Analysis of Gas Prices in Howard County, Maryland
A.3 Technical Discussion—Variables
Download 0.59 Mb.
|
- Navigate this page:
- Figure : Correlation Matrix for Retail and Rack Prices
- Figure : Correlation Matrix for Margin and Rack Price
- Expectations
- Figure : Example of Lagged Rack Variable
- Market Competition
- Figure : Gas Stations per Capita (1,000) by County, 2007–2011
- Reading Dummy Variables
A.3 Technical Discussion—VariablesOn the whole, crude oil prices or “rack prices” are the largest contributor to the fluctuation in the pricing for gasoline. RESI conducted a correlation analysis on wholesale and retail price to demonstrate the dependency of the two variables. The results of this analysis are reported in Figure 14. Figure : Correlation Matrix for Retail and Rack Prices
Sources: OPIS, RESI The correlation matrix in Figure 14 highlights the relationship between the explanatory variable “rack price” and the dependent variable “retail price,” between 2002 and 2012 as an example of the relationship. Since they exhibit a very strong positive relationship, the use of the variables within the regression could create biased results and therefore render any findings invalid. Although correlation does not always indicate that there is a direct causation between retail prices and rack prices, the correlation does indicate that they move in the same direction when changed. If rack prices increase, according to the correlation coefficient matrix above, there will likely be an observed similar upward movement in retail if all other things remain equal. Another factor to consider involves the structure of input prices for final goods in a retail setting. Traditionally, gas stations have no control over the wholesale price of gasoline. Factors that are within the retail station’s control include the number of pumps, pay-at-pump locations, entrances/exits, complementary goods, staff, and hours of operation.39 All of these items represent input costs that will take away from the overall profit margins of a retailer. Despite the homogenous good being gasoline, the differentiation among stations is witnessed through the physical characteristics of the station and the location. This differentiation allows for some price difference among stations within a region.40 Although the rack price is out of the retailer’s control, there do remain some factors of operation that alter gas prices that allow for retailers to dictate how much they budget or spend. RESI determined that using retail prices as the dependent variable may overlook some of those potential factors and create an omitted variable bias within the model. In past economic studies, researchers have attempted to avoid the problematic correlation by using the variable “margin.” Margins would be able to fully reflect the change associated with the rack price, while capturing the underlying costs associated with daily operations for the retailer.41 RESI assumes that the stations can set their own retail prices and, given this control, will use the variable “margin” defined as the difference between the retail price and the rack price less taxes and freight as the endogenous variable in the models. A retailer’s margin will help to capture some of the input costs RESI cannot observe easily and vary across station and region. One input price that is readily available for observation is rack prices and should be kept within the model to accurately show the retailer’s perceived notions of future earnings. To avoid multicollinearity within the model, RESI reviewed margin and rack prices for correlation. Figure 15 shows the correlation matrix for the margin and rack prices. Figure : Correlation Matrix for Margin and Rack Price
Sources: OPIS, RESI Figure 15 highlights that the variable “rack price” is moderately positively correlated with margin and can be used in the model. Ordinarily, economists will avoid using variables with correlations above 0.7 as this would create issues within the analysis that may lead to biased estimates. The above can be read as “rack prices account for approximately 31 percent of the movement in retailer margins.”42 The higher the correlation coefficient is, the more the movement of the dependent variable can be explained by that singular variable, thereby creating potential omitted variable bias and multicollinearity within a regression. Preliminary research also indicates that price movements in retail gas prices have been known to move asymmetrically with their corresponding rack prices, therefore making rack price a good indicator of expectations for current retail prices. RESI used rack prices to capture the potential change to margins given the potential increase to retailers for their good with all other factors of cost remaining equal. Margin will act as the markup necessary to maintain business given the increasing or decreasing costs of rack prices. Other input pricing variables including tax have been kept in the model to demonstrate the impact of those variables on retail prices. Taxes on gasoline purchases are collected by retailers and submitted during their business’s fiscal year. A potential change in the tax rate would be considered when determining retail markup by stations. Income Household income traditionally will impact both the supply and the demand for gasoline prices within a region. A higher average household income within a region may result in an equally higher retail gas price due to cost of living differentiation. In previous studies, income only became a statistically significant variable if firms within a region “are engaged in tacit collusion.”43 Eckert, et al, suggest that, given the localized region, if firms were actively pricing in a collective manner, a higher income in a given proximity may affect retail prices in that localized area.44 Changes in wages may also present a rising cost to employers. RESI attempts to capture the changing income within a region through a household income per capita change variable