What Determines the Profitability of a Retail Gasoline Outlet?

A Study for the Competition Bureau of Canada

---

4. The Data

This section outlines the data employed in the econometric analysis. Data on retail and wholesale prices and sales volume of regular, mid-grade, and premium gasoline were obtained from one station each belonging to five companies. Of these, outlets C and D represent the vertically-integrated firms with refining capabilities, while independent retailers own outlets A and B with no refining capacity.¹ With respect to retail prices and volume, outlets B, C, and D provided data for a relatively long time period (January 2002 – December 2004), while corresponding data from outlet A was over a shorter duration of time (January 2005 – May 2005). However, wholesale prices were provided over similar time periods (January 2002- December 2004 for outlets B, C, and D; August 2002 – April 2005 for outlet A).  The following tables summarize key characteristics of these data.

In most cases average daily retail prices for all types of gasoline are available. The one exception is outlet A, which only provided average monthly data on prices for premium and mid-grade gasoline. In contrast, retail sales volumes and wholesale prices are exclusively in monthly totals. While the average monthly wholesale price experienced by outlets B and A (independent retailers) are obviously based on purchases from vertically-integrated refiners, outlets C and D’s (vertically-integrated firms) average wholesale price is based on an internal transfer price. Specifically, it is obtained by taking the total value of sales by the firm to its affiliated retailer and dividing it by the number of litres to obtain a price per litre. Sample statistics on retail and wholesale prices per litre, excluding taxes, are contained in the tables below.

5. Economic Analysis

5.1 Differences between Outlets A and B (independent retailers)

Tables 3 and 4 consist of the sample statistics of retail and wholesale gasoline prices for outlets A and B (independent retailers) in cents per litre excluding taxes, respectively, over available data. However, for comparison purposes we shall restrict the outlet B data to the same time period (March 2003 – December 2004) as the outlet A data. The summary statistics for outlet B over this time period are contained in Table 5.

The mean sample retail prices (Tables 3 and 5) between outlets A and B are quite similar with respect to regular grade gasoline, being only half a cent apart. The gaps between other grades of gasoline are a bit larger; specifically, outlet B charges approximately two and three cents more, on average, for mid-grade and premium, respectively. The minimum and maximum values for all grades of gasoline are correspondingly higher for outlet B data over the sample period.

However, sample mean wholesale prices across all grades of gasoline are interestingly enough, quite similar between the two stations with minimum as well as maximum values also being close. It is also important to note that sample mean retail and wholesale prices are lower than city averages (Table 11) provided by M.J. Ervin.

The key difference between the two stations seems to be in terms of the mean volume of sales. In order to preserve confidentiality, we cannot report the mean summary statistics. However, we note that using the same time period (March 2003 – December 2004) as with respect to retail and wholesale prices, retail sales volumes at outlet A exceeds twice the amount of corresponding sales at outlet B with respect to all grades of gasoline. This is comforting, as coupled with the lower prices charged by outlet A relative to outlet B, these facts clearly imply a downward sloping demand curve. Of course, there are several caveats associated with this observation, given that they are not even in the same area. But it is nonetheless, reassuring.

The lower price charged by outlet A could also be a function of the fact that it faces five stations within a one kilometer radius as opposed to the four outlets B competes with in a similar radius. Of course, one must be careful in placing undue emphasis on this, as the implication is that an additional station has a considerable marginal impact in terms of prices.       

What does this all mean in terms of revenues and profits? While specific numbers cannot be reported, we note that mean monthly revenues from regular grade gasoline are much higher for outlet A than outlet B. These revenues are roughly 80% of all gasoline sales. Outlet A also earns higher profits from regular grade gasoline sales than outlet B. These profit figures are basically derived from a per litre profit margin (retail price – wholesale price) of 2.57 cents for outlet A and 3.14 cents for outlet B from March 2003 to December 2004. It is interesting to note that the comparable spreads with respect to mid-grade and premium gasoline are much higher for both stations. Specifically, the per litre profit margin experienced by outlet A for mid-grade gasoline is roughly 4 cents and 5 cents for premium gasoline. The corresponding statistics for outlet B are 6 and 9 cents, respectively.² These figures correlate with corresponding Toronto city averages found in Table 11.

