A cluster analysis of petrol profit margins across various regional and urban locations in Australia.

Valadkhani, Abbas ; Chen, George ; Anderson, John 等

1. INTRODUCTION

The present study is one of the few attempts to examine the spread between the retail and wholesale prices using disaggregate data as most of the previous empirical studies on petrol prices in Australia utilised mainly national and/or state level data (see inter alia Donnelly, 1982; Newman and Kenworthy, 1989; Samimi, 1995; Fatai et al., 2004; Dodson and Sipe, 2007; Hensher and Stanley, 2009; Li et al., 2010; Gargett, 2010). Using a unique database, this study provides a cross-sectional comparison of retail profit margins using a hierarchical cluster analysis to determine whether or not the observed large variations in petrol margins can be described as 'excessive'. As an example, the average difference between retail and wholesale prices of petrol varies from 24.1 cents per litre in Tennant Creek to only 2.4 cents per litre in Mackay during the period October 2007-January 2012. This paper seeks to identify whether such large profit spreads can be explained by the extent of economies of scale and scope, the market size and the associated overhead costs. In other words, is the spread between the retail and wholesale prices high mainly in rural, less-populated and remote areas or is it a ubiquitous phenomenon everywhere?

In order to examine these issues, a cluster analysis is conducted to classify all retail locations into various groups each exhibiting similar magnitudes of gross profit spread and a set of control variables. Due to the unavailability of the data for various petrol stations within each of the 109 retail locations, we use two proxy variables to capture the effects of transport cost and the extent of economies of scale and scope, namely population and the distance between retailers and wholesalers. The major objective of this paper is to identify several clusters in which gross profit margins are relatively comparable. This allows us to analyse whether large existing differences in margins across various locations can be justified against the factors associated with location, cost and the size of the market.

The findings presented here not only contribute significantly to our understanding of substantial regional differences in petrol margins, but also have direct practical implications for motorists and provide a guideline for relevant regulatory authorities across various geographical locations. Instead of the prevailing use of aggregate data, our disaggregated study can lead to an in-depth understanding of competition, or lack thereof, and the extent of profiteering in the petrol market. The location-specific results of this study make price monitoring by regulatory bodies more cost effective as they can readily identify and target price setting in those locations pursuing comparatively higher margins. Hence, this study assists both motorists and regulatory bodies to make more informed and objective assessments of retail profit margins within the identified homogeneous clusters, leading to greater efficiency and transparency of the petrol market.

The rest of the paper is structured as follows. Section 2 presents a succinct review of literature on this topic. Section 3 briefly discusses how we conduct our cluster analysis at a disaggregate level for benchmarking purposes. Section 4 describes the sources and summary statistics of the data employed. Section 5 presents the results of our cluster analysis using 109 cross-sectional observations within a trivariate system. Section 6 discusses the policy implications of the study followed by some concluding remarks in Section 7.

2. REVIEW OF LITERATURE

The fuel consumption-economic activity nexus has been examined extensively in the literature since the 1973 oil embargo. According to Hamilton (1983), fuel price hikes were responsible for the majority of post-war U.S. recessions until the mid-1970s. However, Hamilton's study only focused on periods when the economy was subject to significant upward fuel price movements and controls. Hooker (1996) pointed out that such an inverse relationship between fuel prices and economic activity was weakened after 1985 when falling fuel prices failed to stimulate the U.S. economy as anticipated, a view not shared in Hamilton's (1996) subsequent study.

Inter alia Mork (1989) and Ferderer (1996) confirmed Hamilton's initial finding after incorporating the effect of fuel price volatility on the US economy. Brown and Yucel (2002, p.194) analogously summed up this inverse relationship by suggesting that "the classic supply-side effect best explains why rising oil prices slows GDP growth and stimulates inflation." In a recent study, Ford (2011) concludes that vertically integrated oil companies' gross profit margins are actually highest during the time of moderate petrol prices. Contrary to popular belief, these companies' gross profit margins decline when petrol prices are excessively high, and even more so than when these prices are low.

The adverse effect of fuel prices on economic activity has been well documented in many countries around the world. For example, Asafu-Adjaye (2000) finds that a change in fuel consumption Granger causes changes in income in several Asian developing countries. In a follow-up study, Mahadevan and Asafu-Adjaye (2007) broaden the investigation to include a total of 20 developed and developing countries and find that fuel consumption stimulates short-run economic activity. Chontanawat et al. (2008) is one of the studies that provides comprehensive evidence for fuel consumption as a determinant of economic growth. They examined the effect of fuel consumption in 100 countries and concluded that rising fuel costs and reductions in fuel consumption can jeopardise economic growth in the developed countries more than that of the developing countries.

There are a number of studies in the literature which have highlighted the significance of regional differences in the context of petrol prices (see inter alia Hastings and Gilbert, 2005; Simpson and Taylor, 2008; Hosken, et al. 2008; Eckert, 2013; Pennerstorfer and Weiss, 2013). For example, Pennerstorfer and Weiss (2013) compiled a large database consisting of 18 quarterly observations on prices of diesel fuel (December 2000-March 2005) for an unbalanced sample of 595 to 1 370 gasoline stations in Austria. They supplemented price data with geographical/demographical data (i.e. population density, commuting behavior and importance of tourism) as well as individual characteristics of petrol stations in their comprehensive analysis. Based on a measure of spatial clustering of competitors, Pennerstorfer and Weiss (2013) substantiated the effects of local market power on petrol prices with a particular focus on the significance of coordinated pricing behaviour using the difference-in-difference approach (i.e. differentiation between the treatment and control groups). By comparing individual observations before and after the station conversion period, they concluded that spatial clustering of petrol stations lowers competition and increases prices (Pennerstorfer and Weiss, (2013).

There is also an emerging consensus in the extant literature that higher petrol prices and large retail price margins between rural and urban areas in Australia are largely attributable to the lack of competition in the market (Industry Commission, 1994; Australian Competition and Consumer Commission, 1996, 2007; Walker et al., 1997; Department of Parliamentary Services, 2004; Queensland Parliament, 2006). Setting excessive profit margins for fuel prices can adversely affect economic activity, particularly in rural and regional Australia, where petrol is used as a key intermediate input in the production of goods and services. Therefore, many policymakers in Australia have been sensitive to excessive profiteering behaviour in the fuel market since such practices can undermine economic growth. The Australian Competition and Consumer Commission (ACCC) (1996; 2007) launched several inquiries into possible price collusions in the oil industry but was unable to find any significant evidence of systematic price collusion among the major oil companies. However, a recurring theme in these inquiries was the significant difference between fuel prices in different geographical locations.

After reviewing evidence of the ACCC's inquiry in 1996, Walker et al. (1997) concluded that much of the urban-rural fuel price gap may be attributed to the lack of competition among oil importers coupled with limited market power of the independent discount retailers. The systematic rise in petrol prices since 2000 sparked the second wave of the fuel price debate. Eckert (2013, p. 140) summarises the prominence of this issue by pointing out that: "since 2000 alone, over 75 empirical studies of gasoline retailing have been published in English language academic journals, with many more studies existing in working paper form or as reports issued by governments or other agencies or institutes."

In order to enhance competition in the petrol market, the Western Australian state government introduced the fuel price monitoring scheme 'FuelWatch' in January 2001. The FuelWatch scheme has provided a comprehensive dataset on fuel prices down to the station level. After analysing FuelWatch data, the ACCC (2008) failed to identify any attempt by the oil majors to profiteer by manipulating fuel prices. However, Davidson (2008, p.8) cast doubts over the reliability of ACCC's conclusion by pointing to the fact that the ACCC model contained no "diagnostic statistics such as standard errors or p-values that one might expected in any econometric analysis". Using input-output analysis, Valadkhani and Mitchell (2002) demonstrated that although fuel price hikes would not have harmed the Australian economy to the extent as they did in the 1970s, these price hikes would nevertheless adversely affect poorer families. As a result, schemes such as FuelWatch could be readily justified by aiming to protect the interests of lower socioeconomic groups.

