Willingness of farmers to participate in agri-environmental auctions in Finland

Abstract

Auctions have been applied in agri-environmental policy, serving as a key tool to overcome information asymmetries and generate important cost savings for governments. The efficiency of agri-environmental auctions depends on ensuring sufficient participation and avoiding a ‘learning by experience’ situation. In designing successful auctions, it is important to acknowledge farmer characteristics that might increase the adaptability of auctions and also recognize whether past experience affects future participation. This paper uses data from an auction experiment conducted in Nurmijärvi, Southern Finland. We account for socio-demographic, spatial and attitude variables and investigate their effect on the probability of past and future auction participation. Due to the small number of actual participants, we employ a relogit model to correct the coefficient estimates derived by a binary logit model. According to the analysis, large-scale farmers are more likely to have participated in the pilot auction, while older farmers, those engaged full time in farming and less well-trained farmers were less likely to be positive towards future auctions. Past participation was positively and significantly related to prospective auctions. Our findings suggest a strong relationship between attitudes and participation, particularly regarding specific environmental benefits attached to the auction scheme, novelty and financial features as well as the complexity of the auction mechanism. The predicted probability for both past and future participation elicited by the relogit model was consistent with the sample probability, and hence by applying a relogit to correct for rare-event bias we derived more reliable estimates.

Keywords:

• Auction
• participation
• attitude
• relogit model

Acknowledgments

Reserarch funding from the Ministry of Agriculture and Forestry as well as Mr Jonne Lehtimäki's participation in data collection are gratefully acknowledged.

Notes

1. For each alternative, the respondent may obtain a certain level of utility U_y that can be defined as U_i,y = V_i,y + ε_i,y ᗄy,y = 1,0 where V_i,y is the representative utility specified to be linear in the individual's characteristics and chosen alternative, both of which are observable. Conversely, ε_i,y denotes the influences on utility that cannot be measured. The respondent will prefer alternative y = 1 if only U_i,1>U_i,0 or V_i,1 + ε_i,1 > V_i,0 + ε_i,0 or ε_i,0 − ε_i,1 < V_i,1 + V_i,0. It is thus only the difference in utility that matters and not the specific values of U₁ and U₂.

2. For the expression used to estimate the bias in , see Appendix C in King and Zeng (2001b).

3. For a detailed description regarding simulation computations, see King and Zeng (2001b, p. 149).

4. Here, statements for the factor that referred to time may be interpreted along with the notion of ‘time is money’.