Forecasting Automobile Sales: MRN 32,7 Syed Shahabuddin
Forecasting Automobile Sales: MRN 32,7 Syed Shahabuddin
Forecasting Automobile Sales: MRN 32,7 Syed Shahabuddin
www.emeraldinsight.com/0140-9174.htm
MRN
32,7
670
Abstract
Purpose The purpose of this paper is to understand the behavior of the automotive industry
which is very critical to avoid major economic disruptions in the economy. To understand this
industry, one needs to understand its historical performance in relation to many economic factors that
may affect the industry.
Design/methodology/approach Data about automobile sales (in dollars and in units) and many
economic and demographic variables are collected from a variety of sources. Automobile sales are
the dependent variable. However, the variable of automobile sales is divided into foreign and
domestic car makers. The data are regressed using Statistical Package for the Social Sciences (SPSS)
stepwise regression to obtain highly correlated variables.
Findings The results indicate a strong relationship between the economic variables and foreign
car sales, but the relationship between the economic variables and domestic car sales is weak. The
domestic cars sales relationship to the other economic variables should be explored further to
determine possible causes for the weak correlation. One of the possible reasons could be that
domestic car makers use many incentives to influence sales, but data on incentives by model by year
are not available. The addition of this variable as a factor may improve correlation.
Practical implications The results in this study could help the automobile companies better
understand their business, and the auto companies could use the results for possible strategic
decisions. In addition, legislatures in the impacted states could use the results to prepare for
fluctuations in the industry that would result in profound effects on the states in question.
Originality/value This type of analysis is not standard, and the use of multiple economic
variables correlated with domestic and foreign car sales is unique. The study provides a basis for
further research.
Keywords Automotive industry, Sales forecasting, Economic conditions,
United States of America
Paper type Research paper
Introduction
Planning is an essential part of any business activity. However, business plans require
objectives that are based on sales targets, which in turn require demand forecasts.
Thus, forecasting is essential for planning. In addition, forecasts serve as input to
many other business decisions. Obviously, these decisions can be only as good as the
forecast results used to make them.
Sales forecasts are the foundation of planning. The forecasts enable an organization
to have an optimum inventory level, to make appropriate purchasing decisions and to
maintain efficient daily operations. All these affect the profits of the organization.
Therefore, forecasting is critical to profitability.
Demand planning involves the process of creating and affecting demand in the
future. Regardless of method chosen (promotion, etc.), forecasting helps assess the
impact of each possible decision upon demand. Demand management integrates all
aspect of an organizations strengths and weaknesses. It includes not only planning
and forecasting but also coordinating all activities that affect customer demand, e.g.
creating, shaping and fulfilling demand.
Demand forecasting requires projecting what will happen to demand in the future.
Obviously, this requires statistical forecasting methods. Unfortunately, there is still a
gap between the statistical and economic techniques offered by forecasting and the
Forecasters tend to ignore likely changes that may influence the forecast, e.g.
increases in population, increases in competition, technological changes, etc. Any
or all of these factors may affect the organizations sales and can easily be
included.
Forecasting
automobile sales
671
MRN
32,7
672
Using inappropriate computational methods for the data. Each type of data (e.g.
time series, cross-sectional data) requires different forecasting techniques.
Incorrect computational techniques cause errors in forecasts.
Forecasting
automobile sales
673
MRN
32,7
674
The LR 799 model (Tanner, 1977, 1979). This model tried to predict cars per
person in Great Britain using a very few explanatory variables, e.g. GDP per
person and cost of operating a car. However, Tanner found that this model was
too complex and required a large amount of data.
The National Road Traffic Forecast (NRTF) models by Romilly (1998). The
NRTF models include household-based and explanatory models. Both models
use the combination of time series incorporating causal variables. Romilly found
the results encouraging but susceptible to small-sample bias.
The regression model. This is a commonly used model that relates variables and
determines causal relationships. In a regression model, one can use statistical tests
to determine the significance of the model as well as the variables.