It should be noted that Valadkhani (2013 a) found that out of the 28 retail locations exhibiting significant petrol pricing asymmetry, none were from Western Australia, where FuelWatch is effectively monitoring petrol prices unlike the rest of the country. Valadkhani (2013b) examined the day of the week effect in retail prices of unleaded petrol across 114 retail locations in Australia during the period spanning from January 2005 to April 2012. He found that in major capital cities and urban areas prices generally peak on Thursday/Friday and then fall until they reach their cyclical trough on Tuesday. Valadkhani (2013b) also argues that petrol is more expensive in remote and small towns, where the economies of scale and scope are relatively limited and prices are less variable.

3. CLUSTER ANALYSIS OF REGIONAL LOCATIONS

For any given location, our key variable, the mean gross profit spread (margin) is the difference between retail and wholesale prices of petrol averaged over the sample period. It should be noted that there is generally more than one petrol station in each retail location and thus the retail prices in each location at any point of time are already the average of several petrol stations. In other words, the mean retail prices of petrol are averaged over time and over the retail outlets within each location. The wholesale prices are averaged over time only, as retailers purchase petrol from just one outlet which is usually the nearest outport terminal.

After computing the mean profit spread for each of the 109 retail locations, we need to obtain the data on transport costs, the number of service stations in the area (as a proxy for the extent of competition) and the extent of economies of scale and scope. Complete accurate data on the above control variables for various petrol stations within all 109 geographical locations are not available, therefore, we use the distance between retailers and their nearest wholesale distribution terminal as a proxy for transport costs. Consequently the further away each retail location is from its wholesale outport, the higher are expected transport and overhead costs. In addition, population is used as a proxy to capture the size of the market, the density of service stations within a certain geographical location and the extent of economies of scale and scope. When competition is localized in the gasoline market, the information on local differences in demand and cost and the share of informed vs. uniformed consumers are hard or impossible to obtain, thus making the assessment of the effects of coordinated behavior on prices very difficult (Pennerstorfer and Weiss, 2013).

Cluster analysis is a data-reduction technique which can be used to minimise within-group variance, while also maximising between-group variance, leading to a small number of heterogeneous groups with homogeneous contents (Hair et al., 1998). We thus adopt a hierarchical cluster analysis to group the 109 retail locations into several manageable clusters according to the following three variables: mean profit spread, population of the retail location, and distance to the nearest wholesaler. Before conducting a cluster analysis, these three variables are standardised to avoid bias resulting from variables having substantially different magnitudes or being measured in different units. This paper measures the similarity (in terms of the above three variables) between two retail locations, j and k, by the following squared Euclidean distance:

D(j,k) = [3.summation over (i=1)] [([X.sub.ij] - [X.sub.ik]).sup.2] (1)

where [X.sub.ij] and [X.sub.ik] denote the ith variable of locations j and k, respectively. The smaller D(j, k) is, the more similar are locations j and k in terms of the normalised magnitude of the three control variables. In hierarchical cluster analysis, at the beginning of the procedure there are 109 clusters, each representing one retail location. Then, at each stage, the two most similar locations (clusters) are combined until, at the last stage, a single cluster of 109 locations is formed. There are several alternative methods for merging the most similar pair of clusters at each stage namely the average linkage, the nearest centroid sorting, and the complete linkage, which is a conservative decision rule (Hirschberg et al., 1991), because it uses the maximum distance between any two attributes in the two clusters. In practice, compared to the above 3 methods, the Ward method is more widely used. This paper uses Ward's (1963) method, which chooses the two clusters whose merger would result in the smallest increase to the aggregate sum of squared deviations within clusters. The sum of squared deviations within cluster k is defined as follows:

ESS(k) = [summation over (J [member of] K)] [3.summation over (i=1)][([X.sub.ij] - [[bar.X].sub.ik]).sup.2] (2)

where [X.sub.ij] is the ith variable in location j, and [[bar.X].sub.ik] is the ith variable averaged across all locations in cluster k. Given the values of ESS(k), the increment to the aggregate sum of squared deviations within clusters resulting from the merger of cluster k and cluster K to form cluster (k[union]K) is computed by:

[d.sub.Ward](k,K) = [summation over (j [member of] (k [union] K)][3.summation over (i-1)] [([X.sub.ij] - [[bar.X].sub.(K[union]k)])).sup.2] - ESS(k) - ESS(K) (3)

Based on the resulting grouping of homogenous locations within each cluster, cluster analysis can provide a detailed understanding of the pricing behaviour of retailers at various geographical locations and reveal any possible evidence of abnormal pricing practices as presented below.

4. DATABASE

Retail and wholesale petrol prices were obtained from FUELtrac (www.fueltrac.com.au) and Informed Sources (www.informedsources.com) using funding made available under the Australian Research Council's Discovery Projects scheme. Close scrutiny of both databases revealed that there are only 109 retail locations for which consistent and complete price data (with no gap or missing observations) were available over the period 29 October 2007 to 30 January 2012. Petrol stations in these locations purchase petrol from their nearest wholesale outport terminal. In total there are 18 wholesale distribution terminals across 7 states and territories in Australia: 2 are in the State of New South Wales (Newcastle and Sydney), 1 in the Northern Territory (Darwin), 5 in Queensland (Brisbane, Cairns, Gladstone, Mackay, Townsville), 2 in South Australia (Adelaide, Port Lincoln), 2 in Tasmania (Hobart, Devonport), 1 in Victoria (Melbourne) and 5 in Western Australia (Albany, Esperence, Geraldton, Perth, Port Hedland). Population data were obtained from the Australian Bureau of Statistics (2012), and the distance between retail locations and their nearest wholesale terminals was approximated in kilometres using Google map assuming a minimum distance of 5 km.

5. EMPIRICAL RESULTS

Table 1 shows the descriptive statistics and the unit root test results using weekly data spanning from 29 October 2007 to 30 January 2012. During this period on average ten retail locations with the highest average gross profit margin (in cents per litre) were: Tennant Creek (24.1 cents per litre), Eucla (23.9), Alice Springs (21.1), Carnarvon (19.5), Cunnamulla (16.6), Bega (15.4), Cooma (15.3), Hay (14.9), Geraldton (14.4) and Ceduna (14.0). In contrast, the lowest margins were observed at the following retail locations: Mackay (2.4 cents per litre), Townsville (2.7), Bundaberg (4.0), Dalby (4.4), Perth metropolitan (4.7), Cairns (4.7), Caloundra (4.8), Mandurah (4.9), Warwick (5.0), and Brisbane metropolitan area (5.1). Overall it appears that the average margins in urban and more populous metropolitan areas (especially those in Queensland) are conspicuously less than regional areas, where the extent of economies of scope and scale is probably far more limited.

The average gross margin in Sydney metropolitan, as the most populous city in Australia, is 6.3 cents per litre, whereas in Eucla (with a population of only 86 persons) this margin is as high as 23.9 cents. Before computing the mean margin for each of the 109 locations, it is important to ensure that all individual 109 spread series follow a mean reverting pattern during the sample period when the average series are computed. According to the Augmented Dickey-Fuller (ADF) test results in Table 1, the null of unit root is rejected at the 5 percent level of significance or better for all of the 109 spread series. Therefore, we can assert that the spread series fluctuates around their mean values without showing upward or downward trends during the sample period. These results support the view that although the margins between retail and wholesale prices of petrol exhibit significant differences across various geographical locations, they follow a mean reverting pattern over time within their individual retail locations. In the context of the Austrian gasoline market, Pennerstorfer and Weiss (2013) provide convincing evidence that large gasoline price differences can be adequately explained by analysing the link between ownership structure and spatial clustering (i.e. the sequence of stations on a road). It is highly likely that retailers, which are generally members of a network of multi-station firms, can coordinate their pricing attempts within the spatial network due to the lack of competition.

A hierarchical cluster analysis is performed to identify in which comparable locations the average retail gross margins can be considered as relatively too high. To this end, the 109 x 109 proximity matrix is first computed which contains the squared Euclidean distances between all pairs of retail locations. This matrix is not reported here due to its large size, but is available from the author on request.

