BP2: Oil Prices and SUV Sales in the US

The Effects of Oil Price Changes on Light Truck Sales in the US

Robert Avakian

Introduction

In late fall of 2014 US sales of vehicles in the light truck category increased considerably. This increase coincided with a sharp fall in oil prices and steady economic growth. Some news reports linked the drop in oil prices to the spike in SUV sales raising the question as to whether changes in the price of oil have an impact on the sales of large vehicles such as SUVs, trucks and vans.

Background

In the late summer of 2014 oil prices begun a drop that would take them from just over $100 per barrel in late July to a low of $43 by early January of 2015. By early November 2014 news reports began appearing that linked the drop in the price of oil to a ten year high in US sales of trucks and SUVs (The Guardian, Dec 2014). SUV brands ranging from the regular Ford Escape to the luxury Cadillac Escalade experienced sale jumps between 6 and 38% compared to the previous year. Surveys showed that fuel economy which had ranked No. 3 in 2013 on the list of reasons that people buy cars, in 2014 had dropped to No. 6 behind features like styling, brand preference, ride quality and reputation (LA Times, Nov. 2014).

At the same time however, low oil prices also coincided with a with a period of steady economic growth. Employment was back to pre-2009 levels resulting in increased disposable income, a more secure economic outlook and the possibility that oil prices were actually having a marginal or no effect at all on SUV sales.

Determining the actual effect of oil prices on vehicle sales could provide valuable insights for automakers planning marketing campaigns, climate change policymakers and regulators determining new fuel consumption and emissions standards and, behavioral economists measuring the effect of short term incentives in an environment with long-term uncertainty.
Research Design and Methodology

The study analyses the effects of oil prices on vehicle sales over a 30 year period using sales, economic and oil price data from 1985 to 2015. The time interval is chosen based on the historical availability of monthly crude oil price data. All vehicles sales data is seasonally adjusted and divided between the light truck category (SUVs, trucks and vans) and the light car category (compacts, sedans, crossovers). The study focuses specifically on the light truck category sales given their relatively low fuel efficiency and potentially higher susceptibility to high oil prices. Light cars sales are used to check whether oil prices affect the entire industry or just a specific category.

The economic variables used in the analysis include monthly data for GDP growth, GDP per capita and disposable income. The GDP growth rate and GDP per capita are used to measure the effect of economic trends. Disposable income is used as an indicator of the financial means of individuals which, can lower the susceptibility to high oil prices as well as play a major role in the choice of an SUV over a cheaper compact car. A summary statistics table is shown in the appendix.

The data is analyzed using time series Cochrane-Orcutt AR(1) regression that accounts for auto-correlation issues (see appendix for details). Because oil price (and economic indicator) changes are not believed to have an immediate effect but rather take some time to have an actual economic impact, vehicle sales are analyzed using a 1 and 3 month lead.These models assume that oil price changes in August actually impact sales in September or November rather than in August. From a behavioral economics point of view it assumes that people take 1-3 months to internalize the change in gasoline prices and determine whether to buy and what type of vehicle to buy. 

Analysis

As seen in figure 1 there appears to be some correlation as oil price drops or spikes seem to be followed by changes in light truck sales. However, oil prices are not exogenous and can be influenced by economic factors, like GDP, which can also affect vehicle sales.

Figure 1
 

The regression analysis results in table 1 provide a more detailed analysis of the effect that oil price has on vehicle sales when taking into account economic trends such as GDP growth. Models 1-5 measure the effect of oil prices on light truck sales while model 6 measures the same effect on the light cars category.

The results in models 3 and 5 suggest that on average oil prices do have an effect on light truck sales. An approximately $1 increase in the per barrel price of crude oil leads to a 1000 unit decrease in light truck sales. GDP growth on the other hand has a significant positive effect with an average increase of 25 thousand units per 1% of GDP growth. The analysis of model 6 which measured light car sales also shows a positive effect from GDP growth but no significant effect from oil price changes.