5.2 Differences between Outlets C and  D (vertically-integrated firms)    

It is now important to contrast the above results with corresponding figures from the outlets C and D between the March 2003 and December 2004 period. Unsurprisingly, relevant means from Tables 6 and 7 demonstrate that outlet C prices are roughly two cents higher across all grades of gasoline on a per litre basis, relative to prices charged by outlets B and A. These results hold even over the January 2002 – December 2004 time period. One possibility for this is product differentiation; consumers might have a preference for gasoline sold by vertically-integrated firms possibly because of a belief that the gas is cleaner or more reliable than gas sold by independents.3 Another possibility is vertically-integrated firms can product differentiate themselves on the basis of a greater variety of product offerings, relative to smaller independents. A final reason may be the fact that outlet C only faces one other competitor within a one-kilometer circle.

On the other hand, corresponding prices charged by outlet D are lower than C’s. And what is compelling about this result is that it faces more competitors, all of whom happen to be vertically-integrated firms.

However, what is even more intriguing is the fact that outlet C’s wholesale price is remarkably similar to outlet B’s (independent retailer), irrespective of whether we study the Jan 2002 – Dec 2004 or March 2003 – Dec 2004 time period. As a consequence, it is unsurprising to note that outlet C earns significantly more in profits on a per litre basis than outlets B and A (the independent retailers). For example, during the March 2003 – December 2004 sample period, outlet C earned roughly 4.5, 8.5, and 11 cents per litre on regular, mid-grade, and premium gasoline. While outlet D experiences lower wholesale prices, its profit margins are very similar (Tables 8 and 9). As noted above, the corresponding figures for outlet B were 3,6, and 9 cents per litre, while outlet A numbers were 2.5, 4, and 5 cents per litre.

Not only does outlet C enjoy a higher profit margin on a per litre basis but it also has a higher throughput for all grades of gasoline than outlet B and outlet A. Corresponding volumes sold by outlet D are lower than outlet C and higher than outlet B. While outlet D’s regular grade sales are lower than outlet A, it has higher sales in other grades. As a result it is intuitive that outlet C earns much higher average monthly profits than outlets B, D, and A. Another interesting difference is that both outlets C and D (the vertically-integrated firms) earn a lower (higher) percentage of their profits from regular (premium) grade gasoline than the other stations.

5.3 Implications

In summary, our analysis suggests the following;

1. There is a difference in mean retail prices charged by the stations. Outlet C’s price is the highest, followed by outlets B, D, and A.
2. The difference in retail prices between stations increases with the grade of gasoline. For example, although outlet B prices for regular grade gasoline are higher than outlet A, the gap in prices for premium grade gasoline between the two stations is even higher. A similar differential exists between prices charged by outlet B, outlet D, and outlet C. These differences should be expected to the extent that the demand for higher grades of gasoline are more price inelastic than the demand for regular grade gasoline.
3. These prices seem to be accurately measured, as they are similar to city average prices collected by M.J. Ervin.
4. The degree of local competition seems to be a factor as there is a correlation between the number of stations within a one-kilometer radius and retail prices charged by each station. For example, outlet C consistently charges the highest prices and faces just one competitor within a one-kilometer radius. On the other hand, outlet D charged lower retail prices and competes with more stations. However, one must be very cautious in placing a strong emphasis on this observation; the higher retail prices charged by outlet C could very well be the result of other unobserved confounded factors that are impossible to control for with such a limited dataset.
5. On the other hand, average wholesale prices experienced by the different stations are strikingly similar across all grades of gasoline. This observation offers counter evidence to allegations that independent retailers are forced to purchase gasoline at wholesale prices that are significantly different than those enjoyed by affiliates of vertically-integrated firms. However, it is also important to acknowledge that generalizing this result for all independents may not be entirely accurate, given the likely stronger bargaining power of outlet A and outlet B in this regard, relative to other smaller independents.
6. Outlet C enjoys higher average margins for all grades of gasoline followed by outlet D. However, retail margins for regular grade gasoline are not that dissimilar between vertically integrated refiners and independents. Specifically, net revenue figures suggest that the independents obtained a per litre profit of between 2.57-3.14 cents from the sale of regular grade gasoline while vertically-integrated firms earned from 2.5-3.5 cents/litre over a similar sample period.
7. The similarity in wholesale prices and small discrepancy in retail prices suggests that gasoline firms are essentially price-takers – wholesale prices closely track corresponding movements in crude oil prices, while differences in retail prices are probably dictated to a limited extent by factors such as local competition.
8. Differences in average revenue and profits are then largely dictated by throughput, implying that choosing an appropriate location is of paramount significance. Apart from the number of other local competitors, the amount of average throughput will obviously be a function of population density and the amount of local traffic.
9. In this respect, it is important to note that outlet C has the highest volume of sales for all grades of gasoline.
10. Outlets A and B (the independent retailers) obtain a significant portion of their gas revenue and profits from regular grade sales. While sales from regular grade gasoline does constitute a majority of the revenues and profits earned by outlets C and D, a nontrivial portion also comes from the sale of higher grades of gasoline, relative to corresponding returns obtained by outlets A and B.