Table 2 shows how the clusters (or geographical locations) are merged at each stage of the procedure. At Stage 0 there are 109 separate clusters with each containing a single retail location. As shown in Table 2 (columns 2 and 3) at stage 1, Forster (Cluster 40) and Traralgon (Cluster 96) are combined. The number of clusters at the end of Stage 1 is 108 (see Column 5). The clusters that are formed at Stages 2 and 3 also involve the merging of two similar single-location clusters. At Stage 2, locations 57 (Katherine) and 60 (Lakes Entrance) are merged, and at Stage 3 locations 2 (Albany) and 34 (Devonport) are clustered. In this way, the most similar locations continue to merge until stage 37, where Forster (Cluster 40) and Traralgon (Cluster 96), as one cluster, are combined with location 88 (Shepparton). The individual locations, or cluster groupings will continue to merge in the same manner until stage 108, where there will be just one cluster containing all 109 locations.

The agglomeration schedule in Table 2 shows that retail location 4 (Alice Springs), as the most dissimilar location compared to the rest, does not join any cluster until the last stage (see stage 108) and, in so doing, increases the agglomeration coefficient markedly from 223 to 324. Figure 1 shows how clusters (or locations) are formed in a hierarchical clustering by using a dendrogram.

[FIGURE 1 OMITTED]

The maximum change in the agglomeration coefficient is used by some practitioners as a guide to determine the optimum number of clusters. This coefficient denotes the within-cluster sum of squares, aggregated across all clusters that are formed by a given stage of the procedure. Small increments in the agglomeration coefficient mean that relatively homogeneous clusters are being combined at the corresponding stage. Conversely, large increments in the agglomeration coefficient indicate that relatively more heterogeneous clusters are being grouped. The use of maximum change in the agglomeration coefficient as a stopping rule usually suggests too few clusters (Hair et al., 1998). The cubic clustering criterion is another alternative stopping rule that has the tendency to identify too many clusters. In order to overcome the shortcoming associated with adopting too few or too many cluster solutions, we report a range of solutions, varying from two to six. This approach offers a better understanding of the how different retail locations are clustered. However, our preferred final grouping is based on the six-cluster solution because the corresponding change in the agglomeration coefficient rises from one to two digits.

Table 3 reports the cluster analysis results using Ward's method and the normalised values of our three control variables. All five different cluster solutions are reported in columns 2-6 of Table 3. A cursory look at Table 3 reveals that, irrespective of the number of clusters, Melbourne and Sydney (both with large population and minimum distance to their supplying wholesale distributors) appear as separate clusters in all of the cluster solutions. Similarly, Tennant Creek, Eucla and Alice Springs are always grouped together, regardless of the number of clusters. In this cluster, while the mean gross profit margin is higher than that of the other clusters, their populations are relatively lower and at the same time the distances to their corresponding wholesale distributors are also further than others. These groupings can also be seen in the dendrogram presented in Figure 1.

A systematic notation is used to identify each location with a cluster membership code. For example, C6.1 denotes the first cluster within a six-cluster solution, C5.3 indicates the third cluster within a five-cluster solution and so forth. In the six-cluster solution the following 16 locations are separated from cluster 3 in the five-cluster solution (C5.3) and form a more homogeneous group (that is, C6.4): Geraldton, Wynyard, Wonthaggi, Mansfield, Echuca, Colac, Darwin metropolitan, Campbelltown, Ulverstone, Launceston, Burnie, Wollongong, Albany, Devonport, Port Lincoln and Hobart metropolitan.

All clusters in Table 3 are first sorted in terms of the ascending magnitude of the cluster average, using a six-cluster solution, and then within each cluster the corresponding members are further sorted in terms of individual average gross margins (in descending order). In this way, the sixth cluster appears right at the end of Table 3 as it has the highest average margin (that is, 23 cents per litre). Within the sixth cluster solution, the three locations are sorted in descending order, with Tennant Creek (24.1) appearing first and Alice Springs (21.1) last.

6. POLICY IMPLICATIONS OF THE STUDY

By comparing 'apples with apples' in Table 3, one can identify which retail locations charge relatively higher or lower gross margins than their comparable counterparts. In order to facilitate the comparison, all of the retail locations are sorted within a six-cluster solution in terms of a measure of the excess gross profitability index. This index quantifies how much the average gross profitability in a given location is above or below the corresponding cluster's average. For instance, within C6.1, the gross margin for Newcastle is on average 50 percent higher than the average margin for other comparable locations. It is interesting that within this same cluster there are other comparable locations in which the actual average gross margin is well below 7.8 cents per litre (see Table 3).

Another example of excessive margin is Carnarvon (19.5 cents per litre) in Western Australia within cluster C6.5 with a population of 6 333 and an approximate distance of 480 km to its wholesale outport terminal in Geraldton. For comparison purposes, it is interesting to note that the relevant margin in Longreach (12.4 cents per litre) in Queensland is well below that of Carnarvon, where the former has both a smaller population (4 384) and a longer distance to its wholesale outport terminal (786 km). Similarly, both Geraldton and Albany (two Western Australian country towns) in C6.4 are located only 5 kilometres away from their wholesaler, however, the excess profitability index in Geraldton is 29 percent above the cluster average, but 14 percent below the cluster average in Albany. This finding is compounded by the fact that Geraldton and Albany have similar population sizes.

A cursory look at the retail locations appearing at the top of clusters C6.1, C6.3, C6.4 and C6.5 provides convincing evidence that high profit margins are not observed only in remote rural and less populated regional areas. For example, although Rockhampton and Caloundra in C6.1 share similar attributes in terms of population size and the distance to the wholesaler, the excess profitability index is 29 percent above the cluster average in Rockhampton, compared to 8 percent below the cluster average in Caloundra. We also observe the same phenomenon in C6.3 when comparing two similar Victorian country towns, namely Benalla and Ararat, where the excess profitability indices are respectively 38 percent above and 12 percent below the cluster average. These results show that there may be other forces at work in determining gross profit margins.

It is important to explain why there are large profit margins in both rural and urban areas. Foss and Lien (2010) and Dayanandan and Donker (2011) have highlighted the importance of changes in ownership structures in revealing the competitive dynamics and thus profitability of oil and gas firms. It is very useful to explore whether excessive margins can be explained by the extent of competition, ownership structure, local market conditions and its individual characteristics. At this stage, we have not been able to obtain such data for all 109 locations from the ACCC, Fuelwatch or other sources. However, there is widespread view in the literature that large regional price variations are driven by the lack of competition and the number of independent suppliers. For example a similar comprehensive study of the Austrian gasoline market by Pennerstorfer and Weiss (2013) provides convincing evidence that large gasoline price differences can be adequately explained by analysing the link between ownership structure and spatial clustering (i.e. the sequence of stations on a road). An increase in the extent of spatial clustering can lower the degree of competition and hence raise equilibrium prices. It is highly likely that retailers, which are generally members of a network of multi-station firms, can coordinate their pricing attempts within the spatial network due to the lack of competition.

The Australian Institute of Petroleum (2012) report notes that the aggregate retail net profit margin should be in the vicinity of 6 cents per litre. However, this net profit margin appears to be well below the retail gross profit margin of 11 cents per litre identified in this study, even after deducting several cents per litre for overhead costs other than transport. We argue that if gross profit margins in a particular retail location are regarded as relatively excessive, then further careful examination of other factors should also be taken into account as these differences may be reasonably justifiable on other location specific factors.

Of key importance to this study is that it clearly highlights important geographical pricing differences at a disaggregate level. By identifying various urban and regional areas in which profit margins appear to be excessive, the results of this paper provide more transparency in the petrol market particularly for consumers and industries for which petrol is an essential intermediate input. Given the critical role of fuel prices in both regional and urban economies, consumers and regulatory bodies can directly benefit from greater efficiency and transparency of the petrol market. In light of recent debates surrounding suspected profiteering in the petrol industry, our results are both timely and relevant for consumers as well as government regulatory agencies.

