retail price asymmetries in local gasoline markets

12
ELSEVIER Energy Economics18 (1996) 81-92 Energy Economics Retail price asymmetries in local gasoline markets Kevin T. Duffy-Deno office of Energy and Resource Planning~ Utah Department of Natural Resources, 3 Triad Center, Suite 450, Salt Lake City, UT 84180, USA Abstract The experience of the Salt Lake City retail gasoline market during the 1989 to 1993 period suggests that there was asymmetric adjustment. Wholesale and retail price data, unique to the Salt Lake market, are used to test for the existence of three types of price asymmetry during this period. With respect to the magnitude of the pass through of wholesale price changes, evidence of price asymmetry during periods of average weekly wholesale price changes is found. However, during market shocks, the evidence points in the direction of retail price symmetry. Evidence of asymmetry with respect to the timing and pattern of the retail price response is found for periods of both average and market shock changes. Since gasoline is purchased by the consumer in local markets, conclusions reached by studies of market pricing behaviour that rely on national data may not be relevant for local policy decisions. This appears to be the case for the Salt Lake City market. JEL codes: Q40; D40 Keywords: Gasoline prices; Speed of adjustment; Asymmetric adjustment 1. Introduction The typical consumer of gasoline in the USA appears to believe that retail gasoline markets are characterized by price asymmetries. In response to higher crude oil prices and, hence, higher wholesale prices, retail (pump) prices rise quickly and completely. But when wholesale prices decline, retail prices fall at a much slower rate and may not fully adjust to the lower wholesale price environ- ment. This perception has lead many an oil industry executive to defend his pricing strategies before US congressional inquiries. Elsevier Science B.V. SSDI 0140-9883(95)0005 1-8

Upload: kevin-t-duffy-deno

Post on 21-Jun-2016

212 views

Category:

Documents


0 download

TRANSCRIPT

ELSEVIER Energy Economics 18 (1996) 81-92

Energy Economics

Retail price asymmetries in local gasoline markets

Kevin T. D u f f y - D e n o

office of Energy and Resource Planning~ Utah Department of Natural Resources, 3 Triad Center, Suite 450, Salt Lake City, UT 84180, USA

Abstract

The experience of the Salt Lake City retail gasoline market during the 1989 to 1993 period suggests that there was asymmetric adjustment. Wholesale and retail price data, unique to the Salt Lake market, are used to test for the existence of three types of price asymmetry during this period. With respect to the magnitude of the pass through of wholesale price changes, evidence of price asymmetry during periods of average weekly wholesale price changes is found. However, during market shocks, the evidence points in the direction of retail price symmetry. Evidence of asymmetry with respect to the timing and pattern of the retail price response is found for periods of both average and market shock changes. Since gasoline is purchased by the consumer in local markets, conclusions reached by studies of market pricing behaviour that rely on national data may not be relevant for local policy decisions. This appears to be the case for the Salt Lake City market.

JEL codes: Q40; D40

Keywords: Gasoline prices; Speed of adjustment; Asymmetric adjustment

1. Introduct ion

The typical consumer of gasoline in the USA appears to believe that retail gasoline markets are characterized by price asymmetries. In response to higher crude oil prices and, hence, higher wholesale prices, retail (pump) prices rise quickly and completely. But when wholesale prices decline, retail prices fall at a much slower rate and may not fully adjust to the lower wholesale price environ- ment. This perception has lead many an oil industry executive to defend his pricing strategies before US congressional inquiries.

Elsevier Science B.V. S S D I 0140-9883(95)0005 1-8

82 K T. Duffy-Deno /Energy Economics 18 (1996)81-92

Academic and government studies have attempted to validate this perception of price asymmetry through econometric analyses of the US gasoline market. The findings of these studies have been mixed. Since producers rarely sell gasoline directly to retail outlets, studies have examined whether asymmetries exist in the linkage between crude oil and wholesale prices and in the linkage between wholesale and retail prices. With respect to the linkage between crude oil and wholesale prices, Borenstein et al [2] find evidence of both symmetry and asymme- try, depending upon whether wholesale prices are measured by spot prices for generic gasoline or by terminal prices. Although US GAO [9] could not reject the hypothesis of symmetry for average crude oil price changes, they could reject the hypothesis during periods of market shocks. They found that a 10 cents per gallon increase in the spot price of crude oil during a market shock led to a cumulative 10 cents per gallon increase in the spot wholesale price of gasoline. However, a 10 cents per gallon decrease in the crude spot price lead to a cumulative 8 cents per gallon decrease in the spot wholesale price.