5.4 Time Series Analysis

The above conclusions are predicated on comparisons of sample means of retail and wholesale prices over time and across stations. However, relying on sample means might lead to erroneous conclusions if trends in prices are extremely volatile with large deviations around the sample mean. In order to test this possibility we present charts of time-series trends in each of the above variables. Figures 1, 2, and 3 present time-series variation with respect to regular, mid-grade, and premium retail prices, respectively. Similarly, Figures 4, 5, and 6 contain time-series trends for regular, mid-grade, and premium wholesale prices, respectively. Figures 7, 8, 9 contain corresponding trends for retail margins with respect to regular, mid-grade, and premium retail prices, respectively.

Over most of the sample period, the time-series trends in regular retail prices (Figure 1) closely follow the order given by the sample means; in other words, outlet C prices are the highest, followed by outlet B, outlet D, and then outlet A. But it is interesting to note that this order is less pronounced toward the end of the sample. On the other hand, the order is definitely quite clearly visible throughout the time period for the mid-grade (Figure 2) and premium grades (Figure 3).

In contrast, time-series trends in wholesale prices across all grades of gasoline (Figures 4, 5, and 6) clearly demonstrate the similarity in prices experienced by all stations. Corresponding movements in margins across different grades (Figures 7-9) present no surprises. On an average basis, margins enjoyed by outlet C are the highest, but there are instances where their margins are exceeded by corresponding margins from outlet B and outlet A, especially towards the end of the sample period. However, outlet C margins are consistently higher with respect to superior grades of gasoline.

Figure 1 – Average Monthly Retail Prices for Regular Gasoline

Figure 2 – Average Monthly Retail Prices for Mid-grade Gasoline

Figure 3 – Average Monthly Retail Prices for Premium Gasoline

Figure 4 – Average Monthly Wholesale Prices for Regular Gasoline

Figure 5 – Average Monthly Wholesale Prices for Mid-grade Gasoline

Figure 6 – Average Monthly Wholesale Prices for Premium Gasoline

Figure 7– Average Monthly Profit Margin for Regular Gasoline

Figure 8 – Average Monthly Profit Margin for Mid-Grade Gasoline

Figure 9 – Average Monthly Profit Margin for Premium Gasoline

5.5 Econometric Analysis     

While the above analyses are interesting, they may simply reflect the effects of other unobserved factors that also impact movements in retail and wholesale prices over time. In order to understand whether vertically-integrated firms and independents behave similarly, we propose the following straightforward econometric exercises. Specifically, we estimate the: (1) the impact of changes in wholesale prices on retail prices on an individual station basis; and (2) the impact of changes in crude oil prices on wholesale prices on an individual station basis. Similar estimates across stations would imply that vertically-integrated and smaller firms react similarly to crude oil price shocks and thus add greater insight on the profit maximizing behavior of different types of firms.

