in nearly the same order. Thus, the Spearman rank-order test is appropriate to investigate reliability in this study.

Analysis of the rho statistics, similar to inter-item correlations, will also provide evidence for any underlying dimensionality and construct validity. Items with significantly related brand loyalty rankings might be members of the same underlying dimension or factor. Alternatively, items without significantly related brand loyalty rankings would be considered to be either from different underlying factors or maybe one of the items is not related to an underlying dimension. Thus, by looking at the groupings of related loyalty indicators, it should be possible to determine the underlying structure.

Spearman's (1904) test statistic, rho or "r", is calculated with data taken from 'n' pairs (Xi,Yi) of observations from the respondents on the same objects, the fast-food burger outlets or the health clubs. In this study, each Xi and Yi pair would be rankings related to combinations of the loyalty indicators. Refer to Table 1 again to see the rankings of each indicator from highest "6" to lowest "1" for fast-food outlets. The rankings, not the indicator values themselves, are used to calculate the statistic. The r-values and the p-values for each statistical comparison are shown in Table 3 for fast-food outlets and Table 4 for health clubs.

As an example in the fast-food outlets from Table 1 comparing two loyalty indicators--L%OFTOT and LSLFR%, the ranking of brand #3 on each of these indicators is L%OFTOT= 1 (lowest) and LSLFR%= 3 (middle). The reader can find the remaining pairs of rankings for L%OFTOT versus LSLFR%. These associated "rankings pairs" are used to calculate the test statistic. To calculate the "r", the "rankings pairs" are input into the following formula.

R = 1 - 6 [SIGMA][d.sup.2]/n([n.sup.2] - 1)

where 'n' equals the number of paired rankings, in this case six for all the calculations related to fast-food outlets and five for calculations related to health clubs. 'd' equals the absolute difference between the rankings for each outlet: (Xi-Yi). The test statistic ranges between +1 (perfect positive association) and -1 (perfect negative association). In this study, two-tailed tests are performed, giving the general hypotheses for the paired variables: Ho: independently ranked pairs or Ha: related ranked pairs. Note that due to the power of the Spearman test and the small number of comparisons, that large test statistics are not always statistically significant. Therefore, the study chooses to use p=.10 as the cut-off point for significance.

The calculation of the statistic of association, "r" follows for our example. From Table 3, we see that r=.829 indicating a moderately significant relationship (p<.10) between the order of rankings for the two loyalty indicators, L%OFTOT and LSLFR%.

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.]

RESULTS

When evaluating the reliability of the indicators, Table 3 shows that ten out of fifteen possible pairings are statistically significant for fast-food outlets. The remaining pairings all show relatively high r-values, even if not statistically significant. The average rho value is .842 across the fifteen items and above .900 for those that are significant. The large number of significant comparisons between the loyalty indicators is evidence for parallel forms reliability. Thus, support is provided that the different loyalty indicators each provide similar and, therefore, reliable measures of loyalty.

Table 4 shows that seven out of fifteen possible pairings are statistically significant for health clubs. The remaining pairings for health clubs are rather closer to zero. This might suggest that two distinct underlying dimensions are evident. The average rho value is .583 across the fifteen items and above .900 for those that are significant. Again, the moderately large number of significant comparisons between the loyalty indicators is evidence for parallel forms reliability. Thus, support is provided that the different loyalty indicators each provide similar and, therefore, reliable measures of loyalty.

When evaluating construct validity, it is necessary to determine if the indicators adhere to the conceptual definitions derived from the literature review. This study conceptualized loyalty as a two dimensional construct comprised of attitudes (preference) and behaviors (usage). Therefore, it might be expected that the indicators will all group in one of three ways. First, they all group together, exhibiting a single overall loyalty dimension. Second, they may split into two groups as suggested earlier, each characteristic of usage or preference. A third possibility is that there will be more than two underlying dimensions, thus not providing support for the validity of our construct as defined.

