Factors Affecting the Choice, Intensity, and Allocation of Irrigation Technologies by U.S. Cotton Farmers Bijay K. Pokhrel1 Krishna P. Paudel Eduardo Segarra
Selected Paper prepared for presentation at the Southern Agricultural Economics Association Annual Meeting, San Antonio, Texas, February 6-9, 2016.
Copyright 2016 by [Bijay K. Pokhrel, Krishna P. Paudel, and Eduardo Segarra]. All rights reserved. Readers may make verbatim copies of this document for non-commercial purposes by any means, provided that this copyright notice appears on such copies.
1
Pokhrel is a graduate student and Paudel is a professor at Louisiana State University and LSU AgCenter, Baton Rouge, Louisiana. Eduardo Segarra is a professor at Texas Tech University, Lubbock, Texas. This work is partially funded by Cotton Incorporated and supported by faculty at the University of Florida, Louisiana State University, Mississippi State University, North Carolina State University, the University of Tennessee, and Texas Tech University. Corresponding author: Krishna Paudel, Email:
[email protected]
1
Factors Affecting the Choice, Intensity, and Allocation of Irrigation Technologies by U.S. Cotton Farmers Abstract We use 2013 cotton precision survey data to understand the adoption of irrigation technologies by cotton farmers in 14 U.S. states. We find that water saving irrigation technologies such as sub-surface drip and trickle irrigation technologies are adopted by farmers who produce higher irrigated yield and by those farmers who are from Southern Plains (Texas and Oklahoma). There are ten irrigation technologies that farmers can adopt in cotton production in these 14 U.S. states. The intensity of irrigation technologies as measured by the number of irrigation technology adopted in cotton production is affected by irrigated cotton yield realized, land holding, education, computer use, and cotton farmers being from Southern Plains. A multivariate fractional regression model is used to identify land allocation under different irrigation technologies. Results indicate that age of an operator, cover crop, information sources used, per acre irrigated yield, education, and cotton farmer being from Southern Plains are significant variables affecting the land allocation under different irrigation technologies. Key Words: Irrigation technology, irrigation area allocation, intensity, multivariate fractional regression, water saving irrigation technologies JEL classifications: D22, Q16, Q25
2
Factors Affecting the Choice, Intensity, and Allocation of Irrigation Technologies by U.S. Cotton Farmers Frequent drought and erratic rainfall patterns necessitate the wise use of limited water resources. Cotton farmers can increase yield and net returns by adopting an irrigation technology, but the choice of an irrigation technology depends on geographical location, land area, and other socioeconomic variables. Cotton farmers have multiple irrigation technologies available at their disposal; some farmers use only one technology while others use multiple technologies. Adoption of trickle and subsurface irrigation technologies such as drip irrigation helps to increase water use efficiency. Farmers' choice to adopt new technologies depends on various factors like education, access to the technologies, age, size of land, and proximity to the urban area (Koundouri, Nauges and Tzouvelekas 2006). Other factors affect irrigation technology adoption as well. Some farmers adopt an irrigation technology based on whether it is appropriate for their field and how technology helps them to increase profit potential and avoid the risk associated with fluctuating and unpredictable weather pattern. Farmers have been concerned about decreasing aquifer level, dwindling surface water supply, and erratic rainfall patterns. As a result, in many areas farmers want to be a steward of water resources. This has resulted in switching from furrow irrigation and center pivot system to a more water efficient system such as subsurface drip and other precision mounted irrigation systems (Lichtenberg et al. 2015). Farmers consider one or more technologies, evaluate the usefulness and then decide to adopt the technology that they perceive would meet their economic and environmental goals (Byerlee and De Polanco 1986; Leathers and Smale 1991). Farmers' positive experience with one technology is likely to affect the adoption of that or similar other technologies (Weber 2012). 3
Scarce water and associated high cost of extraction encourage farmers to adopt more water efficient technologies (Dinar and Yaron 1990; Green et al. 1996; Ding and Peterson 2005; Pfeiffer and Lin 2014). Advance irrigation technologies such as subsurface drip irrigation, low energy precision application system, and variable rate irrigation systems need much higher investment than a traditional irrigation system. The invention of efficient water saving irrigation technologies has helped to optimize ground water use. Due to escalating energy costs and a declining water table, farmers would like to adopt efficient irrigation technologies. Physical factors that affect the choice of irrigation technologies are terrain (slope), soil type and type of crops planted. Farmers would adopt an irrigation technology if a given irrigation technology needs little maintenance and provides needed water for crop growth and maximum yield efficiency. In water deficit areas, farmers can utilize water more prudently using water saving irrigation technologies. In fact, in these water deficit areas cotton farmers are changing from conventional to more efficient irrigation technologies (Weinheimer et al. 2013). Additionally, farmers are laser land leveling so that water movement occurs more efficiently from one end to the other end of the field. Farmers are not necessarily restricted by the choice of one irrigation technology. Some land such as land with lower soil quality or with various levels of slopes may require a combination of irrigation technologies for farmers to maximize profit. Likewise, investing money to install water saving irrigation technologies in better soil and level land may not be profitable (Lichtenberg 1989; Dinar and Yaron 1990; Negri and Brooks 1990; Shrestha and Gopalakrishnan 1993; Green et al. 1996).