Data
To forecast automobile sales in the USA, relevant demographic and economic variables
were selected, and data were collected from many sources, e.g. Bureau of Labor
Statistics, Federal Reserve Bank and Moodys economy.com, economagic.com, US
Commerce Department. Further, quarterly data from 1959 to 2006 were amassed on
total automobile sales in units and in dollars, and automobile sales data were also
collected on domestic cars and trucks as well as foreign brand cars. The variables and
equations used are listed below. The independent variables are durable industrial
demand, durable personal consumption, discount rate, non-durable industrial goods
demand, personal consumption, GNP, GDP, population, leading economic indicators,
M1 (money that is liquid), M2 (M1 plus short-term invested assets), and M3 ( M2 and all
institutional funds). All these variables are assumed to affect the purchase of
automobiles and other consumer goods. However, some economic variables may make
take a long time to show their effects. Thus, in order to find whether variables have lag
relationships, variables were lagged one quarter. To understand whether any or all of
these variables can predict automobile sales in units or value of shipment (in dollars),
multiple regression method was used.
Two models were tested:
A model without a lag:
(1)
Analyses
Data on automobile sales in units and in dollars and all the twelve independent
variables were regressed using Statistical Package for the Social Sciences (SPSS).
Multiple regression models with stepwise method were run.
Regressing the 12 unlagged independent variables with total (domestic and foreign)
automobile demand in units, the equation with significant variables is:
Y 111922:95X1 2:48X2 0:08X3 28:84X4 0:3X6 1:8X8 11:72X9 1:69X10
t
8:452:5
2
2:71
R 0:75 and F 30
8:02
3:3
3:8
6:4 4:57
4:6
The selected variables have significant t values, and the R2 of 0.75 indicates that these
variables (equation) can explain the automobile sales in units with 75 percent accuracy.
In other words, 75 percent of the automobile sales in units can be explained by this
equation. Durban-Watson statistic is inconclusive about serial correlation and analysis
of errors indicates no heteroscedasticity.
When demand for domestic cars (in units) alone is regressed with the 12 unlagged
independent variables, the equation with significant variables is:
Forecasting
automobile sales
675
MRN
32,7
t
2
R 0:74
676
4:05
and F 44
10:01
7:57
9:81
The variables have significant t values, and the R2 of 0.74 indicates that this equation can
predict the sales in units with 74 percent accuracy. In other words, 74 percent of the
automobile sales in units can be explained by this equation. Durban-Watson statistic is
inconclusive about serial correlation and analysis of errors indicates no heteroscedasticity
When demand for foreign cars (in units) alone is regressed with the 12 unlagged
independent variables, the equation with significant variables is:
Y 1172:7 0:85 X1 25:27 X4 0:195 X7 6:83 X9 0:23 X12
t
15:5 5
13:6
14:3 18
14:06
R2 0:91
F 186
and
All the variables have significant t values, and the R2 of 0.91 indicates that this
equation can predict with 91 percent accuracy the sales. This indicates that the demand
for foreign cars (in units) is highly correlated with the significant variables. DurbanWatson statistic is inconclusive about serial correlation and analysis of errors indicates
no heteroscedasticity
Regressing the unit sales of trucks (from 1976 to 2006) only with unlagged
variables, the equation with significant variables is:
Y 768 1:34 X1 3:93 X2 32:4 X4 0:454 X11
19:614:41
t
2
R 0:99
and
22:13
F 2;036
12:71 14:8
Durban-Watson dw 1:2 k 4; n 123
All the variables have significant t values, and the R2 of 0.99 indicates that this
equation can explain with 99 percent accuracy the sales of trucks. Durban-Watson
statistic is inconclusive about serial correlation and analysis of errors indicates no
heteroscedasticity
In order to determine whether the value (in dollars) of automobiles sold could
provide high predictive results, the sale of automobiles in dollars was regressed with
the same 12 unlagged independent variables. The analyses of the sales value of
automobile with the unlagged independent variables resulted in:
Y 4432 53:45 X1 181:38 X2 9:08 X12
t
4:34 5:7
20:61
11:99
R2 0:996
and
F 5;955
The four variables have significant t values, and the R2 of 0.996 indicates that this
equation can predict with 99.6 percent accuracy the sales of total value of automobile
shipment. Both R2 and F are very significant and indicate high predictive power.
Durban-Watson statistic is inconclusive about serial correlation and analysis of errors
indicates no heteroscedasticity.