Several studies have found evidence of asymmetry in the linkage between wholesale and retail prices. Karrenbrock [4], using US monthly data, found that 69% of an increase in the average wholesale price is passed on to consumers within the first month, while only 22 to 32% of a wholesale price decrease is passed on during the first month. He could not, however, reject the hypothesis that the cumulative impact of a wholesale price increase and decrease on retail prices was the same. Borenstein et al [2] found that retail prices adjust more quickly to wholesale (terminal) price increases than decreases using weekly data. GAO examined the relationship between wholesale spot prices at the New York harbour and a national average of unleaded gasoline prices. They could not reject the hypothesis of price symmetry between wholesale and retail prices for both average wholesale price changes and those that occur during market shocks. Using data similar to GAO, Norman and Shin [6] also could not reject the hypothesis of price symmetry between retail gasoline and both wholesale gasoline and crude oil prices using a non-linear adjustment model. The quadratic adjustment model used by Norman and Shin was also used by Bacon [1] in a study of the UK gasoline market. Bacon found that retail adjustment to a wholesale price increase took, on average, 8.75 weeks while the adjustment to a wholesale price decrease took, on average, 9.54 weeks. He further found evidence of the rise in retail prices being sharper and less gradual than declines in retail prices.

A summary of the literature on asymmetries in the gasoline market is presented in Table 1. Findings on three types of asymmetry are shown: (i) the magnitude of the wholesale price change pass-through; (ii) the average length of the adjustment length; and (iii) the pattern of the adjustment (i.e., does all of the adjustment occur immediately or is it spread out over the adjustment period.).

Although these studies differ in the price series employed, frequency of data, historical time period examined, and model specification, they are all similar in that they base their findings on some measure of national prices. An argument can be made, however, that the findings based on such data may not be applicable to smaller, local markets. And it is the local market in which gasoline is purchased by

K. T. Duffy-Deno / Energy Economics 18 (1996) 81-92 83

the typical consumer. This study uses unique data for a specific local gasoline market, Salt Lake City, Utah, to test for the existence of retail price asymmetries. 1

The use of national prices glosses over the fact that there are distinct and important differences between local retail gasoline markets that can affect the transmission of wholesale price changes to the pump. Foremost, the complexity of the distribution network can differ between local markets. The number of distribu- tors, the number and type of ownership of retail outlets, historical contract relationships between distributors and dealers and between suppliers (e.g., refiner- ies) and distributors, and state laws affecting motor fuel marketing, are all factors that can affect how wholesale price changes are ultimately reflected in pump prices. In addition, local markets differ in the number of wholesale suppliers as well as in whether refineries are located in the region. For example, there are 12 suppliers of wholesale gasoline to the Salt Lake market, 5 of whom have refineries located in the Salt Lake area.: Pricing behaviour in this market may be quite different as compared to a local market in which supply-side competition is more limited and/or wholesale supplies must be either entirely trucked or piped in from distant sources. Moreover, these differences suggest that pricing behaviour in any given local market may be quite different from that inferred using US data which averages out these regional differences. 3

The findings of the current study suggest that for average weekly changes in wholesale prices during the 1989 to 1993 period, the Salt Lake City retail gasoline market was characterized by price asymmetries. During periods of market shocks, however, the evidence points more toward price symmetry with respect to the cumulative impact of a wholesale price change on the average, after tax retail price. For both average weekly changes and those that occur during market shocks, asymmetries with respect to the timing and pattern of the adjustment in the average net retail price are found to exist. The remainder of this paper presents the estimation methodology and data sample and explains these findings in greater detail.