4
Farmers may decide to adopt a center pivot irrigation technology in one land parcel whereas they may choose to adopt a furrow irrigation technology in another land parcel they are cultivating. Fractions of land allocated under each irrigation system may be impacted by several physical and socioeconomic characteristics. Fractional response variables are frequently encountered in decision making (Papke and Wooldridge 1996; 2008). This necessitates the adoption decision to be analyzed using a fractional regression model. We follow the method developed by Murteira and Ramalho (2014) to identify the impacts of explanatory variables to allocate land under multiple irrigation technologies. The objectives of this study are three-fold. First, we identify the variables that affect adoption of water saving irrigation technologies in cotton farming. Second, we identify why farmers adopt multiple irrigation technologies and factors affecting their decision to adopt multiple irrigation methods. Third, we use a fractional regression model to determine the allocation of land under different irrigation technologies.
Methods Farmersβ decision to adopt drip or trickle water saving irrigation technologies Farmers maximize expected utility by either adopting or not adopting water saving irrigation technologies. If a technology adoption generates higher utility than nonadoption, then farmers adopt the technology and vice versa. Suppose Y is a binary (0 or 1) variable, in our case whether farmers use water saving irrigation technologies or not and explanatory variables X influence the outcome of Y, which is explained by the following expression: (1)
ππ (π = 1|π) = πΉ(π β² πΎ)
5
Where ππ represents probability, πΉ is the cumulative distribution function (CDF) of the standard normal distribution and parameter πΎ is estimated by maximum likelihood. This model can be estimated using a limited dependent variable model such as a probit model.
Farmersβ decision to adopt multiple irrigation technologies Let ππ be the observed number of irrigation technologies used by an π π‘β cotton farmer. The ππ is assumed to be independent with Poisson distribution. The parameters πΌ depend on explanatory variables X which affect the number of irrigation technologies adopted. The conditional mean can be expressed as πΈ(ππ |ππ ) = ππ = exp(πΌπ ππ ) , π = 1,2, β¦ , π where ππ denotes the intensity parameter. We can represent the Poisson density function as (2) ππ(ππ = π€) = π(ππ ) =
π π ππ ππ π
ππ !
ππ = 0,1,2 β¦ , π.
This type of model can be estimated using Poisson, negative binomial or zero inflated Poisson models.