Forecasting
automobile sales
t
2
R 0:994
and
10:3
F 7; 553
All variables have significant t values, and the equation has an R of 0.994, which
indicates that this equation can predict with 99.4 percent accuracy the sales of
automobile in dollar. In other words, 99.4 percent of the automobile sales in dollar can
be explained by the equation. Durban-Watson statistic is inconclusive about serial
correlation and analysis of errors indicates no heteroscedasticity
By lagging the variables by one quarter and regressing the total (domestic and
foreign) units of automobile sales, the equation is:
Y 10776 3:18 X1 0:298 X2 0:08 X3 22:53 X4 2:25 X6 1:77 X8
t
R2 0:72 and F 25
The variables have significant t values, but the R2 of 0.72 not a high value. The
equation can explain the automobile sale in units with only 72 percent accuracy.
Durban-Watson statistic is inconclusive about serial correlation and analysis of errors
indicates no heteroscedasticity
Regressing the unit sales of just the domestic cars with variables lagged one
quarter, the equation is
Y 8470 7:5 X1 0:074 X3 53:34 X4 1:5 X8 0:93 X10 0:14 X12
t
6:48 5:01 7:5
7:5
5:52 4:30 2:18
R2 0:73 and F 35:25
These variables have significant t values, but, again, the R2 of 0.73 is not a high value, and
the equation can predict with only 73 percent accuracy. Durban-Watson statistic is
inconclusive about serial correlation and analysis of errors indicates no heteroscedasticity
Regressing the unit sales of foreign cars with variables lagged one quarter, the
equation is
Y 904 1:66 X1 20:1 X4 0:235 X7 4:13 X9 0:238 X12
t
9:3 4:2
8:67
15:55 6:4
13:45
R2 0:91
and
F 176
Here the variables have significant t values, and the R2 of 0.91 indicates that this
equation can predict with 91 percent accuracy the sales of foreign cars. Durban-Watson
statistic is inconclusive about serial correlation and analysis of errors indicates no
heteroscedasticity.
677
MRN
32,7
When regressing the unit sales of trucks (from 1976 to 2006) with variables lagged
one quarter, the equation is
Y 345 2:5 X2 36:62 X4 0:192 X8 0:367 X11
1:77 4:8
t
2
678
R 0:98
and
10:46
F 1; 070
2:92
9:4
The variables again have significant t values, and the R2 of 0.98 indicates that this
equation can predict with 98 percent accuracy the sales of trucks. Durban-Watson
statistic is inconclusive about serial correlation and analysis of errors indicates no
heteroscedasticity
The results of all the models are summarized in Table I.
As can be seen, foreign cars sales (lagged or unlagged) has a high correlation with
significant variables, which means that it can be predicted with high accuracy
compared to domestic cars. Domestic cars are harder to predict with the same
independent variables. This obviously indicates that domestic cars sales are hard to
predict due to many other factors, such as discounts, quality problem, and amount of
advertising, are affecting sales. In contrast, demand for trucks can be predicted with
high accuracy. However, the equation for trucks is based on data from 1976 to 2006, so
it cannot be said with certainty that the future demand for trucks can be predicted with
such a high accuracy, because the novelty and the competitive pressure of foreign car
manufacturers will affect the demand for domestic trucks as well. As a result, the
domestic truck market may eventually show the same weak relationship with the
significant independent variables as the domestic cars.
Conclusion
The automobile industry is a major component of the US economy. Especially
in states where automobiles dominate the manufacturing sector, decreases in
automobile sales can have dire economic consequences for the state and its people.
Therefore, it is critical that the automobile industry plan its business carefully and to
be aware of the upturns and downturns that might be coming so as to prevent or
ameliorate any economic shock those changes might have on the countrys or the
states economy.
To accomplish this, automobile manufacturers and dealers must carefully develop
their business plans. For a plan to be effective, it must be reliable. The reliability of a
plan increases if it is based on realistic goals. The goals can be realistic only if they are
based on an analysis of the conditions affecting the business. Such analysis requires
use of a forecasting method that allows the incorporation of as many conditions as
feasible that might affect the business and that can be used to predict demand.
R2
Table I.