2. Methodology

The methodology employed to test the hypothesis of price symmetry in the Salt Lake City retail gasoline market follows that of Karrenbrock [4] and Wolfram

1The current study examines only the linkage between wholesale and retail prices in the Salt Lake City market because of the unavailability of an accurate refinery acquisition cost time series for Salt Lake City's refineries. 2The average number of suppliers in PAD 4 (Rocky Mountain) reported by the Oil Price Information Seroice is 7. 3An advantage to using a local market sample to test for price asymmetries is that there is a much closer match between wholesale and retail prices t h a n i s the case when national data are used. This is particularly true in the current study because of the use of wholesale price data obtained from survey's of Salt Lake City's suppliers. Matching these city specific wholesale and retail data should increase the likelihood of obtaining an accurate portrayal of the dynamics in the retail gasoline market.

84 /~ T. Duffy-Deno / Energy Economics 18 (1996) 81-92

Table 1 Summary of literature reviewa

Sample

Study Level Frequency Model

Crude to Wholesale Borenstein et al [2] National Weekly G A O [9] National Weekly

Borenstein et al [2] GAO [9]

Magnitude of pass through

1. Average CR ( + , - ) : 83% 2 Shock CR ( + ): 100%

Shock CR ( - ): 80%

Timing (lag length) CR ( + , - ): 3 weeks Average and shock CR ( + , - ): 1 week

Linear error correction Linear adjustment

Asymmetry Pattern No difference All adjustment occurs within 1 week

Wholesale to retail Bacon [1] Ci ty/nat ional

Karrenbrock [4] National

Norman and Shin [6] National

Borenstein et al [2] National GA O [9] National

Bacon [1]

Karrenbrock [4]

Norman and Shin [6]

Borenstein et al [2]

Magnitude of pass through WH ( + , - ) : 100%

WH ( + , - ): 100%

1. (monthly) WH ( + , - ): 90%

2. (weekly) WH ( + , - ) : 85%

Bi weekly

Monthly

Monthly/weekly

Weekly Weekly

Timing (lag length) WH ( + ): 8.75 weeks

WH ( - ): 9.54 weeks

WH ( + , - ): 2 months

W H ( + , - ) : < 100%

GAO [9] WH ( + , - ): 100 %

WH ( + , - ) : 3 weeks

WH ( + , - ) : 4 months

Non-linear (quadratic) adjustment Linear adjustment Non-linear (quadratic) adjustment Linear error correction Linear adjustment

Asymmetry Pattern R ( + ) sharper and less gradual than R ( - )

WH ( + ): 67 % in first month WH ( - ): 33 % in first month

1 (Monthly) WH ( + , - ): 67 % first mt 2. (Weekly) WH ( + , - ): 14 % first wk

WH ( + ): 90 % in first 2 weeks WH ( - ): 30 % in first 2 weeks

WH ( + , - ): 50 % in first 4 weeks

a CR: Crude Price; WH: wholesale price; R: retail price.

K. T. Duffy-Deno / Energy Economics 18 (1996)81-92 85

(1971). The methodology is based on a markup model of retail gasoline pricing:

N R , = a o + a l W , (1)

wher N R t and W, are the net retail (after tax) and wholesale price of gasoline in cents per gallon. The use of net retail price removes any confounding effects caused by changes in state a n d / o r federal excise taxes over time. The effect of a change in the wholesale price on the net retail price is

N R t - N R , _ 1 = al(W t - Wt_ , ) (2)

Distinguishing between positive and negative price changes is required in order to uncover whether there is a differential effect on net-retail price between a wholesale price increase and a wholesale price decrease. This distinction can be made by using