7. CONCLUSION

Changes in petrol prices attract a great deal of attention from consumers since they spend a significant share of their income on this commodity. This paper examines the extent to which the average spread between retail and wholesale prices of petrol across 109 rural, urban geographical locations in Australia are relatively excessive. For this purpose, we conduct a hierarchical cluster analysis using a disaggregated database (during the period 29 October 2007-30 January 2012) not freely available to the public. We postulate that population (as a proxy for the market size, competition and the extent of economies of scale and scope) and the distance between retailers and wholesalers (as a proxy for transport costs) are the major determinants of the observed sizable differences in the gross margins across various geographical locations. The results indicate that retailers have enough market power to determine their margins, particularly when the local market is subject to less competition, potentially causing substantial cost inefficiencies for consumers and industries with high reliance on petrol for transportation and production.

Our cluster analysis classifies all of the 109 retail locations into six heterogeneous groups with each group containing homogenous and comparable contents in terms of the standardised magnitudes of the following control variables: the averaged spread between retail and wholesale prices of petrol, population, and the distance between the retailers and wholesalers. Within the six-cluster solution, we have ranked all of the 109 retail locations in terms of the excess-gross profitability index, quantifying the extent to which the average gross profitability in a particular retail location is above or below the corresponding cluster average.

Contrary to popular belief, our results provide compelling new evidence that excessively high profit margins are not necessarily observed only in remote and isolated rural areas. In other words, large gross profit margins do also exist in major urban areas such as Newcastle, Rockhampton, Geelong, North Coast, Wangaratta, Shepparton, Geraldton and Bega (see the top of each of the 6 clusters in Table 3). Despite petrol retailers being exposed to high levels of competition and the economies of scale and scope in these locations, they continue to charge relatively higher profit margins from motorists than their counterparts in other similar locations. Therefore, this study can provide important policy implications for consumers and regulators given the recent debates in relation to the whereabouts of suspected profiteering in the petrol market, which can equally affect both urban and regional economies.

ACKNOWLEDGEMNTS: We wish to thank two anonymous referees, whose invaluable inputs and comments considerably improved an earlier version of this article. The usual caveat applies. This research was supported under Australian Research Council's Discovery Projects funding scheme (DP120102753). The views expressed herein are those of the authors and are not necessarily those of the Australian Research Council.

REFERENCES

Asafu-Adjaye, J. (2000). The relationship between energy consumption, energy prices and economic growth: time series evidence from Asian developing countries. Energy Economics, 22(6), pp. 615625.

Australian Bureau of Statistics (2012). Regional Population Growth. cat. no. 3218, ABS, Canberra.

Australian Competition and Consumer Commission (1996). Inquiry Into the Petroleum Products Declaration. Australian Government Publishing Service, Canberra.

Australian Competition and Consumer Commission (2007). Petrol Prices and Australian Consumers: Report of the ACCC Inquiry Into the Price of Unleaded Petrol. Australian Government Publishing Service, Canberra.

Australian Competition and Consumer Commission (2008). ACCC Issues Details of Further FuelWatch Econometric Analysis. Press release, May. Online version accessed 15 August 2012, www.accc.gov.au.

Australian Institute of Petroleum, (2012). Facts About Petrol Prices. Online version accessed 23 September 2012 http://www.aip.com.au/pricing/pdf/Facts%20About%20Petrol%20 Prices.pdf

Brown, S.P.A. and Yucel, MK. (2002). Energy prices and aggregate economic activity: an interpretative survey. Quarterly Journal of Economics and Finance, 42(2), pp. 193-208.

Chontanawat, J., Hunt, L.C. and Pierse, R. (2008). Does energy consumption cause economic growth? evidence from a systematic study of over 100 countries. Journal of Policy Modeling, 30(2), pp. 209-220.

Davidson, S. (2008). Secret econometric business: watching FuelWatch and the ACCC. Agenda, 15(4), pp. 5-18.

Dayanandan, A. and Donker, H. (2011). Oil prices and accounting profits of oil and gas companies. International Review of Financial Analysis, 20, pp. 252-257.

Department of Parliamentary Services (2004). Petrol Pricing in Australia: Issues and trends, Current Issues Brief no. 10 (2003-04). Information and Research Services, Parliamentary Library, Department of Parliamentary Services, Canberra.

Dodson, J. and Sipe. N. (2007). Oil vulnerability in the Australian city: assessing socioeconomic risks from higher urban fuel prices. Urban Studies, 44(1), pp. 37-62.

Donnelly, W.A. (1982). The regional demand for petrol in Australia. Economic Record, 58(4), pp. 317-327.

Eckert, A. (2013). Empirical studies of gasoline retailing: a guide to the literature. Journal of Economic Surveys, 27(1), pp. 140-166.

Fatai, K., Oxley, L. and Scrimgeour, F.G. (2004). Modelling the causal relationship between energy consumption and GDP in New Zealand, Australia, India, Indonesia, the Philippines and Thailand. Mathematics and Computers in Simulation, 64(3), pp. 431-445.

Ferderer, P.J. (1996). Oil price volatility and the macroeconomy. Journal of Macroeconomics, 18(1), pp. 1-26.

Ford, G.S. (2011). An investigation into the relationship of retail gas prices on oil company profitability. Applied Economics, 43(27), pp. 4033-4039.

Foss, N.J. and Lien, L.B. (2010). Ownership and competitive dynamics. Quarterly Journal of Austrian Economics, 13, pp. 3-30.

Gargett, D. (2010). Petrol prices in Australia. Australasian Transport Research Forum 2010. Adelaide, Australia.

Hair, J.F., Anderson, R.E. and Tatham, R.L. (1998). Multivariate Data Analysis. Upper Saddle River, NJ: Prentice Hall.

Hamilton, J.D. (1983). Oil and the macroeconomy since World War II. Journal of Political Economy, 91(2), pp. 228-248.

Hamilton, J.D. (1996). This is what happened to the oil price macroeconomy relationship. Journal of Monetary Economics, 38(2), pp. 215-220.

Hastings, J. and Gilbert, R. (2005). Market power, vertical integration and the wholesale price of gasoline. The Journal of Industrial Economics, 53(4), pp. 469-492.

Hayashi F. (2000). Econometrics, Princeton University Press: Princeton, NJ.

Hensher, D.A., and Stanley, J.K. (2009). What if petrol increased to $10 per litre in Melbourne? Implications for travel behaviour and public transport demand. Road and Transport Research, 18(3), pp. 61-68.

Hirschberg, J.G., Maasoumi, E. and Slottje, D.J. (1991). Cluster analysis for measuring quality of life across countries. Journal of Econometrics, 50(1-2), pp. 131-150.

Hooker, M.A. (1996). What happened to the oil-macroeconomy relationship? Journal of Monetary Economics, 38(2), pp. 195-213.

Hosken, D.S., McMillan, R. and Taylor, C.T. (2008). Retail gasoline pricing: what do we know? International Journal of Industrial Organization, 26, pp. 1425-1436.

Industry Commission (1994). Petroleum Products, Report no. 40, July. Australian Government Publishing Service, Canberra.

Li, Z., Rose, J.M. and Hensher, D.A. (2010). Forecasting automobile petrol demand in Australia: an evaluation of empirical models. Transportation Research Part A, 44(1), pp. 16-38.

Mahadevan, R. and Asafu-Adjaye, J. (2007). Energy consumption, economic growth and prices: a reassessment using panel VECM for developed and developing countries. Energy Policy, 35(4), pp. 2481-2490.

Mork, K.A. (1989). Oil and the macroeconomy when prices go up and down: an extension of Hamilton's results. Journal of Political Economy, 97(3), pp. 740-744.

Newman, P.W.G. and Kenworthy, J.R. (1989). Gasoline consumption and cities. Journal of the American Planning Association, 55(1), pp. 24-37.

Pennerstorfer, D. and Weiss, C. (2013). Spatial clustering and market power: evidence from the retail gasoline market. Regional Science And Urban Economics, 43(4), pp. 661-675.DOI: 10.1016/j.regsciurbeco.2013.04.002

Queensland Parliament (2006). Inquiry Into Petrol Pricing in Queensland. Legislative Assembly of Queensland, Brisbane.