Table 1. Truck and Car Sales by Oil Prices between 1985-2015

	(1)	(2)	(3)	(4)	(5)	(6)
DV: Seasonally Adjusted US Light Truck Sales (in thousands)	Truck Sales with 3 Month Time Lead	Truck Sales with 1 Month Time Lead	Truck Sales with 3 Month Time Lead	Truck Sales with 3 Month Time Lead	Truck Sales with 3 Month Time Lead	Car Sales with 3 Month Time Lead
Price of WTI Crude Oil	-0.189	-1.004	-1.154*	-1.793	-1.222*	-1.803
	(0.366)	(1.914)	(2.214)	(1.152)	(2.113)	(1.201)
Disposable Income per Capita		0.040*	0.015	0.014	0.015	-0.016
		(2.000)	(0.752)	(0.705)	(0.776)	(1.141)
% GDP Growth Month-to-Month		1.824	25.991**	24.344*	25.065*	20.479*
		(0.191)	(2.739)	(2.428)	(2.542)	(1.983)
GDP Per Capita		-0.006	0.005	0.015	0.013	0.032**
		(0.649)	(0.644)	(0.732)	(0.664)	(2.624)
Square of Oil Price				0.004		0.010
				(0.398)		(1.070)
Time				-0.929	-0.857	-3.403***
				(0.473)	(0.435)	(3.848)
Oil over $100					1.226
					(0.083)
Constant	605.054***	-403.970	-45.023	-169.034	-178.301	691.381**
	(9.725)	(1.245)	(0.141)	(0.388)	(0.409)	(3.174)
Observations	360	360	360	360	360	360
R-squared	0.000	0.063	0.076	0.079	0.078	0.328

Note:*Significant at 5%; **significant at 1%; ***significant at 0.1%. * p < 0.05, ** p < 0.01, *** p < 0.001
Source: Bureau of Economic Analysis (BEA) and Energy Information Agency (EIA)

Conclusions

The results of the analysis suggest that oil prices on average do have an effect on light truck sales. The fact that there is no oil price effect on light car sales but a similar GDP growth effect seems to confirm the susceptibility of the light truck category to higher oil prices. It is possible that while light cars are purchased as a necessity whenever economic conditions allow it, SUVs and trucks are seen by some buyers as more of a luxury which requires more favorable economic as well as oil price conditions in order to justify the leap in category and price. 

A Bloomberg analysis suggests that "even in an environment of high gas prices, much of the growth in sales in recent years has come from...light trucks which were already up nearly 10 percent in the year through October, while cars were up a mere 1.4 percent (Bloomberg, Dec. 2014)." This interpretation can still be consistent with the above presented results as it suggests that GDP growth has a greater effect than oil prices on light truck sales but leaves room for oil prices to further influence the buyers final decision. This analysis would be consistent with the late 2014 spike of already growing light truck sales.

The above mentioned difference in pre-oil price drop sales between light trucks and cars (10% vs 1.4%) could be due to a latent demand that had been so far kept down by the effects of the financial crisis and unemployment. A further study focusing on controlling for the effects of economic crises could be conducted to further clarify the issue.

------------------------------------------------------------------------------------------------------ 

Appendix

Table 2. Summary Statistics from 1985-2015

	Mean	Std. Dev.	Min.	Max.
Seasonally Adjusted Light Truck Sales  (in Thousands)	563.6	150.35	305.	1049.1
Seasonally Adjusted Car Vehicle Sales (in Thousands)	687.97	123.55	372.2	1254.2
Price of WTI Crude Oil	42.2	30.53	11.3	133.9
Disposable Income per Capita	30616.12	4811.37	22675.8	37832
% GDP Growth Month-to-Month	0.66	0.58	-2.1	1.9
GDP Per Capita	35566.38	11443.96	0.6	55319
Observations	364

Source:  Bureau of Economic Analysis (BEA) and Energy Information Energy (EIA)

Trend Effects Concerns 

Given the variance in the values of the variables shown in figures 2,3 and 4 there is no concern that steadily increasing or decreasing values over long periods of time can create trending effects that give false indications of significant effect. In regression models 4, 5 and 6 the variable "Time" is used to account for any trend effects of the growth in the date variable. 

Figure 2.


Figure 3.


Figure 4.


Auto-correlation Correction

A Durbin-Watson test provided a p-value<0.001 which indicated considerable autocorrelation in the regression models. A Prais–Winsten first order auto-regressive estimation was used to correct for the effects of serial correlation.

Regression Models

Model 2 uses a 1 month lead for the dependent variable representing truck sales and is the only model where disposable income is significant. However, with a t-statistic of 2.000 the variable is barely significant and the model is dropped in further developed instead favoring the 3 month lead models.

Models 4 and 6 include a variable that has the squared values of oil prices. This variable is included to check whether rising oil prices have an parabolic or plateauing effect on car sales. The variable does not show significant effect in either model. However, its inclusion in model 4 results in the standard oil price variable becoming insignificant.

In model 5 the dummy variable "Oil over $100" is included to check whether prices over the $100 threshold have a significant effect on sales. The variable is not statistically significant.