A N R t = b l W I t + b 2 W D t + e t (3)

where

A N R t = R t - R t _ 1

WI,=W,-W,_,

W O t ~ . W t - Wt_ 1

e t = error term

i f ( W t - W t _ J) > 0

if(W, - Wt_ 1) < 0

and = 0 otherwise

and = 0 otherwise

Finally, a more general specification that allows for lags in adjustment is

s

A N R t = b o + ~ _ ~ b l , i W l t _ i + Y'~ b 2 , j W D t _ j + e, (4) i=1 j = l

where r and s are the number of lags for wholesale price increases and decreases. Following Karrenbrock, the general specification allows for investigation of three

types of price symmetry. The first type, magnitude symmetry, is whether the total, or cumulative, effect of a wholesale price increase on the net retail price is the same as that of a wholesale price decreases. Testing the hypothesis that they are the same requires testing whether or not Ebl, i = Eb2, j. The second type, timing symmetry, is whether the lag lengths are the same for wholesale price increases and decreases (i.e., whether r = s). The third type of price symmetry can be referred to as pattern symmetry. The hypothesis of such symmetry cannot be rejected if r = s and each bl, i = b2, j4

4Norman and Shin [6] employ a non-linear adjustment model in their study of US gasoline markets, the same model used by Bacon [1] to analyse the UK gasoline market. This methodology does not allow for tests for pattern or timing symmetry but was still used to analyse the Salt Lake City data. Although estimation convergence could not be achieved, the non-convergence findings were suggestive of asymmetry.

86 K T. Duffy-Deno / Energy Economics 18 (1996) 81-92

3. Data

The estimation of Equation (4) is based on a data sample consisting of weekly, average Salt Lake City prices for regular unleaded gasoline during the 1989 to 1993 period. 5 Wholesale prices are reported by the subscription Oil Price Information Service (OPIS). These prices are obtained through a weekly survey of Salt Lake City's motor fuel distribution terminals conducted by OPIS. Currently there are 12 major suppliers of gasoline to the Salt Lake market, 5 of whom have refineries located in the immediate region. Wholesale prices used in this study are the average of both unbranded and branded regular unleaded rack prices as well as of all of the terminal suppliers in Salt Lake City. 6 Although quoted rack prices may be higher than actual prices paid by reseUers as a result of discounts, they generally are closer to the actual prices paid in the local market than are spot prices. Rack prices reflect local market conditions as well as premia for supply certainty and price stability relative to spot prices.

Average retail prices are reported by the Oil and Gas Journal (OGJ). OGJ contracted for a weekly survey of unleaded regular, self-serve pump prices in major US cities, including Salt Lake City, during the 1989 to 1993 period. Definitionally, the Salt Lake City market consists of Salt Lake County. Net retail prices are the surveyed average pump price less state and federal excise taxes (currently 19 and 18.4 cents).

The sample size was constrained by two factors. First, the Office of Energy and Resource Planning did not subscribe to OPIS prior to 1989. Secondly, OGJ switched to conducting the retail price survey in-house starting in January, 1994. Discussions with OGJ suggests that the reliability of the data for Salt Lake City diminished appreciably as a result of this change.

Fig. 1 presents these weekly average wholesale and retail data for the 1989 to 1993 period. The 260 week sample shows a fair degree of variation in gasoline prices. The Exxon Valdez oil spill (early 1989) and the Persian Gulf Conflict (late 1990) both occurred during this period and contributed to this variation. The average retail price in the Salt Lake City market reached a high of US $1.29 per gallon during the Persian Gulf conflict in late 1990. The lowest average retail price during this five-year period was US $0.95 per gallon in early 1992.

Statistics on the week-to-week price changes for this sample are shown in Table 2. The table first shows, for both average wholesale and net retail prices, the number of occurrences of positive and negative week-to-week changes. The aver- age and maximum magnitudes of these changes are also shown. The largest week to week increases in wholesale and net retail prices occurred during the Persian Gulf conflict (US $0.142 and US $0.071). The largest week to week decrease in average wholesale prices (US $0.08) also occurred during this period but the largest

51t is important to use as small a data sampling interval as possible (i.e., smaller than the suspected length of the adjustment lag) to avoid or minimize the bias in the estimated adjustment speed (Maddala, [5]). The smallest data sampling interval available for the Salt Lake City market is weekly. 6Branded prices are paid by resellers who purchase gasoline from a supplier selling under its trademark name, while unbranded prices are paid for gasoline that is not sold under any trademark.