Samimi, R. (1995). Road transport energy demand in Australia: a cointegration approach. Energy Economics, 17(4), pp. 329-339.

Simpson, J., Taylor, C. (2008). Do gasoline mergers affect consumer prices? the marathon Ashland petroleum and Ultramar diamond Shamrock transaction. Journal of Law and Economics, 51, pp. 135-152.

Valadkhani, A., (2013a). Do petrol prices rise faster than they fall when the market shows significant disequilibria? Energy Economics, 39, pp. 66-80.

Valadkhani, A. (2013b). Seasonal patterns in daily prices of unleaded petrol across Australia. Energy Policy, 56(3), pp. 720-731.

Valadkhani, A., Mitchell. W.F. (2002). Assessing the impact of changes in petroleum prices on inflation and household expenditures in Australia. Australian Economic Review, 35(2), pp. 122-132.

Walker, G., Murphy, T., and Hicks. J. (1997). Country petrol pricing: evidence and recommendations. Australasian Journal of Regional Studies, 3(1), pp. 85-98.

Ward, J.H. (1963). Hierarchical grouping to optimize an objective function. Journal of the American Statistical Association, 58(301), pp. 236-244.

Abbas Valadkhani

Swinburne Business School, Swinburne University of Technology, Hawthorn, VIC, 3122, Australia. E-mail: abbas@swin.edu.au

George Chen

UNE Business School, University of New England, Armidale, NSW, 2351, Australia. E-mail: gchen2@une.edu.au

John Anderson

UNE Business School, University of New England, Armidale, NSW, 2351, Australia. E-mail: John.Anderson@une.edu.au

Table 1. Descriptive Statistics and the Unit Root
Test Results (October 2007-January 2012).

No.   Retailer          Mean (a)   Rank     Standard
      location                            deviation (a)

1     Adelaide Metro      5.6       95        2.43
2     Albany              9.6       47        3.67
3     Albury              5.3       98        3.77
4     Alice Springs       21.1      3         5.01
5     Ararat              6.9       79        3.84
6     Bairnsdale          7.0       77        3.64
7     Ballarat            5.6       94        4.06
8     Bathurst            9.6       46        3.65
9     Bega                15.4      6         2.87
10    Benalla             10.8      34        3.20
11    Bendigo             6.8       80        3.50
12    Bowen               5.8       91        5.83
13    Brisbane Metro      5.1      100        5.40
14    Broken Hill         11.8      27        3.70
15    Bunbury             8.3       63        2.61
16    Bundaberg           4.0      107        4.39
17    Burnie              10.4      40        2.97
18    Caboolture          5.4       97        5.57
19    Cairns              4.7      104        5.16
20    Caloundra           4.8      103        5.92
21    Campbelltown        11.4      30        2.88
22    Canberra Metro      9.2       55        3.69
23    Carnarvon           19.5      4         5.19
24    Casino              7.7       69        5.07
25    Ceduna              14.0      10        3.16
26    Charters Towers     7.6       71        5.06
27    Coffs Harbour       10.6      36        3.05
28    Colac               11.9      24        3.81
29    Cooma               15.3      7         2.87
30    Cowra               8.8       59        3.28
31    Cunnamulla          16.6      5         6.28
32    Dalby               4.4      106        5.20
33    Darwin Metro        11.5      28        4.00
34    Devonport           9.5       52        3.53
35    Dubbo               9.6       51        3.73
36    Echuca              11.9      23        3.81
37    Emerald             6.6       83        4.84
38    Eucla               23.9      2         6.74
39    Forbes              13.3      12        2.85
40    Forster             9.7       44        5.23
41    Geelong             6.2       86        2.48
42    Geraldton           14.4      9         3.90
43    Gladstone           5.2       99        4.91
44    Gold Coast          5.8       92        5.13
45    Goondiwindi         7.6       70        6.42
46    Goulburn            7.9       66        3.60
47    Grafton             9.1       56        4.61
48    Griffith            11.5      29        3.74
49    Gympie              5.9       89        6.12
50    Hervey Bay          6.9       78        5.52
51    Hay                 14.9      8         3.41
52    Hobart Metro        8.7       61        3.88
53    Horsham             12.3      21        2.85
54    Inverell            10.6      35        4.18
55    Ipswich             5.5       96        5.33
56    Kalgoorlie          10.0      42        4.15
57    Katherine           11.1      32        4.70
58    Kempsey             9.9       43        3.51
59    Kingaroy            6.2       87        5.63
60    Lakes Entrance      11.0      33        3.79
61    Launceston          10.6      39        3.37
62    Lismore             8.2       64        4.74
63    Longreach           12.4      20        6.18
64    Mackay              2.4      109        5.37
65    Mandurah            4.9      102        3.11
66    Mansfield           12.7      17        3.17
67    Maryborough         5.7       93        5.32
68    Melbourne Metro     7.3       74        2.07
69    Mildura             12.5      19        3.47
70    Moree               12.7      16        4.05
71    Mt Gambier          8.7       60        3.88
72    Murray Bridge       7.3       75        2.63
73    New Norfolk         7.4       72        4.53
74    Newcastle           7.8       68        2.48
75    North Coast         6.1       88        5.32
76    Orange              11.8      25        3.04
77    Parkes              12.6      18        3.09
78    Perth Metro         4.7      105        2.60
79    Port Augusta        8.1       65        4.26
80    Port Lincoln        9.2       53        3.60
81    Port Macquarie      9.0       57        3.54
82    Port Pirie          8.7       62        2.12
83    Portland            12.8      15        3.08
84    Renmark             5.8       90        4.94
85    Rockhampton         6.7       82        5.35
86    Roma                9.6       50        5.82
87    Sale                9.2       54        3.01
88    Shepparton          10.2      41        3.56
89    Sunbury             7.3       73        2.15
90    Swan Hill           13.4      11        2.40
91    Sydney Metro        6.3       85        1.99
92    Tamworth            11.2      31        2.76
93    Taree               6.7       81        4.33
94    Tennant Creek       24.1      1         5.47
95    Townsville          2.7      108        4.62
96    Traralgon           9.6       49        3.54
97    Ulverstone          10.6      38        3.20
98    Victor Harbour      7.8       67        3.70
99    Wagga Wagga         12.3      22        3.59
100   Wangaratta          10.6      37        2.43
101   Warrnambool         8.8       58        3.42
102   Warwick             5.0      101        5.68
103   Whyalla             6.6       84        5.84
104   Wodonga             7.1       76        3.07
105   Wollongong          9.7       45        2.14
106   Wonthaggi           13.2      14        2.44
107   Wynyard             13.2      13        2.63
108   Yarrawonga          9.6       48        4.25
109   Yass                11.8      26        2.94
      Average             9.5                 4.00

No.   Retailer          ADF t stat.   P-value
      location            (b) (c)

