The Impact of Extension Services on Farm Level Outcomes: An Instrumental Variable Approach Anthony Cawley, Walsh Fellow REDP, Teagasc & NUI Galway Supervisors Dr. Kevin Heanue, REDP, Teagasc Prof. Cathal O’Donoghue, REDP, Teagasc Prof. Maura Sheehan, Edinburgh Napier University & NUI Galway Dr. Rachel Hilliard, Dept. of Management, NUI Galway Outline Introduction • Contribution Context • Literature Review and Research Hypotheses Methodology • Issues, Approach, Instruments and Model Data • Dependent Variables, Explanatory Variables, Summary Statistics Preliminary Results • Relevance and Validity of Instruments, Impact of Extension Conclusion • Policy Implications Future Work • Next steps of PhD Introduction Agricultural extension is the process of transferring specialist knowledge and technology transfer from a policy/academic level to farm level Used to build capabilities of clients, through improved problem solving, decision making and management Many studies have shown that interaction with extension influences farmers’ technology adoption decisions, productivity and profitability levels However, specifically quantifying the causal economic return and controlling for the inevitable endogeneity is less common Thus, this paper applies an Instrumental Variable approach to combat the endogeneity bias Context Extension as policy instrument: • Driver of performance • Risk management tool • Mitigation • Market failure Teagasc Advisory Programme: 1. Business and Technology 2. Environmental/Good Farm Practice 3. Rural Development 4. Adult Training and Life Learning Literature Review Impact of Extension • Positive impact on productivity (Davis et al., 2012) • High benefit cost ratio (Wang, 2041) • Positive effects on technology adoption (Garforth et al., 2013) • Positive impact on gross margin (Lӓpple et al., 2013) • Results can be variable (Feder and Anderson, 2004) • Krishna and Patnam (2014) Diminishing Returns Impact using IV estimation • Increased impact of education on earnings using IV approach given measurement error (Card, 1995) • OLS estimates of education on earnings not significantly downward biased (Callan and Harmon, 1999) • Increased effects of obesity on medical expenditure using IV (Cawley and Meyerhoefer, 2012) • Increased impact of agri education on family farm income (Heanue & O’Donoghue, 2014) Research Objectives Most studies do not account for endogeneity as central focus Therefore estimated results do not account for differences between participants due to self selection, measurement error and omitted variable bias, leading to inconsistent and biased estimations We expect that extension engagement will positively impact farm level outcomes Therefore 2 hypotheses are tested: 1. Extension services positively impact farm level outcomes (family farm income per hectare) 2. The impact of extension on farm income is robust whilst addressing the issue of endogeneity Methodological Challenge: Endogeneity Definition: • Correlation between regressors and error term Causes: • Omitted Variable Bias • Self Selection Bias • Measurement Error Solution: • Instrumental Variable Regression (dependent on validity of instruments) Process: • 2 Stage Least Squares Regression: 1. Regress Instruments on endogenous variable 2. Insert predicted values of variable into structural equation & regress Instruments Requirements: 1. Must be correlated with the endogenous regressor (Relevant) 2. Not necessary to explain dependent variable in original regression (Valid) 3. Must be exogenous to the error term Distance to Local Office Policy Change • Larger distance to local office expected to reduce likelihood of engagement • Expected to negatively affect decision to participate • Does not belong in original regression • Introduction of the SFP in 2005 increased advisory clients by 20% • Expected to positively affect decision to participate • Does not belong in original regression Interactive Term • Shows the effect of distance conditional on the policy change • Expected to negatively affect decision to participate • Does not belong in original regression Functional Form of Model ^ Data Requirements • Participants vs Non Participants, farm outcomes Data Source • • • Teagasc National Farm Survey (NFS) which is collected for the Farm Accountancy Data Network (FADN) required by the European Union It collects an annual panel data set of approximately 1,100 farms nationwide in Ireland weighted to represent the total farming population Collects data on farm profile, output, revenue, costs, and production Variables • • • • Dependent Variable: Family Farm Income / ha Main Explanatory Variables: Advisory Contact Controls: Land, System, Size, Labour, Soil, Region, Stock, farm characteristics Instruments: Distance to advisory office, policy change, interaction term Descriptive Statistics – Policy change 20% overall increase in client numbers (40,700) 42,623 cont. Selected Summary Statistics NFS – Weighted Years 2000-2013 Variable Description Mean SD FFI/ha Family farm income per ha 456.4 432.9 Ln FFI/ha Log of family farm income per ha 5.919 0.938 Advisory Contact = 1 if Teagasc Client 0.532 0.499 Dist. Adv Office Distance to advisory office (km) 10.69 8.600 SFP Year = 1 if year is after 2005 0.655 0.475 Stocking Rate Stocking density per hectare 1.359 0.633 Farm Size Farm size in utilisable hectares 88.59 82.94 Age Age of farmer 54.12 17.33 Off farm Job = 1 if employed off farm 0.352 0.478 Sub sample chosen: 1.Must be in the NFS sample before 2005 2.But not a Teagasc Client before 2005 Results – NFS weighted (n = 9,086) Model OLS 1 Instrument 1st stage Advisory Contact .1903*** (.0196) Policy 2 Instruments 2nd stage 1st stage 2nd stage 1st stage 2nd stage .3511*** (.0414) .3501*** (.0413) .3462*** (.0412) .5248*** Distance .5252*** .5573*** -.0021*** .0005 Interact -.0033*** CD Wald F Statistic R2 Cent. R2 Sargan 3 Instruments 2638.9 1330.0 889.8 .2200 .3634 .2142 .0000 .3646 .2143 .7504 .3651 .2146 .2017 Conclusions • Clear positive causal relationship between extension participation and farm income • Instruments were proven valid therefore endogeneity is addressed and estimations are more consistent • OLS underestimates benefits of advisory contact • Policy implications: Participation in Teagasc beneficial to clients Effectiveness of extension services should be promoted Future Work • Submit for publication • Disaggregate the analysis into more intensive specific forms of extension and measure impact with greater precision • Examine the role of the adviser and facilities on knowledge transfer • Bolster the management theoretical contribution, particularly in relation to the process of knowledge transfer