K T. Duffy-Deno/Energy Economics 18 (1996) 81-92 87

1.40

1.30

C t.20 0

m - - 1.10 O=

C.~ 1.00 I,,..

0.90 Q. 0.80

~) 0.70

0.60 > ,< o.so

0.40

0.30 1989 1990 1991 1992 1993

1 Wholesale 1 Net Retail 1 Retail I

Fig. 1. Weekly Salt Lake city regular unleaded gasoline prices

week to week decrease in average net retail prices occurred during early 1993 (US $0.116).

4. Findings

A free form lag specification was used to estimate Equation (4). Initially, arbitrarily long lags for both W/t and WD t were included. Working backwards, F-

Table 2 Weekly price changes: Salt Lake City Regular Unleaded Gasoline 1989 to 1993

Average wholesale Average net retail

Number of increases 133 126 Number of decreases 121 114 Number of no changes 5 19

Maximum increase (US$) 0.142 0.071 Maximum decrease (US$) 0.080 0.116

Average increase (US$) 0.012 0.012 Average decrease (US$) 0.013 0.013

88 K~ T. Duffy-Deno / Energy Economics 18 (1996) 81-92

and t-tests were used to test if any of the lags individually or as a group could be eliminated as statistically insignificant. The results of this procedure were verified by initially specifying a one week lag and then adding lags if singly or jointly significant. The results were also verified by more formal tests such as the Akaike information criteria on (Judge et al [3]). A 95% level of confidence was employed in all testing.

In addition, a polynomial distributed lag specification was also estimated but yielded estimates essentially identical to those discussed below. The estimated intercept accounts for other fixed factors that may affect the average retail price. First differences minimiTes problems associated with first-order autocorrelation as evidenced by the Durbin-Watson statistics.

4.1. Average wholesale price changes

The findings from the symmetry tests are presented in Table 3. This indicates that the Salt l a k e City retail market exhibited all three types of price asymmetry. Foremost, the net retail price took, on average, longer and adjusted more fully to a wholesale price increase than a decrease. Four and two weeks were the average adjustment lengths for a wholesale price increase and decrease. Moreover, the point estimate suggests that, on average, a 10 cents per gallon increase in the wholesale price ultimately lead to an 8 cents increase in the net retail price. However, the null hypothesis that the point estimate equals 1 cannot be rejected suggesting the adjustment may be close to complete. A 10 cents per gallon decrease in the wholesale price ultimately lead to, on average, a 4.6 cents decline in the net retail price. The fact that the null hypothesis that the point estimate equals 1 can be rejected suggests that the average net retail price did not adjust completely to wholesale price decreases during this time period. The magnitude of the asymmetry is verified by the rejection of the null hypothesis that Ebl, i = Eb2, i"

The pattern of the adjustment in the average net retail price also differs between an increase and decrease in the average wholesale price. Over half (53%) of the adjustment to a wholesale price increase occurs within the first week with the remaining 47% spread out over the following three weeks. The adjustment in the average net-retail price in response to a wholesale price decrease appears to be spread evenly across two weeks. 7

4.2. Market shocks

The nature of the adjustment of the average net retail price to changes in the average wholesale price may depend upon the degree of the wholesale price change. The average net retail price may adjust differently during periods when the average wholesale price is either rapidly rising or falling as compared with normal

7The sample period includes the sharp prices increases that occurred during the Persian Gulf Conflict in late 1990. A dummy variable was included, both separately and interactively, to account for this period but was never statistically significant.