An analysis of Table 3 reveals the possible groupings of loyalty indicators for fast-food outlets. This should help provide us with an understanding of the underlying dimensionality and provide evidence of construct validity. As can be seen from the table, there are really two groups evident and are divided as expected. Group 1 refers to a preference dimension and includes L%OFTOT + LSLFR%. Group 2 refers to a usage dimension and includes L%OFTOT + LUSED% + LMOST% +L2MST% + LLAST%. As noted, L%OFTOT seems to capture both dimensions and might be useful as an overall indicator of loyalty by itself, since it appears to correlate highly with the usage group and the preference group. Note that the single true attitudinal measure does not correlate highly with any of the items in the other group, thus providing evidence for convergence and discrimination within and among the two dimensions--important for construct validity. Therefore, evidence is provided for construct validity in fast-food outlets through the as expected groupings of indicators and also with ample evidence for discrimination and convergence.

An analysis of Table 4 shows the possible groupings of loyalty indicators for health clubs.

As noted in Table 4, there are really two groups evident and are divided as expected. Group 1 refers to a preference dimension and includes L%OFTOT + LSLFR%. Group 2 refers to a usage dimension and includes LUSED% + LMOST% +L2MST% + LLAST%. Note that there is even more evidence of convergence and discrimination as we might expect in the health clubs than the fast-food outlets, therefore providing evidence for construct validity. Also, it appears that the loyalty indicators are more representative of the underlying dimensions with health clubs than with fast-food outlets, where a single overall dimension could be an alternative to the conceptually defined two dimensions.

To determine if the variables are useful, they must show a predicted relationship to other known constructs--referred to as criterion validity if the data is gathered in a single timeframe, as in this study. The literature claims an explanation for developing brand loyalty is to reduce risk by finding and sticking with brands which are reliable (c.f. Deering & Jacoby 1972, Derbaix 1983). From this explanation we would expect loyal buyers to have used a lesser number of brands than other buyers and, additionally, we would expect loyal buyers to consider fewer brands than other buyers (c.f. Eroglu et al 1983, Brown & Wildt 1987, Stewart & Punj 1982). A single loyalty indicator is selected, the L%OFTOT, to represent the six indicators in the study since L%OFTOT is really the only indicator available at the respondent-level. For fast-food outlets, L%OFTOT is significantly negatively correlated with both the number of brands used (p=.000) and the size of the consider-set (p=.044) of the respondents. For health clubs, L%OFTOT is also significantly negatively correlated with both the number of brands used (p=.000) and the size of the consider-set (p=.015) of the respondents. Therefore, the selected loyalty indicator exhibits criterion validity by correlating with other known constructs in the expected manner.

The study also addresses external validity through the repetition of the analysis in two retailer types, fast-food outlets and health clubs. Although the statistics are not identical in the comparison, the results are basically similar. Both reveal reliability through parallel forms testing. Also, using identical measures eliminates many alternative explanations in the repeated data collection. Additionally, the evident groupings of the loyalty variables are found, in both studies, to be similar to the predicted dimensionality, revealing a preference dimension and a usage dimension. In addition, evidence is provided for criterion validity in both retail settings, with the L%OFTOT indicator highly correlated with two related constructs in a manner that we would expect.

DISCUSSION/IMPLICATIONS

The outline and testing of multiple indicators for loyalty is the primary purpose of the study. The analysis revealed that loyalty is comprised of two underlying dimensions: a preference or attitude factor and a usage or behavioral factor. The analysis provided support for the indicators as reliable equivalent forms to be used in measuring loyalty. Additionally, the analysis provided support for construct validity by showing convergence-divergence and a grouping of the variables as predicted by theory. Also, a single indicator Loyalty-% of total use (L%OFTOT) is shown to correlate with other related constructs in an expected manner, thus providing evidence for criterion validity. Finally, the study provides evidence for external validity with similar results in two separate retailer types.

