Interaction between price, quality and country of origin when estimating automobile demand: the case of Turkey

Abstract

We estimate the demand for new automobiles in Turkey immediately following its entry to Custom Union with the European Union in 1996. We use quarterly data on price, quantity, quality, country of origin and product characteristics of the new automobile sales market and macroeconomic variables to estimate market demand during 1996–1999 using dynamic Generalized Least Squares Estimation method. In addition to including quality, we also include country of origin, to serve as another proxy for quality and to capture differences in cost-increasing and demand-reducing trade restrictions imposed on cars from certain countries as well as taste differences. Our results indicate that country of origin as well as quality matter for automobile demand in Turkey. Consistent with other aggregate automobile demand models for other countries, the demand for new automobiles is found to be price inelastic in the short run.

Acknowledgements

We are grateful to Oya Pinar Ardic, Burcay Erus and an anonymous referee for their invaluable comments and suggestions. We thank Natiq Mustafayev and Pinar Uysal for their able research assistance; Murad Cakir and Yusuf Soner for helpful discussions. The usual disclaimer applies. We acknowledge financial support from Bogazici University Research Fund No. 04C102 and Automobile Distributors Association of Turkey. Alper acknowledges financial support from TUBA-GEBIP (Turkish Academy of Sciences–Young Scientists Scholarship Program).

Notes

¹ On 1 January 1996 the Customs Union between the European Union and Turkey came into effect, whereby Turkey has adapted its trade regime and other relevant legislation to that of the EU becoming the only nonmember country – with the exception of Malta, San Marino and Andorra – which has signed a customs union agreement with the European Union.

² See, among others, Berument and Gunay (2003) and Ertugrul and Selcuk (2001).

³ Out of 185 models, 65 was omitted from the data set due to missing price data over the substantial portion of the period.

⁴ For example, Honda Civic Sedan produced in Turkey is classified as a domestic origin automobile, although some parts of it are imported.

⁵ Another approach is to run a hedonic price regression (used by Levinsohn, 1988) to group the cars in neighbourhood to determine a segment.

⁶ Specifically, the real interest rate is calculated using the following formula: where r_t is the real interest rate at time t, i_t is the nominal interest rate at time t, and is the expected inflation rate which is assumed to be equal to π _t .

⁷ Theoretically, we expect a positive relation between the demand of an automobile and the availability of credit. However, during 1996–1999, the public sector borrowing requirement in Turkey was very high and the main asset item in the commercial banks’ balance sheets was government securities rather than household credits. Hence, a priori we do not expect real credit volume to be significant.

⁸ With the Customs Union, Turkey (a) abolished its tariff and nontariff protection against the EU for all goods covered by the Customs Union, (b) progressively aligned itself to the EU Common external Tariff 2 and (c) assumed the preferential trade system of the EU through Free Trade Agreements with third world countries.

⁹ Against the alternatives of fixed and random effects models, we estimate using the FGLS method. This is due to two reasons. First, based on the redundant fixed effects likelihood ratio test statistics, we conclude that there are no significant fixed effects. Moreover, the main contribution of the article is to inquire whether country-of-origin matters for automobile demand in Turkey following the customs union agreement in 1996. The use of intercept dummy variables denoting country-of-origin is not compatible with fixed effects model. Second, given the small number of temporal observations, inclusion of the lagged dependent variable to account for the presence of autocorrelation precludes the random effects model. Moreover, Wu–Hausman test statistics indicate no random effects against the alternative of fixed effects model when the lagged dependent variable is taken out of the model.

¹⁰  The inclusion of the lagged dependent variable is also appropriate since it captures first-order serial correlation in the disturbance terms.

¹¹ Among others, see Stiglitz (1987).

¹² The endogeneity issue may be addressed econometrically using an instrumental variable approach. However, in our unique data set, there are no cross-section instruments available. Hence, we caveat the reader on the endogeneity issue for models including quality.

¹³ When prices of all the models change simultaneously at the same rate, the relative prices do not change, hence the substitution effect across different models will be zero. Thus the change in the prices will affect the automobile demand only through income effect. While EU and non-EU imported origin automobiles exhibit normal good characteristics with positive income effect, the domestic origin automobiles exhibit no income effect.