K. T. Duffy-Deno / Energy Economics 18 (1996)81-92

Table 3 Symmetry findings for Salt Lake City market: average wholesale price changes

89

AW, > 0 AW, < 0

Timing Number of weeks lagged 4

Magnitude Cumulative price impact 0.798 t-value: F.b = 0 7.056 t-value: F.b = 1 1.786

t-value: ~bl , i = ~b2, )

Pattern Coefficient: t 0.425 Coefficient: t - 1 0.103 Coefficient: t - 2 0.081 Coefficient: t - 3 0.188

R 2

D W

2.023

0.30 2.04

0.455 4.574 5.487

0.209 0.246

or average periods. Some industry experts argue that in the face of rapidly rising wholesale prices, retailers will not increase their prices as much to avoid raising the ire of consumers and decreased sales. When wholesale prices begin to decline, retail prices decline at a slower rate as retailers attempt to recoup lower margins experienced when wholesale prices were increasing quickly (GAO [9]). Others, though, disagree and assert that there is no price asymmetry during periods of market shocks. GAO [9] could not reject the hypothesis of wholesale to retail price symmetry during periods of market shocks using national, weekly data from January 1984 to March 1991.

A market shock is defined, in this study, as an event that leads to a week to week change in the average wholesale price greater than a certain amount. For example, a 'mild' market shock can be defined as one which results in a percentage change in the average wholesale price greater than or, where o- is the standard deviation from the mean of the wholesale price percentage change distribution. Using this definition of a market shock results in a redefinition of W/t and WD t as follows:

W/t = Wt - Wt_ 1

= 0 otherwise

WD,- Wt- W~_1

= 0 otherwise

if ( W t - Wt_ 1) > 0

i f ( W t - Wt_ 1) < 0

and I%AW~I > or

and I%AWtl > or

Thus, W/t and WD t are defined only if the percentage change in W t is relatively large. A more extreme price shock is defined to occur when the %AW t > 2o- which

90 /~ T. Duffy-Deno / Energy Economics 18 (1996)81-92

Table 4 Symmetry findings for the Salt Lake City market: market shocks

I%AW, I > ~r I%AW~I > 2or

8 w, > o AW, < 0 AW, > 0 AW,<O

Timing Number of weeks lagged 4 3 4 3

Magnitude Cumulative price impact 0.893 0.635 1.096 0.765 t-value: Eb = 0 7.385 5.824 6.490 4.378 t-value: Eb = 1 0.881 3.350 0.567 1.342 t-value: ~1, i = ~b2, y 1.526 1.399

Pattern Coefficient: t 0.433 0.209 0.445 0.236 Coefficient: t - 1 0.102 0.154 0.216 0.257 Coefficient: t - 2 0.079 0.272 0.192 0.273 Coefficient: t - 3 0.280 0.243

R 2 0.30 0.21 DW 2.04 1.84

leads to

W~t : Wt - Wt_ 1

= 0 otherwise

WD,= W,- W,-1 = 0 otherwise

i f (W t - W t _ ~ ) > 0

i f (W t - Wt_ 1) < 0

and I%AW~I > 2¢r

and I%AW~I > 2or

The results of estimating Equation (4) for these two definitions of market shocks are shown in Table 4. 8

Comparing the results shown in Table 4 with each other and with those shown in Table 3 for average wholesale price changes yields some interesting findings. First, the average net retail price took one week longer to adjust to a wholesale price decrease during a market shock than under normal or average circumstances. On the other hand, adjustment to a wholesale price increase took four weeks during both market shocks and normal circumstances.

Second, the hypothesis that the cumulative impact on the average net-retail price of a wholesale price increase equals 1 during a market shock cannot be rejected. However, rejection of the same hypothesis for a wholesale price decrease depends upon how the market shock is defined. Using tr for the percentage change in W t

8GAO (1993) included a wholesale gasoline stock variable in their estimating equation to help explain why there may be asymmetric responses to wholesale price changes induced by market shocks. Monthly data on gasoline stocks are provided by Salt Lake City's refineries. However, a weekly stock variable created through interpolation was never statistically significant.