1     Adelaide Metro      -10.96       0.00
2     Albany               -5.16       0.00
3     Albury               -6.12       0.00
4     Alice Springs        -5.08       0.00
5     Ararat               -6.38       0.00
6     Bairnsdale           -4.71       0.00
7     Ballarat             -4.82       0.00
8     Bathurst             -5.16       0.00
9     Bega                 -4.74       0.00
10    Benalla              -5.51       0.00
11    Bendigo              -6.30       0.00
12    Bowen                -4.68       0.00
13    Brisbane Metro       -5.62       0.00
14    Broken Hill          -5.68       0.00
15    Bunbury              -5.16       0.00
16    Bundaberg            -5.64       0.00
17    Burnie               -5.89       0.00
18    Caboolture           -5.52       0.00
19    Cairns               -5.69       0.00
20    Caloundra            -5.94       0.00
21    Campbelltown         -5.61       0.00
22    Canberra Metro       -7.60       0.00
23    Carnarvon            -4.62       0.00
24    Casino               -4.70       0.00
25    Ceduna               -4.94       0.00
26    Charters Towers      -4.68       0.00
27    Coffs Harbour        -6.49       0.00
28    Colac                -5.08       0.00
29    Cooma                -6.01       0.00
30    Cowra                -5.99       0.00
31    Cunnamulla           -4.93       0.00
32    Dalby                -5.37       0.00
33    Darwin Metro         -4.21       0.01
34    Devonport            -5.06       0.00
35    Dubbo                -4.60       0.00
36    Echuca               -5.28       0.00
37    Emerald              -5.81       0.00
38    Eucla                -3.86       0.02
39    Forbes               -5.96       0.00
40    Forster              -4.01       0.01
41    Geelong              -7.61       0.00
42    Geraldton            -5.16       0.00
43    Gladstone            -4.07       0.01
44    Gold Coast           -5.77       0.00
45    Goondiwindi          -4.88       0.00
46    Goulburn             -4.79       0.00
47    Grafton              -4.79       0.00
48    Griffith             -5.02       0.00
49    Gympie               -4.81       0.00
50    Hervey Bay           -4.67       0.00
51    Hay                  -5.59       0.00
52    Hobart Metro         -5.38       0.00
53    Horsham              -5.99       0.00
54    Inverell             -5.54       0.00
55    Ipswich              -6.45       0.00
56    Kalgoorlie           -5.46       0.00
57    Katherine            -4.54       0.00
58    Kempsey              -5.34       0.00
59    Kingaroy             -5.16       0.00
60    Lakes Entrance       -4.15       0.01
61    Launceston           -5.79       0.00
62    Lismore              -4.77       0.00
63    Longreach            -4.60       0.00
64    Mackay               -5.76       0.00
65    Mandurah             -5.22       0.00
66    Mansfield            -4.85       0.00
67    Maryborough          -4.72       0.00
68    Melbourne Metro      -8.35       0.00
69    Mildura              -5.87       0.00
70    Moree                -6.05       0.00
71    Mt Gambier           -5.39       0.00
72    Murray Bridge        -5.30       0.00
73    New Norfolk          -6.26       0.00
74    Newcastle            -7.06       0.00
75    North Coast          -3.73       0.02
76    Orange               -5.97       0.00
77    Parkes               -5.54       0.00
78    Perth Metro          -5.74       0.00
79    Port Augusta         -5.45       0.00
80    Port Lincoln         -5.18       0.00
81    Port Macquarie       -4.39       0.00
82    Port Pirie           -5.99       0.00
83    Portland             -5.51       0.00
84    Renmark              -4.92       0.00
85    Rockhampton          -5.33       0.00
86    Roma                 -4.43       0.00
87    Sale                 -5.64       0.00
88    Shepparton           -4.34       0.00
89    Sunbury              -8.42       0.00
90    Swan Hill            -4.90       0.00
91    Sydney Metro         -8.61       0.00
92    Tamworth             -6.18       0.00
93    Taree                -5.13       0.00
94    Tennant Creek        -4.85       0.00
95    Townsville           -5.99       0.00
96    Traralgon            -5.75       0.00
97    Ulverstone           -5.47       0.00
98    Victor Harbour       -5.22       0.00
99    Wagga Wagga          -4.68       0.00
100   Wangaratta           -6.01       0.00
101   Warrnambool          -6.68       0.00
102   Warwick              -3.85       0.02
103   Whyalla              -4.52       0.00
104   Wodonga              -5.69       0.00
105   Wollongong           -7.47       0.00
106   Wonthaggi            -6.41       0.00
107   Wynyard              -5.98       0.00
108   Yarrawonga           -4.90       0.00
109   Yass                 -5.93       0.00
      Average

Note.--(a) cents per litre. (b) The Schwarz information
criterion is utilised to select the optimal lag length,
including both an intercept term and a time trend
variable. (c) Following Hayashi (2000), given T=223
weekly observations, the upper bound in search of the
optimum lag length is assumed to be 14.

Source: the Authors

Table 2. Agglomeration Schedule.

Stage   Cluster Combined              Agglomeration    No. of
                                      Coefficients    clusters

        Cluster 1(a)   Cluster 2(a)

1            40             96            .000          108
2            57             60            .001          107
3            2              34            .001          106
4            67             84            .003          105
5            18             55            .004          104
6            73             89            .006          103
7            77             83            .009          102
8            30             47            .012          101
9            46             62            .016          100
10           58            108            .021           99
11           6              50            .025           98
12           11             93            .030           97
13           76            109            .035           96
14           81            101            .040           95
15           2              80            .046           94
16           57             92            .051           93
17           35             56            .058           92
18           8              87            .065           91
19           53             70            .073           90
20           20             65            .081           89
21           29             51            .089           88
22           17             97            .097           87
23           9              29            .105           86
24           12             59            .114           85
25           72             98            .124           84
26           39             90            .135           83
27           64             95            .147           82
28           21             28            .160           81
29           6             103            .173           80
30           15             46            .186           79
31           16             32            .201           78
32           81             82            .216           77
33           45            104            .233           76
34           10            100            .250           75
35           19             43            .269           74
36           48             99            .290           73
37           40             88            .312           72
38           39             77            .336           71
39           5              24            .359           70
40           14             69            .383           69
41           30             79            .412           68
42           12             49            .441           67
43           36             66            .470           66
44           27             35            .502           65
45           41             85            .534           64
46           72             73            .572           63
47           7             102            .610           62
48           71             86            .650           61
49           37             45            .693           60
50           8              40            .736           59
51           11             26            .780           58
52          106            107            .828           57
53           53             76            .890           56
54           18             19            .954           55
55           7              20            1.022          54
56           17             61            1.092          53
57           41             75            1.166          52
58           9              25            1.242          51
59           5              15            1.319          50
60           12             67            1.399          49
61           38             94            1.487          48
62           2              52            1.581          47
63           14             48            1.678          46
64           10             58            1.778          45
65           17             33            1.889          44
66           21             36            2.004          43
67           30             81            2.130          42
68           6              37            2.263          41
69           13             78            2.402          40
70           5              11            2.555          39
71           53             57            2.709          38
72           27             71            2.874          37
73           3              6             3.054          36
74           14             54            3.238          35
75           17            105            3.431          34
76           44             74            3.625          33
77           42            106            3.820          32
78           8              10            4.076          31
79           7              41            4.337          30
80           12             16            4.613          29
81           68             91            4.931          28
82           22             30            5.261          27
83           7              18            5.598          26
84           2              17            5.956          25
85           39             53            6.417          24
86           31             63            6.988          23
87           8              22            7.617          22
88           21             42            8.305          21
89           1              44            9.037          20
90           5              72            9.825          19
91           14             27           10.915          18
92           3              12           12.008          17
93           7              64           13.118          16
94           9              23           14.294          15
95           2              21           17.019          14
96           3              5            19.806          13
97           14             39           22.604          12
98           1              7            25.517          11
99           9              31           29.443          10
100          4              38           33.786          9
101          3              8            38.826          8
102          9              14           46.269          7
103          1              13           58.055          6
104          2              3            71.928          5
105          1              2            92.378          4
106          4              9            138.109         3
107          1              68           223.185         2
108          1              4            324.000         1

Note.--(a) See the first and second columns in Table 1 to
identify the retail location. Source: the Authors

Table 3. Gross profitability index using a six-cluster solution.

