280 likes | 371 Views
Factors affecting forward pricing behaviour: Implications of alternative regression model specifications. By H Jordaan and B Grové. Outline. Introduction Data and Procedures Results Conclusions. Introduction.
E N D
Factors affecting forward pricing behaviour: Implications of alternative regression model specifications By H Jordaan and B Grové
Outline • Introduction • Data and Procedures • Results • Conclusions
Introduction • SA agricultural sector experienced long history of state intervention (1930s to mid 1990s) • Period of regulation ended with Marketing of Agricultural Products Act of 1996 ordering demise of control boards • Farmers became responsible for marketing of their own produce • Became exposed to price risk
Introduction (Cont…) • Three markets emerged for grain in SA • Cash (spot) mkt • Forward contracting mkt • Derivatives mkt (Futures & options) • Although forward pricing can reduce price risk few farmers tend to use it (both in South Africa and internationally) Forward pricing
Introduction (Cont…) • Some researchers expected growth in the use of forward pricing (i.e. due to learning curve effect) • From the literature growth is lacking • Ample research was done to determine why producers do not use forward pricing methods
Introduction (Cont…) • Number of different modelling specifications were used over the years • Tobit models • Multinomial logit models • Binary models (Logit & Probit) • Two-step approach • Binary sub-model (Adoption decision) • Linear sub-model (Quantity decision) Single decision Separate decisions
Introduction (Cont…) • PROBLEM – If variables are forced to influence decision in certain way • May overlook important factors restricting use of forward pricing methods • May also explain lack of growth in use of forward pricing methods
Introduction (Cont…) • OBJECTIVE • Compare results from three model specifications • Determine implications of alternative model specifications in the South African context • Single decision vs Separate decisions
Data and Procedures • Used primary data from study by Jordaan & Grové (2007) • Questionnaire survey conducted (October 2005) • 78 randomly drawn farmers were personally interviewed • 50 produced maize • 22 used forward pricing methods
Data and Procedures (Cont…) • SINGLE DECISION framework – dependent variable specified as proportion of crop which is forward priced • Non-adopters forward priced 0% of their crop • 56% did not use forward pricing methods • Dependent variable is censored at threshold of 0 • Need model that can effectively model censored dependent variable
Data and Procedures (Cont…) • Tobit model relates observed outcomes of yi > 0 to an index function (Heij et al., 2004): Where if if
Data and Procedures (Cont…) • Second and third specifications relate to TWO-STEP approach • 1) Adoption decision – • binary choice whether or not respondent used forward pricing methods • binary logit regression
Data and Procedures (Cont…) • 2)Quantity decision • choice on proportion of crop that is forward priced, CONDITIONAL TO RESPONDENT HAVING USED FORWARD PRICING METHODS • OLS regression
Two-step approach Results NA NA NA
Results - + + + -
Conclusions • Modelling forward pricing behaviour as single decision assumes same variables influence two forward pricing decisions • This assumption is again proven wrong in this study • Importance of risk aversion is again highlighted • Only significant factor in all three specifications • Confirms findings by McNew and Musser (2000)
Conclusions (Cont…) • TOBIT specification fails to expose • producers’ perception that forward pricing is risky marketing alternative • adopters of forward pricing methods tend to use portfolio of risk management tools
Conclusions (Cont…) • Conclusions based on results where variables are forced to influence two decisions similarly may • overlook restrictive impact of certain factors on the use of forward pricing methods • explain the lack of growth in the use of forward pricing methods
Conclusions (Cont…) • Thus, study provides more evidence that quantity decision should be modelled conditional of rather than jointly with adoption decision (Katchova & Miranda, 2004)