The econometric specification we employ is quite simple. The dependent variable is the retail price of a specific grade of gasoline. The key covariates are the corresponding current average monthly wholesale price and the previous month’s average monthly wholesale price for the particular grade of gasoline, purchased by the station. We also control for time shocks by using month specific fixed effects. Of course, the estimation methodology is Ordinary Least Squares (OLS).⁴  Further, two different types of specifications are employed in order to evaluate the sensitivity of our results; a levels as well as a log-log model.⁵ 

Table 12 consists of estimates of the impact of current and lagged wholesale prices on current retail prices for regular grade gasoline.⁶ The regressions with pooled outlet A/outlet B data consists of a station fixed effect to distinguish between the two.⁷ The first relevant result is the similarity in coefficient estimates across all columns. The levels model from column (1) implies that controlling for other factors, a 1 cent/litre increase in wholesale prices is significantly correlated with a 0.87 cent/litre increase in outlet A/outlet B retail prices. The coefficient estimate from the corresponding levels specification using outlets C and D data is remarkably similar; specifically it implies that a 1 cent/litre increase in wholesale prices is significantly correlated with a 0.87 (0.807) cent/litre increase in retail prices charged by outlet C (outlet D). Corresponding estimates from log-log models are also similar. While using outlet A/outlet B data implies that a 1% increase is wholesale prices is associated with a 0.822% increase in retail prices, outlet C (outlet D) data suggests that a 1% increase in wholesale prices is correlated with a 0.79% (0.73%) increase in its retail prices. Further, the R2 is very high across all specifications, implying that the variation in retail prices at the station level are almost exclusively explained by movements in wholesale prices.

Tables 13 and 14 contain similar estimates of the impact of mid-grade and premium wholesale prices on mid-grade and premium retail. Again, it can be seen that there is virtually no difference in coefficient estimates of changes in wholesale prices on corresponding movements in retail prices. Empirical estimates from outlet A/outlet B, outlet C, and outlet D data yield very similar results. As in Table 12, the R2 across empirical specifications in Tables 13 and 14 is very high.

What are the implications of these findings? Basically, changes in wholesale prices predominantly explain variation in retail prices charged by all firms in the sample. Further, all these firms respond quite similarly to changes in wholesale prices.

Tables 15, 16, and 17 contain similar estimates of the impact of crude oil price shocks on regular, mid-grade and premium wholesale prices, respectively. Mirroring previous findings, changes in wholesale prices to crude oil price shocks are extremely similar across stations, irrespective of whether we employ log-log or levels models. One difference is that sum of coefficient estimates of crude oil prices is slightly less than one, as opposed to being equal to or slightly greater than one as found by most other studies. These results are probably due to the aggregated nature of data used by other studies.

These results yield some rather firm and credible evidence against the existence of “raising the rivals costs” and instead demonstrate that wholesale prices experienced by vertically-integrated firms and smaller independents follow similar patterns. 

Table 12 - Regular Grade Gasoline
(Dependent Variable – Retail Price of Regular Grade Gasoline)

Table 13 - Mid Grade Gasoline
(Dependent Variable – Retail Price of Mid Grade Gasoline

Table 14 - Premium Grade Gasoline
(Dependent Variable – Retail Price of Premium Grade Gasoline)

Table 15 - Regular Wholesale Gasoline
(Dependant Variable - Wholesale Price of Regular Gasoline)

Table 16 - Mid-Grade Wholesale Gasoline
(Dependant Variable -Wholesale Price of Mid-Grade Gasoline)

Table 17 - Premium Wholesale Gasoline
(Dependant Variable -Wholesale Price of Premium Gasoline)

---

Footnotes

1 Outlet E is also owned by a vertically-integrated firm; however, we are only able to use its data in the profitability analysis due to some missing information.

2 These figures are from Tables 3 and 5.

3 This is of course, despite the fact that the gasoline sold by the independent firms is obtained from vertically-integrated firms.

4 Coefficient estimates are Newey-West corrected for second order autocorrelation.

5 In a levels model, a coefficient estimate is interpreted as the change in “y” associated with a unit change in “x” holding everything else constant. On the other hand, in a log-log model a coefficient estimate is interpreted as the % change in “y” associated with a 1% change in “x” holding everything else constant.
 
6 Standard errors are in parentheses. *, **, *** refers to statistical significance at the 1%, 5%, and 10% levels of significance.

7 We pooled the Outlet B and Outlet A data together, as the Outlet A data by themselves consist of only 16 observations, which is too small to conduct any credible regression analyses.