0.75
0.74
0.91
0.99
0.996
Unlagged
F
30
44
186
2,036
5,955
25
35.25
175
1,070
7,553
Regression allows for such analysis. Therefore, to forecast automobile sales, twelve
possible economic factors were analyzed to establish a relationship between the
variables and the automobile sales. The data were analyzed using regression without a
lag and with a one quarter lag. Using the t-test, some variables were dropped due to
their insignificant relationship. Total automobile sales (domestic and foreign car sales)
in units, domestic car sales, foreign car sales, truck, and value (in dollars) of automobile
shipment were analyzed. Relationships of some of the independent variables with total
sales in dollars, foreign cars, and trucks were highly correlated with both lagged and
unlagged independent variables.
Based on the analysis, foreign car sales also proved to be highly correlated to the
economic variables, while domestic car and total car sales were not significant and had
a very weak relationship. Therefore, it can be safely concluded that it is possible to
predict foreign car sales, truck sales, and the total sales of automobile in dollar with
high degree of confidence. However, these results and the domestic car sales can
further be improved by segmenting cars and trucks by size and price range.
Some might counter that using current economic indicators to forecast current sales
cannot be theoretically supported. However, the purpose of analyzing the relationship
of sales of cars and trucks with both lagged and unlagged variables was to find
whether current information would indicate close relationships. Apparently, both
lagged and unlagged indicated the same strength of relationship with the dependent
variables. In other words, the strength of the relationships did not improve or
deteriorate depending on whether variables were lagged or not.
In addition, some independent variables show high correlation with each other, e.g.
GNP and personal consumption. However, removing one or two significant
independent variables did not change the strength of the correlation significantly;
therefore, they were not eliminated in the study.
This study has correlated many economic and demographic independent variables
with the sales of automobiles in the USA. The results and relationships can further be
improved by the substitution of or the inclusion of subsets of the variables such as
analysis of each automobile company individually, of each model car, and of vehicles in
different price ranges. However, such ses are not possible for this researcher without
the cooperation of each automobile manufacturer, as the required data is not available
publicly. For now, this study can be used as a predictive model of the overall
automobile industry and as the basis for further study.
References
Abu-Eisheh, S.A. and Mannering, F.L. (2002), Forecasting automobile demand for economics in
transition: a dynamic simultaneous equation system approach, Transportation Planning
and Technology, Vol. 25 No. 4, pp. 311-31.
Armstrong, J.S. (1985), Long-Range Forecasting, John Wiley, New York, NY.
Armstrong, J.S. (1986), Research on forecasting; a quarter-century review, 1960-1984, Interface,
Vol. 5 No. 1, pp. 89-103.
Armstrong, J.S., Morwitz, V.G. and Kumar, V. (2002), Sales forecasting for existing consumer
products and services: do purchase intentions contribute to accuracy?, International
Journal of Forecasting, Vol. 16 No. 4, pp. 383-97.
Byrnes, J.C. (1964), Consumer intention to buy, Journal of Advertising Research, Vol. 4, p. 49.
Carlson, R.L. and Umble, M.M. (1980), Statistical demand for automobile and their use for
forecasting in an energy crises, The Journal of Business, Vol. 53 No. 2, pp. 193-204.
Forecasting
automobile sales
679
MRN
32,7
680
Dargay, J. and Gately, D. (1999), Incomes effect on car and vehicle ownership, worldwide:
1960-2015, Transport Research, Vol. 33 No. 2, pp. 101-38.
Department of Labor (2004), Consumer spending in the United States, available at:
www.bls.gov/ro5/cesmid.htm (accessed 10 December 2007).
Fauvel, Y. and Samson, L. (1991), Intertemporal substitution and durable goods: an empirical
analysis, Canadian Journal of Economics, Vol. 24 No. 1, pp. 192-205.
Fitzsimmon, G.J. and Ajzen, I. (1975), Belief, Attitude, Intention, and Behavior, Addison-Wesley,
Reading, MA.
Fulton, G.A., Sean, A.M, Donlard, R.G., Lucie, G.S. and Barbara, C.R. (2001), Contribution of the
Automobile Industry to the US Economy in 1998: The Nation and Its Fifty States, The
University of Michigan Institute of Labor and Industrial Relations, The University of
Michigan Transportation Research Institute, Office for Industrial Automotive Transport
and the Center for Automotive Research, Ann Arbor, MI.