With the provision and testing of reliable and valid multiple indicators, future studies which include loyalty will have a choice of indicators to use in the analysis. If an investigator wants to use a single indicator, due to space or time limitations, then L%OFTOT may be acceptable since it captures both dimensions in some settings. Otherwise, any of the other indicators can be applied, depending on the preferences of the researchers. Note, however, that the indicators in the study do not specifically address brand insistence or brand preference or brand avoidance.

While the findings are interesting, readers are cautioned about applying the results to other areas. The weak power of the test statistic actually may serve to increase our confidence in the conclusions drawn from the study. However, other complications may arise with a stronger test. As usual, one should wonder whether the findings will also be evident in other retailers, other samples, and other markets. Plus, more study is necessary to evaluate additional indicators of the preference dimension of loyalty. The study provides two possible indicators for this dimension, but did not test them for validity and reliability.

REFERENCES

Badinger, A.L. & J. Rubinson (1997). The jeopardy in double jeopardy. Journal of Advertising Research, 37(6), 37-50.

Brown, J.J. & A.R. Wildt (1987). Factors influencing evoked set size. Working Paper Series, University of Columbia at Missouri

Cooper, D.R. & P. Schindler (2001). Business research methods (7th Edition). New York, NY: Mcgraw-Hill Publishing.

Derbaix, C. (1983). Perceived risk and risk relievers: An empirical investigation. Journal of Economic Psychology, 3, 19-38.

Deering, B.J. & J. Jacoby (1972). Risk enhancement and risk reduction as strategies for handling perceived risk.

Proceedings of the 3rd Annual Convention of the Association for Consumer Research, M. Venkatesan, ed., Chicago: ACR, 404-416

Dick, A. & K. Basu (1994). Customer loyalty: Toward an integrated framework. Journal of the Academy of Marketing Science, 22(Spring), 99-113.

Donthu, N. (1994). Double jeopardy in television program choice. Journal of the Academy of Marketing Science, 22(2), 180-186.

Ehrenberg, A. & G.J. Goodhardt (2002). Double jeopardy revisited, again. Marketing Research, 14(1), 40-42.

Eroglu, S.A., G.S. Omura & K.A. Machleit (1983). Evoked set size and temporal proximity to purchase. Proceedings of the AMA Educators' Conference, P. Murphy, ed., Chicago: American Marketing Association, 97-101.

Etzel, M.J., B.J. Walker & W.J. Stanton (2004). Marketing (13th Edition). New York, NY: McGraw-Hill/Irwin Publishing.

Gounaris, S. & V. Stathakopoulos (2004). Antecedents and consequences of brand loyalty: An empirical study. Journal of Brand Management, 11(4) 283-307.

Lehmann, D.R. & R.S. Winner (1997). Analysis for marketing planning (4th edition). Burr Ridge, IL: Irwin.

Martin Jr, C.R. (1973). The theory of double jeopardy. Journal of the Academy of Marketing Science, 1(2), 148-153.

McDowell, W.S. & S.J. Dick (2001). Using TV daypart 'double jeopardy effects' to boost advertising efficiency. Journal of Advertising Research, 41(6), 43-51.

McPhee, W.N. (1963). Formal theories of mass behavior. New York: The Free Press

Michael, J.H. & P.M. Smith (1999). The theory of double jeopardy: An example from a forest products industry. Forest Products Journal, 49(3), 21-26.

Murphy, P.E. & B.M. Enis (1986). Classifying products strategically. Journal of Marketing, 50(July), 24-42.

Pleshko, L.P. & R.A. Heiens (1997). An empirical examination of patient loyalty: An application of the customer loyalty classification framework to the healthcare industry. Journal of Customer Service in Marketing and Management, 11(2), 105-114

Pleshko, L.P. & R.A. Heiens (1996). Categories of customer loyalty: An application of the customer loyalty classification framework in the fast-food hamburger market. Journal of Food Products Marketing, 3(1), 1-12

Raj, S.P. (1985). Striking a balance between brand popularity and brand loyalty. Journal of Marketing, 49(Winter), 53-59.