K~ T. Duffy-Deno /Energy Economics 18 (1996) 81-92 91

yields conflicting findings. The hypothesis that Y'.b = 1 for a wholesale price decrease can be rejected. However, the hypothesis that Ebl, i = Y'-b2, j cannot be rejected. Using 2o- for the percentage change in W t, though, yields consistent findings. The hypothesis that the cumulative impact of a wholesale price change on the average net retail price equals 1 cannot be rejected for both a wholesale price increase and decrease. Moreover, the hypothesis that Ebl, i = Eb2, j also cannot be rejected. The evidence appears to point in the direction of magnitude symmetry during markets shocks. 9

Finally, the pattern of the adjustment process during market shocks is similar to what occurred under normal or average circumstances. That is, in the case of a wholesale price increase, a large portion of the cumulative increase in the average net retail price occurred within one week with the remaining adjustment being spread out fairly evenly over the remaining three weeks. In the case of a wholesale price decrease, the adjustment in the average net-retail price was spread out evenly over the entire three-week adjustment period.

5. Summary

The experience of the Salt Lake City, Utah, retail gasoline market during the 1989 to 1993 period appears to support the perception that asymmetric adjustment in retail gasoline prices occurs. During periods of average wholesale price changes, a 10 cents per gallon increase in the average, weekly, Salt Lake City wholesale price of unleaded gasoline lead to a cumulative 8 cents per gallon rise in the average net retail price. On the other hand, a 10 cents per gallon wholesale price decrease lead to only a 4.6 cents decline in the average net retail price. However, the evidence suggests that during periods of market shocks, periods characterized by relatively large changes in the average wholesale price, the hypothesis of price symmetry with respect to the amount of the pass through of the wholesale price change cannot be rejected.

Consistent evidence of timing and pattern asymmetry was found during this time period. The average net retail price took four weeks to adjust to a wholesale price increase but only two to three weeks to adjust to a wholesale price decrease. Moreover, a large portion (at least 40%) of the cumulative increase in the average net retail price occurred within one week of a wholesale price increase with the remaining adjustment spread out fairly evenly over the remaining three weeks. In the case of a wholesale price decrease, the adjustment was spread out evenly over the entire two- to three-week adjustment period.

Just as it is difficult to draw inferences on price asymmetry for local markets from national data, it is difficult to draw inferences on pricing behaviour in other local gasoline markets based on the Salt Lake City experience. Local gasoline

9These findings should be considered only as suggestive in light of the relatively weaker explanatory power of the model when using 2~r for the percentage change in W t.

92 K. T. Duffy-Deno / Energy Economics 18 (1996) 81-92

markets differ in several ways that may affect the transmission of wholesale price changes to the pump price. An extension of the work reported on in this paper would be to test the hypothesis of price symmetry in other local gasoline markets using similar market specific data, methodology, and time period. If a large enough sample of markets was studied, then a correlation between the symmetry findings and market characteristics may yield greater insights into the workings of local gasoline markets.

Acknowledgements

The views expressed in this paper are the author's and do not necessarily reflect those of the Utah Department of Natural Resources or the State of Utah.

References

[1] Bacon, Robert, 'Rockets and feathers: the asymmetric speed of adjustment of UK retail gasoline prices to cost changes' Energy Economics 1991 13 211-218

[2] Borenstein, A, Cameron, Colin and Gilbert, Richard Do Gasoline Prices Respond Asymmetrically to Crude Oil Price Changes? National Bureau of Economic Research, Working Paper No. 4138 (1992)

[3] Judge, George G, Hill, R Carter Griffiths, William E Lutkepohl Helmut and Lee, Tsoung-Chao Introduction to the Theory and Practice of Econometrics John Wiley and Sons, New York (1988).

[4] Karrenbrock, Jeffrey, 'The behavior of retail gasoline prices: symmetric of not?' Federal Reserve Bank of St. Louis Review 1991 73 19-29

[5] Maddala, G.S., Econometrics McGraw-Hill, New York (1977) [6] Norman, Donald and Shin, David Price Adjustment in Gasoline and Heating Oil Markets, American

Petroleum Institute, Research Study No. 060 (1991) [7] Oil and Gas Journal, (various issues) [8] Oil Price Information Service United Communications Group, Rockville, MD (various issues) [9] US General Accounting Office; Analysis of the Pricing of Crude Oil and Petroleum Products,'

GAO/RCED-91-17 (1993)