Garcia-Ferrar, A., Dell, J.H. and Martin-Arroyon, A.S. (1997), Univariate forecasting comparison:
the case of the Spanish automobile industry, Journal of Forecasting, Vol. 16 No. 1, pp. 1-17.
Harris, E.S. (1986), Forecasting automobile output, Federal Reserve Bank of New York,
Quarterly Review, pp. 40-42.
Justin, F.T. (1996), Consumer buying intentions and purchase probability: an experiment in
survey design, Journal of the American Statistical Association, Vol. 61 No. 315, p. 658.
Lee, M., Elango, B. and Schnaars, S.P. (1997), The accuracy of the conference boards buying plan
index: a comparison of judgmental vs. extrapolation forecasting methods, International
Journal of Forecasting, Vol. 13 No. 1, pp. 127-35.
McAlinden, S.P., Hill, K. and Swicki, B. (2003), Economic Contribution of Automotive Industry to
the US Economy An Update, Center for Automotive Research, Ann Arbor, MI, available
at: www.cargroup.org/pdfs/Alliance-Final.pdf (accessed 20 December 2007).
Mahmoud, E. (1984), Accuracy of forecasting: a survey, Journal of Forecasting, Vol. 3 No. 2,
pp. 139-51.
Morrison, D.G. (1979), Purchase intentions and purchase behavior, Journal of Marketing, Vol. 43
No. 2, pp. 65-74.
Morwitz, V.G. and Schmittlein, D. (1992), Using segmentation to improve sales forecasts based
on purchase intent: which intenders actually buy?, Journal of Marketing Research, Vol. 29
No. 4, pp. 391-405.
Nerlove, M. (1957), A note on long-run automobile demand, Journal of Marketing, Vol. 22 No. 1,
pp. 57-64.
Pickering, J.F. (1978), The durable purchasing behavior of the individual household, European
Journal and Marketing, Vol. 122 No. 2, pp. 178-93.
Prevedouros, P.D. and Ann, P. (1998), Automobile ownership in Asian countries: historical trend,
and forecast, ITE Journal, Vol. 68 No. 2, pp. 24-29.
Sanders, N.R. (1997), The status of forecasting in manufacturing firms, Production and
Inventory Management Journal, Vol. 5 No. 1, pp. 32-36.
Shahabuddin, S. (1987), Why forecasts are wrong? The Journal of Business Forecasting
Methods and Systems, Vol. 6 No. 3, p. 16.
Suits, D.B. (1958), The demand for new automobile in the United States, 1926-1956, Review of
Economics and Statistics, Vol. 40 No. 3, pp. 273-80.
Tanner, J.C. (1977), Car ownership trend and forecast, Department of the Environment,
Department of Transport, TRRL Report LR799, Crowthorne.
Tanner, J.C. (1979), Choice of model structure for ownership forecasting, Department of the
Environment, Department of Transport, TRRL Report LR799, Crowthorne.
Thompson, P.A. and Noordewier, T. (1992), Estimating the effect of consumer incentive program
on domestic automobile sales, Journal of Business Economic Statistics, Vol. 10 No. 4,
pp. 408-47.
Forecasting
automobile sales
Wu, D.-M. (1965), An empirical analysis of household durable goods expenditure, Econometrics,
Vol. 33 No. 4, pp. 761-80.
Further reading
Barlevy, G. (2002), The sullying effects of recessions, The Review of Economic Studies, Vol. 69
No. 1, pp. 65-96.
Berkovec, J. (1985), New car sales and used car stocks: a model of the automobile market, The
Rand Journal of Economics, Vol. 16 No. 2, pp. 195-214.
Boustead, T. (1997), The US economy to 2006, Monthly Labor Review, Vol. 120 No. 11, pp. 6-22.
Bunn, D. and Wright, G. (1991), Interaction of judgemental and statistical forecasting methods:
issues and analysis, Management Science, Vol. 37 No. 5, pp. 501-20.
Clark, L. (1981), Time lags: a controversy in monetary policy: a note, American Economist,
Vol. 25 No. 2, pp. 63-68.