Robinson, J. (1979). Brand strength means more than market share. Journal of Advertising Research, 19(October), 5, S.83-87.

Rosenberg, L. & J. Czepial (1983). A marketing approach to customer retention. Journal of Consumer Marketing, 2, 45-51.

Smith, T. & K. Basu (2002). A view from the top: The impact of market share dominance on competitive position. Journal of Brand Management, 10(1), 19-33.

Spearman, C.E. (1904). The proof and measure of association between two things. American Journal of Psychology, 15, 72-101.

Stewart, D.W. & G. Punj (1982). Factors associated with changes in evoked set among purchasers of automobiles.

Proceedings of the AMA Educators' Conference, B.J. Walker et al, eds., Chicago: American Marketing Association, 61-65.

Tellis, G. (1988). Advertising exposure, loyalty, and brand purchase: A two-stage model of choice. Journal of Marketing Research, 25(May), 134-144.

Larry P. Pleshko, Kuwait University

Table 1: Fast Food Burger Outlets: Loyalty Statistics and Rankings

Items/Brands #3 #2 #6 #1 #5 #4

L%OFTOT 19% 26% 28% 28% 30% 42%
Rank 1 2 4 3 5 6
LSLFR% 9% 7% 9% 7% 9% 17%
Rank 3 1 4 2 5 6
LUSED% 23% 31% 56% 70% 76% 84%
Rank 1 2 3 4 5 6
LMOST% 2% 7% 16% 22% 16% 38%
Rank 1 2 3.5 5 3.5 6
L2MST% 4% 9% 10% 21% 29% 28%
Rank 1 2 3 4 6 5
LLAST% 3% 7% 8% 16% 25% 41%
Rank 1 2 3 4 5 6

Table 2: Health Clubs: Loyalty Statistics and Rankings

Items/club #2 #9 #3 #10 #13

L%OFTOT 100% 86% 70% 75% 89%
Rank 5 3 1 2 4
LSLFR% 67% 43% 22% 58% 8%
Rank 5 2 1 3 4
LUSED% 4% 9% 11% 12% 44%
Rank 1 2 3 4 5
LMOST% 4% 7% 7% 10% 41%
Rank 1 2.5 2.5 4 5
L2MST% 0% 1% 3% 3% 5%
Rank 1 2 3.5 3.5 5
LLAST% 4% 8% 14% 18% 47%
Rank 1 2 3 4 5

Table 3: Fast-food Burger Outlets: Spearman Rho Statistics

 LSLFR% LUSED% LMOST%

L%OFTOT .829 (<.10) .943 (<.02) .857 (<.10)
LSLFR% -- .714 (n.s.) .571 (n.s.)
LUSED% -- .914 (<.05)
LMOST% --
L2MST%

 L2MST% LLAST%

L%OFTOT .886 (<.05) .943 (<.02)
LSLFR% .657 (n.s.) .714 (n.s.)
LUSED% .943 (<.02) 1.00 (<.01)
LMOST% .800 (n.s.) .914 (<.05)
L2MST% -- 943 (<.02)

Table 4: Health Clubs: Spearman Rho Statistics

 LSLFR% LUSED% LMOST%

L%OFTOT .900 (<.10) -.300 (n.s.) -.250 (n.s.)
LSLFR% -- -.100 (n.s.) -.150 (n.s.)
LUSED% -- .950 (<.10)
LMOST% --
L2MST%

 L2MST% LLAST%

L%OFTOT -.550 (n.s.) -.300 (n.s.)
LSLFR% -.350 (n.s.) .150 (n.s.)
LUSED% .950 (<.10) 1.00 (<.01)
LMOST% .900 (<.10) .950 (<.10)
L2MST% -- .950 (<.10)