Retailer          Cluster Membership Codes           AGM    REPI
Location                                                    (%)

                  6      5      4      3      2

Newcastle         C6.1   C5.1   C4.1   C3.1   C2.1   7.8    50
Rockhampton       C6.1   C5.1   C4.1   C3.1   C2.1   6.7    29
Geelong           C6.1   C5.1   C4.1   C3.1   C2.1   6.2    19
North Coast       C6.1   C5.1   C4.1   C3.1   C2.1   6.1    17
Gold Coast        C6.1   C5.1   C4.1   C3.1   C2.1   5.8    12
Ballarat          C6.1   C5.1   C4.1   C3.1   C2.1   5.6    8
Adelaide Metro    C6.1   C5.1   C4.1   C3.1   C2.1   5.6    8
Ipswich           C6.1   C5.1   C4.1   C3.1   C2.1   5.5    6
Caboolture        C6.1   C5.1   C4.1   C3.1   C2.1   5.4    4
Gladstone         C6.1   C5.1   C4.1   C3.1   C2.1   5.2    0
Brisbane Metro    C6.1   C5.1   C4.1   C3.1   C2.1   5.1    -2
Warwick           C6.1   C5.1   C4.1   C3.1   C2.1   5.0    -4
Mandurah          C6.1   C5.1   C4.1   C3.1   C2.1   4.9    -6
Caloundra         C6.1   C5.1   C4.1   C3.1   C2.1   4.8    -8
Cairns            C6.1   C5.1   C4.1   C3.1   C2.1   4.7    -10
Perth Metro       C6.1   C5.1   C4.1   C3.1   C2.1   4.7    -10
Townsville        C6.1   C5.1   C4.1   C3.1   C2.1   2.7    -48
Mackay            C6.1   C5.1   C4.1   C3.1   C2.1   2.4    -54
Cluster average   C6.1                               5.2    0
Melbourne Metro   C6.2   C5.2   C4.2   C3.2   C2.1   7.3    7
Sydney Metro      C6.2   C5.2   C4.2   C3.2   C2.1   6.3    -7
Cluster average   C6.2                               6.8    0
Benalla           C6.3   C5.3   C4.1   C3.1   C2.1   10.8   38
Wangaratta        C6.3   C5.3   C4.1   C3.1   C2.1   10.6   36
Shepparton        C6.3   C5.3   C4.1   C3.1   C2.1   10.2   31
Kempsey           C6.3   C5.3   C4.1   C3.1   C2.1   9.9    27
Forster           C6.3   C5.3   C4.1   C3.1   C2.1   9.7    24
Bathurst          C6.3   C5.3   C4.1   C3.1   C2.1   9.6    23
Yarrawonga        C6.3   C5.3   C4.1   C3.1   C2.1   9.6    23
Traralgon         C6.3   C5.3   C4.1   C3.1   C2.1   9.6    23
Sale              C6.3   C5.3   C4.1   C3.1   C2.1   9.2    18
Canberra Metro    C6.3   C5.3   C4.1   C3.1   C2.1   9.2    18
Grafton           C6.3   C5.3   C4.1   C3.1   C2.1   9.1    17
Port Macquarie    C6.3   C5.3   C4.1   C3.1   C2.1   9.0    15
Warrnambool       C6.3   C5.3   C4.1   C3.1   C2.1   8.8    13
Cowra             C6.3   C5.3   C4.1   C3.1   C2.1   8.8    13
Port Pirie        C6.3   C5.3   C4.1   C3.1   C2.1   8.7    12
Bunbury           C6.3   C5.3   C4.1   C3.1   C2.1   8.3    6
Lismore           C6.3   C5.3   C4.1   C3.1   C2.1   8.2    5
Port Augusta      C6.3   C5.3   C4.1   C3.1   C2.1   8.1    4
Goulburn          C6.3   C5.3   C4.1   C3.1   C2.1   7.9    1
Victor Harbour    C6.3   C5.3   C4.1   C3.1   C2.1   7.8    0
Casino            C6.3   C5.3   C4.1   C3.1   C2.1   7.7    -1
Goondiwindi       C6.3   C5.3   C4.1   C3.1   C2.1   7.6    -3
Charters Towers   C6.3   C5.3   C4.1   C3.1   C2.1   7.6    -3
New Norfolk       C6.3   C5.3   C4.1   C3.1   C2.1   7.4    -5
Sunbury           C6.3   C5.3   C4.1   C3.1   C2.1   7.3    -6
Murray Bridge     C6.3   C5.3   C4.1   C3.1   C2.1   7.3    -6
Wodonga           C6.3   C5.3   C4.1   C3.1   C2.1   7.1    -9
Bairnsdale        C6.3   C5.3   C4.1   C3.1   C2.1   7.0    -10
Hervey Bay        C6.3   C5.3   C4.1   C3.1   C2.1   6.9    -12
Ararat            C6.3   C5.3   C4.1   C3.1   C2.1   6.9    -12
Bendigo           C6.3   C5.3   C4.1   C3.1   C2.1   6.8    -13
Taree             C6.3   C5.3   C4.1   C3.1   C2.1   6.7    -14
Emerald           C6.3   C5.3   C4.1   C3.1   C2.1   6.6    -15
Whyalla           C6.3   C5.3   C4.1   C3.1   C2.1   6.6    -15
Kingaroy          C6.3   C5.3   C4.1   C3.1   C2.1   6.2    -21
Gympie            C6.3   C5.3   C4.1   C3.1   C2.1   5.9    -24
Renmark           C6.3   C5.3   C4.1   C3.1   C2.1   5.8    -26
Bowen             C6.3   C5.3   C4.1   C3.1   C2.1   5.8    -26
Maryborough       C6.3   C5.3   C4.1   C3.1   C2.1   5.7    -27
Albury            C6.3   C5.3   C4.1   C3.1   C2.1   5.3    -32
Dalby             C6.3   C5.3   C4.1   C3.1   C2.1   4.4    -44
Bundaberg         C6.3   C5.3   C4.1   C3.1   C2.1   4.0    -49
Cluster average   C6.3                               7.8    0
Geraldton         C6.4   C5.3   C4.1   C3.1   C2.1   14.4   29
Wynyard           C6.4   C5.3   C4.1   C3.1   C2.1   13.2   18
Wonthaggi         C6.4   C5.3   C4.1   C3.1   C2.1   13.2   18
Mansfield         C6.4   C5.3   C4.1   C3.1   C2.1   12.7   13
Echuca            C6.4   C5.3   C4.1   C3.1   C2.1   11.9   6
Colac             C6.4   C5.3   C4.1   C3.1   C2.1   11.9   6
Darwin Metro      C6.4   C5.3   C4.1   C3.1   C2.1   11.5   3
Campbelltown      C6.4   C5.3   C4.1   C3.1   C2.1   11.4   2
Ulverstone        C6.4   C5.3   C4.1   C3.1   C2.1   10.6   -5
Launceston        C6.4   C5.3   C4.1   C3.1   C2.1   10.6   -5
Burnie            C6.4   C5.3   C4.1   C3.1   C2.1   10.4   -7
Wollongong        C6.4   C5.3   C4.1   C3.1   C2.1   9.7    -13
Albany            C6.4   C5.3   C4.1   C3.1   C2.1   9.6    -14
Devonport         C6.4   C5.3   C4.1   C3.1   C2.1   9.5    -15
Port Lincoln      C6.4   C5.3   C4.1   C3.1   C2.1   9.2    -18
Hobart Metro      C6.4   C5.3   C4.1   C3.1   C2.1   8.7    -22
Cluster average   C6.4                               11.2   0
Carnarvon         C6.5   C5.4   C4.3   C3.3   C2.2   19.5   56
Cunnamulla        C6.5   C5.4   C4.3   C3.3   C2.2   16.6   33
Bega              C6.5   C5.4   C4.3   C3.3   C2.2   15.4   23
Cooma             C6.5   C5.4   C4.3   C3.3   C2.2   15.3   22
Hay               C6.5   C5.4   C4.3   C3.3   C2.2   14.9   19
Ceduna            C6.5   C5.4   C4.3   C3.3   C2.2   14.0   12
Swan Hill         C6.5   C5.4   C4.3   C3.3   C2.2   13.4   7
Forbes            C6.5   C5.4   C4.3   C3.3   C2.2   13.3   6
Portland          C6.5   C5.4   C4.3   C3.3   C2.2   12.8   2
Moree             C6.5   C5.4   C4.3   C3.3   C2.2   12.7   2
Parkes            C6.5   C5.4   C4.3   C3.3   C2.2   12.6   1
Mildura           C6.5   C5.4   C4.3   C3.3   C2.2   12.5   0
Longreach         C6.5   C5.4   C4.3   C3.3   C2.2   12.4   -1
Horsham           C6.5   C5.4   C4.3   C3.3   C2.2   12.3   -2
Wagga Wagga       C6.5   C5.4   C4.3   C3.3   C2.2   12.3   -2
Orange            C6.5   C5.4   C4.3   C3.3   C2.2   11.8   -6
Yass              C6.5   C5.4   C4.3   C3.3   C2.2   11.8   -6
Broken Hill       C6.5   C5.4   C4.3   C3.3   C2.2   11.8   -6
Griffith          C6.5   C5.4   C4.3   C3.3   C2.2   11.5   -8
Tamworth          C6.5   C5.4   C4.3   C3.3   C2.2   11.2   -10
Katherine         C6.5   C5.4   C4.3   C3.3   C2.2   11.1   -11
Lakes Entrance    C6.5   C5.4   C4.3   C3.3   C2.2   11     -12
Inverell          C6.5   C5.4   C4.3   C3.3   C2.2   10.6   -15
Coffs Harbour     C6.5   C5.4   C4.3   C3.3   C2.2   10.6   -15
Kalgoorlie        C6.5   C5.4   C4.3   C3.3   C2.2   10     -20
Roma              C6.5   C5.4   C4.3   C3.3   C2.2   9.6    -23
Dubbo             C6.5   C5.4   C4.3   C3.3   C2.2   9.6    -23
Mt Gambier        C6.5   C5.4   C4.3   C3.3   C2.2   8.7    -30
Cluster average   C6.5                               12.5   0
Tennant Creek     C6.6   C5.5   C4.4   C3.3   C2.2   24.1   5
Eucla             C6.6   C5.5   C4.4   C3.3   C2.2   23.9   4
Alice Springs     C6.6   C5.5   C4.4   C3.3   C2.2   21.1   -8
Cluster average   C6.6                               23.0   0