Conlisk, J. (1981), Bears theorem with infinite lags, Journal of Economic Dynamics and Control,
Amsterdam, Vol. 3 No. 1, p. 43.
Dohinue, R.M. (1993), Evaluating the role of judgement plays in forecasting accuracy, Journal
of Forecasting, Vol. 12 No. 2, p. 81.
Eberly, J. (1994), Adjustment of consumers durable stocks: evidence from automobile
purchases, Journal of Political Economy, Vol. 102 No. 3, pp. 403-36.
Fernandez, P.V. (1999), Forecasting home appliances sales: incorporating adoption and
replacement, Journal of International Consumer Marketing, Vol. 12 No. 2, pp. 39-61.
Greenspan, A. and Cohen, D. (1999), Motor vehicle stocks, scrappage, and sales, The Review of
Economics and Statistics, Vol. 81 No. 3, pp. 369-83.
Higgins, B. (1982), Should the federal reserve fine tune monetary growth?, Economic Review
Federal Reserve Bank of Kansas City, Kansas City, MO, pp. 3-17.
Jain, C.L. and Chen, T.P. (1992), The role of judgment in business forecasting, Industrial
Management, Vol. 34 No. 6, pp. 1-3.
Khan, J. (1986), Gasoline prices and the used automobile market: a rational expectations asset
price approach, Quarterly Journal of Economics, Vol. 110 No. 2, pp. 323-39.
Lawrence, M.J., Edmundson, R.H. and OConnor, M.J. (1986), The accuracy combining
judgemental and statistical forecast, Management Science, Vol. 32 No. 12, pp. 1521-32.
Lee, B.S. (1992), Causal relations among stock returns, interest rates, real activity and inflation,
Journal of Finance, Vol. 47 No. 4, pp. 1591-603.
Lee, D. and Kreutzer, D. (1982), Lagged demand and a perverse response to threatened property
rights, Economic Inquiry, Vol. 23 No. 2, pp. 353-61.
Mehra, Y.P. and Spencer, D.E. (1979), The St. Louis equation and reverse causation: the evidence
reexamined, Southern Economic Journal, Vol. 45 No. 4, p. 1104.
Mukherjee, T.K. and Naka, A. (1995), Dynamic relations between macroeconomic variables and
the Japanese stock market: an application of a vector error-correction model, Journal of
Empirical Research, Vol. 18 No. 2, pp. 223-37.
Nasseh, A. and Strausss, J. (2000), Stock prices and domestic and international macroeconomic
activity: a cointegration approach, Quarterly Review of Economic and Finance, Vol. 40
No. 2, pp. 229-45.
681
MRN
32,7
682
Regional Spending Patterns in the US and Metropolitan Areas in the Midwest, 2001-2002 (2004),
Bureau of Labor Statistics, News Release, 22 January.
Romelly, P., Song, H. and Liu, X. (1995), Modeling and forecasting car ownership in Britain,
Journal of Transport Economics and Policy, Vol. 32 No. 2, pp. 165-85.
Rosenbum, M.S. (1995), Lags in the effect of monetary policy, Economic Review, Federal
Reserve Bank of Atlanta, Atlanta, GA, Vol. 70 No. 10, pp. 20-34.
Souleles, N. and Agarwal, S. (2003), The response of consumer spending and debt to tax rebates:
evidence from the CEX and the household credit accounts, Manuscript.
Webby, R. and OConnor, M.J. (1996), Judgemental and statistical time series forecasting: a
review of literature, International Journal of Forecasting, Vol. 12 No. 1, p. 91.
About the author
Syed Shahabuddin, Professor of Management Science and Operations Management, received his
PhD from the University of Missouri, Columbia, Missouri. He has been with Central Michigan
University since 1980. Before joining Central Michigan University, he taught at the University of
Notre Dame, South Bend, Indiana. He has served as chairman of two departments at Central
Michigan University, and was a Fulbright scholar. Dr Shahabuddin has published more than 50
articles and two books, entitled Management Science and Programming in Basic, and a third
book, Business Statistics, to be published shortly. He has chaired many sessions and presented
many papers in the Decision Science Institute and the INFORMS national meetings. Syed
Shahabuddin can be contacted at: [email protected]
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.