Quick-service expenditure in Ireland: parametric vs. semiparametric analysis

Abstract

The aim of this study is to identify the economic and socio-economic factors influencing Irish households' expenditure on quick-service meals, a particularly dynamic component of the foodservice industry, and to determine the extent to which these factors have changed over the course of the 1990s. Maximum likelihood estimation and semiparametric alternatives are considered with the conclusion that in this instance semiparametric techniques do not offer a viable alternative to maximum likelihood estimation of tobit models, even in the presence of heteroscedasticity and non-normality. The results revel that household income, place of residence, commuters and household size have significant and positive influences on quick-service expenditure. Older families, single households and married couples, together with homeowners, display reduced expenditure. The opportunity cost of time is positively related to quick-service expenditure, consistent with theory, while health knowledge has a negative impact on quick-service consumption.

Acknowledgements

C. Keelan was funded for this research by a Walsh Fellowship from Ashtown Food Research Centre, Teagasc and this support is gratefully appreciated. The authors would like to acknowledge feedback received at the 79th Agricultural Economics Society Annual Conference in Nottingham in April 2005, in particular, from Dr. Philip Lund and Dr. Geir Gustavsen. The authors would also like to acknowledge the helpful comments of the editor and an anonymous referee.

Notes

¹In keeping with most other studies in this area, this article classifies foods ‘at home’ and ‘away from home’ based on where the food was prepared or obtained, not where it was consumed (Lin et al., 2001).

²Authors own calculations based on Irish Household Budget Survey data (Central Statistics Office, 1997, 2002).

³Lazaridis (2002) disaggregated the Greek market into expenditure in restaurant, in coffee houses and in takeaway and canteens.

⁴The 1994/5 HBS and the 1999/2000 HBS are hereafter referred to as 1994 and 1999.

⁵EU adult equivalence scales give the first adult a weight of 1, each other adult 0.7 and each child under 14 years a weight of 0.5.

⁶It can be hypothesized that the greater the proportion of household members in the labour force, the less likely they will have time for food preparation in the home leading to a greater reliance on processed foods and other time-saving choices such as FAFH and quick-service meals.

⁷See for example, McCracken and Brandt (1987), Byrne et al . (1998) and Stewart et al . (2004) amongst others.

⁸In the HBS the head of household is the oldest person in the household and given that the completion of the expenditure diary is in itself a task indicative of household management. The choice of the head of household as the household manager can easily be justified.

⁹In the case of expenditure on food items, such as FAFH, we assume that the household's decision to purchase a food item depends on the utility they derive from the consumption of that item. It may be the case that the household dislikes the food item to such an extent that they attach a negative level of utility to it. In such a case, all we would observe in the expenditure data are zero values. As such, the tobit model, which takes account of censoring of this kind in the dependent variable, is appropriate in this case.

¹⁰A number of studies considering the least absolute deviations approach without censoring as an alternative to the standard parametric models for dealing with censoring found the former to be superior in most cases. See for example, Fullerton (1998), Yoo et al . (2000) and Yoo (2001).

¹¹For consistency, bootstrapping is also used to compute the residuals of the SCLS model.

¹²The respective ado files were obtained at http://elsa.berkeley.edu/~kenchay. For consistency, the residuals of the tobit model estimated using MLE are computed using bootstrapping.

¹³The results of a Lagrange Multiplier test for heteroskedasticity and Pagan and Vella's (1989) moments based test for non-normality are presented in the Appendix.

¹⁴See Newman et al . (2003) for an example of an application of the adjusted model.

¹⁵In contrast, it should be noted that Stewart et al . (2004) found that larger households spend less on all segments of FAFH, supporting their hypothesis that such households benefit from economies of scale in food preparation at home.