in the regression. As noted, despite firms having foresight about wage changes, wage contracts are often slow to change and therefore create at least a period lag between the proposed change and the real change in wages.45 The use of household income per capita in this model is a proxy variable for county wealth. The higher the wealth within the county is, the greater the potential for higher gas prices. Expectations Expectations are inherent within pricing models. In determining the pricing of a good, a retailer reviews their historical demand and costs as well as potential future costs of their inputs. In this model, RESI will attempt to capture the expectations of retailers through change variables and some lagged terms. Lagged variables in statistical regressions are explanatory variables that shift the first period’s observed variable to the next period. An example using the variable “rack” is given in Figure 16. Figure : Example of Lagged Rack Variable
Sources: OPIS, RESI Lagged variables are used primarily in forecasting models to determine the future variable change. Here, RESI will use the lagged-one period explanatory variable to predict current prices among retailers. For example, retailers in 2004 may consider the rack prices today, as well as 2002 and 2003, when determining future prices. Current prices give a baseline of where future rack prices might be in 2005, but the past prices allow for them to determine what the trend of rack prices has been. Under this assumption, RESI assumes retailers keep records on past rack prices or they are able to be easily found and referenced by retailers. In Section 3.1, RESI mentioned the use of a “rack price change” to show the retailer’s perception of future rack prices for their goods given the difference between current and past period prices. This variable will include the use of a lagged-one period explanatory variable (rack prices) to estimate the trend seen by retailers in the current period when reviewing past rack prices. Economic literature that explains asymmetric price movement and its association with retail and rack gasoline pricing is well documented.46 RESI assumes retailers have this knowledge when pricing their fuel and will price according to the previous period’s rack prices. The inclusion of the current tax and lagged-one period variable for taxes in Section 3.1 is done so given the difference between State fiscal year and calendar year. Adherence to tax legislation in a region does not always coordinate with a retailer’s fiscal year. For example, the revision to the Maryland Gas Tax will take place on July 1, 2013, which is in accordance with the State Fiscal Year.47 Retailers consider their fiscal years according to the Internal Revenue Service as a “consecutive 12 months ending on the last day of any month except December.”48 Therefore, a retailer’s projection of pricing in accordance with obtaining the largest margins may need to take into account the current tax base and the previous tax base to estimate the actual taxes for a fiscal year. RESI attempts to capture this using the variable “tax” and “tax lagged-one period.” Market Competition Gasoline is a homogenous good, and competition, or the number of competitors in a given region, will affect the gasoline retailer’s margins. A larger base of suppliers within a given region may produce pricing wars among retailers to keep prices higher or work toward lowering prices. Figure 17 shows the number of reporting stations by county from 2007 to 2011. Figure : Gas Stations per Capita (1,000) by County, 2007–2011 Sources: RESI, OPIS Howard County’s population is relatively similar in size to Harford County’s population, but Harford County’s number of stations per capita (in thousands) is higher than Howard County’s. Montgomery has the largest population of the observed eight counties but the fewest gas stations per capita (in thousands). Based on these observations, RESI set out to determine whether the restrictiveness of a municipality’s zoning had an influence on gas prices. The variable “lg_station1000” reflects the log of number of stations per 1,000 residents. This variable helps to determine the potential supply of competitors available within a given region and its effect on margin over time. RESI also reviewed the demographics of the area using a population variable to determine the potential base of consumers. Population increases are noted with potential increases of gas stations within a region historically, but this influx of new competitors can lead to a decrease in current existing retailers’ margins if they cannot change to compete.49 Usher and Evans find that, if the firms cannot alter to compete, they will ultimately leave the market.50 Population is used as a change variable between the current and past periods to represent a potential growth within a retailer’s region. The impact from the growth may not be instantaneous, but the observed growth may give incentive for new firms to move into the region. Reading Dummy Variables One can use the formula below to convert the impact multipliers reported in Figure 4 in Section 3.0: 
The β in the equation above represents the impact multiplier of the dummy variable from the analysis in Section 3.0. Directory: Media Media -> Applications to Drill Media -> Unicef voices of Youth Chat Theme: Children and aids nigeria and Zimbabwe 13 October 2006 Background Media -> Tsunami Terror Alert: Voices of Youth Media -> Biblical Eschatology Presentation by: D. Paul Beck May 4, 2016 Ground Rules Media -> Guide to completing the collection using the Omnibus system Media -> The milk carton kids Media -> Events Date and Location Media -> The Gilded Age: The First Generation of Historians by H. Wayne Morgan University of Oklahoma, April 18, 1997 Media -> Analysis of Law in the United Kingdom pertaining to Cross-Border Disaster Relief Prepared by: For the 30 June 2010 Foreword Media -> Cuba fieldcourse 2010 Download 0.59 Mb. Share with your friends:
|