Regional differences in technical efficiency and technological gap of Norwegian dairy farms: a stochastic meta-frontier model

ABSTRACT

This paper compares technical efficiencies (TEs) and technological gap ratios (TGRs) for dairy farms in regions of Norway, accounting for differences in working environments. We used the state-of-the-art stochastic meta-frontier approach to estimate TEs and TGRs to account for regional heterogeneity, and the ‘true’ random-effect model to account for farm effects. The dataset used was farm-level balanced panel data for 24 years (1992–2014), with 5442 observations from 731 dairy farms. The results of the analysis provide empirical evidence of small regional differences in TEs, TGRs, and input use. Furthermore, the results may provide support for the more regionally specific agricultural policy, in terms of support schemes and structural regulations.

KEYWORDS:

• Dairy farms
• metafrontier
• heterogeneity
• technical efficiency

JEL CLASSIFICATION:

• R58
• O52
• C23
• D24

Acknowledgments

We thank the two anonymous referees for many insightful comments and suggestions that helped to improve this article. We are grateful for the financial assistance of the Research Council of Norway which supported the project with grant number 225330/E40.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ Some researchers have employed Bayesian estimators that resemble the random parameter model in assuming a stochastic model with exponentially distributed inefficiency. For further information, refer to Koop and Steel (2001), Tsionas (2002), and Assaf (2011).

² In this study, we used the ‘true’ random-effect model and not the ‘true’ fixed-effect model. Estimates (not reported here) show reasonably low correlation between farm/firm effects and the regressors (less than approximately 0.5). In addition, we used an unbalanced panel in which 25% of the sample has four or fewer observations per farm (i.e. panel data with a large share of short time period/time series). In cases like this, based on Clark and Linzer (2015), a fixed-effect model exacerbates measurement error bias and the random-effect model is preferable. Another drawback of the features of fixed-effects models is that they cannot be used to investigate time-invariant causes of the dependent variables.

³ The tests are not reported here due to space constraints, but are available upon request from the principal author.