Retailer          Population   Distance to   Wholesale      Wholesaler
Location          (Persons)    Wholesaler    Distributor    state
                               Km

Newcastle         552776       5             Newcastle      NSWACT
Rockhampton       78643        109           Gladstone      QLD
Geelong           180805       75            Melbourne      VIC
North Coast       335273       95            Brisbane       QLD
Gold Coast        600475       80            Brisbane       QLD
Ballarat          97810        115           Melbourne      VIC
Adelaide Metro    1212982      5             Adelaide       SA
Ipswich           172738       40            Brisbane       QLD
Caboolture        158988       50            Brisbane       QLD
Gladstone         52949        5             Gladstone      QLD
Brisbane Metro    2074222      5             Brisbane       QLD
Warwick           12659        156           Brisbane       QLD
Mandurah          89559        70            Perth          WA
Caloundra         51991        94            Brisbane       QLD
Cairns            153075       5             Cairns         QLD
Perth Metro       1738807      5             Perth          WA
Townsville        176347       5             Townsville     QLD
Mackay            87324        5             Mackay         QLD
Cluster average   434857       49
Melbourne Metro   4137432      5             Melbourne      VIC
Sydney Metro      4627345      5             Sydney         NSWACT
Cluster average   4382389      5
Benalla           9129         212           Melbourne      VIC
Wangaratta        29018        252           Melbourne      VIC
Shepparton        50373        190           Melbourne      VIC
Kempsey           29581        280           Newcastle      NSWACT
Forster           18372        164           Newcastle      NSWACT
Bathurst          34561        203           Sydney         NSWACT
Yarrawonga        5727         282           Melbourne      VIC
Traralgon         31105        164           Melbourne      VIC
Sale              14782        214           Melbourne      VIC
Canberra Metro    417860       287           Sydney         NSWACT
Grafton           24798        315           Brisbane       NSWACT
Port Macquarie    44793        245           Newcastle      NSWACT
Warrnambool       34193        265           Melbourne      VIC
Cowra             12940        309           Sydney         NSWACT
Port Pirie        18169        224           Adelaide       SA
Bunbury           70037        172           Perth          WA
Lismore           32617        198           Brisbane       NSWACT
Port Augusta      14725        306           Adelaide       SA
Goulburn          22225        197           Sydney         NSWACT
Victor Harbour    14219        83            Adelaide       SA
Casino            11414        229           Brisbane       NSWACT
Goondiwindi       11437        348           Brisbane       QLD
Charters Towers   12978        136           Townsville     QLD
New Norfolk       5230         35            Hobart         TAS
Sunbury           36658        40            Melbourne      VIC
Murray Bridge     19724        78            Adelaide       SA
Wodonga           51899        322           Melbourne      VIC
Bairnsdale        11282        282           Melbourne      VIC
Hervey Bay        61691        294           Brisbane       QLD
Ararat            8215         205           Melbourne      VIC
Bendigo           92934        154           Melbourne      VIC
Taree             49453        169           Newcastle      NSWACT
Emerald           18410        370           Gladstone      QLD
Whyalla           23430        267           Port Lincoln   SA
Kingaroy          14601        225           Brisbane       QLD
Gympie            50011        169           Brisbane       QLD
Renmark           9834         257           Adelaide       SA
Bowen             14515        203           Townsville     QLD
Maryborough       28520        256           Brisbane       QLD
Albury            107086       327           Melbourne      NSWACT
Dalby             11419        208           Brisbane       QLD
Bundaberg         69500        185           Gladstone      QLD
Cluster average   39273        222
Geraldton         37842        5             Geraldton      WA
Wynyard           11530        66            Devonport      TAS
Wonthaggi         6529         136           Melbourne      VIC
Mansfield         7998         192           Melbourne      VIC
Echuca            2261         221           Melbourne      VIC
Colac             10857        152           Melbourne      VIC
Darwin Metro      128073       5             Darwin         NT
Campbelltown      772          130           Hobart         TAS
Ulverstone        9760         21            Devonport      TAS
Launceston        106655       100           Devonport      TAS
Burnie            17729        48            Devonport      TAS
Wollongong        293503       85            Sydney         NSWACT
Albany            36551        5             Albany         WA
Devonport         25639        5             Devonport      TAS
Port Lincoln      14739        5             Port Lincoln   SA
Hobart Metro      216656       5             Hobart         TAS
Cluster average   57943        72
Carnarvon         6333         480           Geraldton      WA
Cunnamulla        1217         804           Brisbane       QLD
Bega              34035        425           Sydney         NSWACT
Cooma             10524        399           Sydney         NSWACT
Hay               3315         421           Melbourne      NSWACT
Ceduna            3828         403           Port Lincoln   SA
Swan Hill         22275        343           Melbourne      VIC
Forbes            9818         375           Sydney         NSWACT
Portland          11531        363           Melbourne      VIC
Moree             14465        315           Brisbane       NSWACT
Parkes            15267        358           Sydney         NSWACT
Mildura           50909        542           Melbourne      VIC
Longreach         4384         786           Gladstone      QLD
Horsham           14125        300           Melbourne      VIC
Wagga Wagga       59005        458           Sydney         NSWACT
Orange            40062        258           Sydney         NSWACT
Yass              15450        279           Sydney         NSWACT
Broken Hill       19703        516           Adelaide       NSWACT
Griffith          26001        461           Melbourne      NSWACT
Tamworth          48262        306           Newcastle      NSWACT
Katherine         9967         317           Darwin         NT
Lakes Entrance    12070        318           Melbourne      VIC
Inverell          5013         569           Sydney         NSWACT
Coffs Harbour     53798        390           Brisbane       NSWACT
Kalgoorlie        32841        390           Esperance      WA
Roma              7191         476           Brisbane       QLD
Dubbo             38383        394           Sydney         NSWACT
Mt Gambier        26206        436           Adelaide       SA
Cluster average   21285        424
Tennant Creek     3555         989           Darwin         NT
Eucla             86           894           Port Lincoln   WA
Alice Springs     27589        1498          Darwin         NT
Cluster average   10410        1127

Notes: This Table is first sorted in terms of the ascending
magnitudes of the corresponding cluster averages, whereby the
cluster average for C6.1, C6.2, C6.3, C6.4, C6.5 and C6.6 are 5.2,
6.8, 7.8, 11.2, 12.5 and 23.0 (cents per litre), respectively.
Then, the retail locations within each cluster are sorted in terms
of the descending values of their gross margins. AGM=Average Gross
Margin (Cents Per Litre). REPI=Relative excess (gross) profitability
index measures how much the gross profit margin in a given location
is above or below the corresponding cluster's average.

Source: the Authors