Farmersβ decision to adopt different proportion of land under each irrigation technology Cotton farmers use one or combination of different irrigation technologies, which cover 100% of irrigated areas in their farm. Suppose π = (π¦1 , π¦2 , π¦3 , β¦ . π¦π ) represents the fraction of area to be irrigated using ππ‘β irrigation technology. These dependent variables possess fractional values that lie between 0 and 1 which is explained by a 1 * K vector of explanatory variablesπ = (π₯1 , π₯2 , π₯3 , β¦ . π₯π ) . The solution of this types of variables is addressed by the nonlinear function satisfying 0 β€ π(. ) β€ 1, where π(. ) is non- linear model (Papke and Wooldridge 1996). The conditional mean of dependent variable is explained by 6
(3) πΈ(π|π) = π(ππ½) where π is a known function satisfying 0 < π(. ) < 1 condition, π½ is a k*1 vector and π is a matrix of independent variables. The following function estimates π½ by maximizing Bernoulli log-likelihood (4) πΏπΏ(π½)=βπ π=1 ππ πππ[π(ππ π½)] + (1 + ππ )πππ[1 β π(ππ π½)] where N is the number of cotton farmers. The estimated parameter will only be consistent and asymptotically normal if πΈ(π|π) is correctly defined. For univariate case, a fractional regression model based on quasi-likelihood and logistic conditional mean functions is useful (Papke and Wooldridge 2008; Murteira and Ramalho 2014). Here, regression is done simultaneously by adopting a multivariate specification as cotton farmers select multiple irrigation technologies and those are correlated. The following functional form of generalization of univariate specification to a multivariate specification with multinomial logit link and multivariate Bernoulli distribution (Murteira and Ramalho 2014) is useful for the fractional regression. Let πΈ(π|π) = πΊ(π; π½) = [πΊ1 (π, π½1 ), β¦ . . , πΊπ (π, π½π )] be the M vector of conditional mean function
with
its
componentsπΈ(π¦π |π), π = 1, β¦ . . π,
with
πΊπ = πΊπ (π, π½π ),
where
conditional mean varies 0 < πΊπ < 1 for all m and βπ 1 πΊπ = 1. The following multinomial logit specification is used; (5) πΊπ =
exp(ππ½π ) π βπ=1 exp(ππ½π )
π = 1, β¦ β¦ , π
where πΊπ is the fraction of ππ‘β component of irrigation method used by producers and subsequently it follows the multivariate Bernoulli (MB) distribution (Murteira and Ramalho 2014). Therefore, the individual contribution to the log-likelihood is as follows;
7
πΊππ
π (6) πΏπππΏπ (π½) = βπ π=1 π¦ππ πΏπππΊππ = βπ=1 πΊππ + πΏπππΊππ
Where πΊπ,π = 1-βπ π=1 πΊπ,π . Then the quasi-maximum likelihood estimator is found by maximizing log-likelihood of all cotton farmers (N) π
πΏπΏ(π½) = β πππ πΏπ (π½) π=1
Where π½Μ is consistent and asymptotically normal. Data, Variables used and its justification The 2013 Southern Cotton Farm Survey data is used for this research which contains 14 states (Alabama, Arkansas, Florida, Georgia, Kansas, Louisiana, Mississippi, Missouri, North Carolina, Oklahoma, South Carolina, Tennessee, Texas and Virginia). Data regarding location, years of cotton grown, age, livestock, crop rotation, cover crop, education, computer used for farm management, accessibility of precision equipment, irrigation technology, cotton yield, and household income were collected through the questionnaire. The survey procedure suggested by Dillman (1978) was adopted to collect comprehensive information which comprises of the questionnaire, a postage-paid return envelope, and a cover letter describing the purpose of this survey. These questionnaires were sent to the 13,566 selected producers after removing 272 duplicate addresses based on the 2011 marketing year. Pre-notification letter was sent on January 18, 2013 to each producer mentioning that the questionnaire about precision farming is coming in two weeks, and requested them for the participation. On February 1, 2013 initial mailing of the survey questionnaire was sent to the each producer followed by reminder postcard on February 8, 2013. Again, a follow - up mail was sent to the non-respondent producers on three weeks i.e. February 22, 2013. The second and final mails were sent to the producers explaining the importance of the survey, the questionnaire, and included prepaid envelope and reminder that 8
producers do not have to respond in the questionnaire if they did not grow cotton during the period from 2008 to 2012. Out of 13,566 mailed questionnaires, 75 were non-respondent, 66 returned due to the wrong addresses, and remaining 263 were either having other farming, retired or deceased. 13.68% responded questionnaires were received by July 15, 2013. Table 1 indicates that most of the surveyed cotton farmers (633) adopted center pivot irrigation technology followed by 259 producers adopting furrow, 133 adopting sub-surface drip, 28 adopting big or travelling gun, 24 adopting flood, 12 adopt hand move, 11 adopt slide roll , and 4 cotton producers each equally share solid set/fixed and trickle irrigation methods. New water saving irrigation technologies comprising of sub-surface drip and trickle irrigation technologies are generated for the probit regression. The dependent variable for the probit model is 1 if a cotton farmer has adopted the water saving irrigation technologies, otherwise it is zero. There are in total 10 different irrigation technologies used by cotton farmers in the study region. These irrigation technologies were counted and the intensity is the number of total irrigation technologies adopted by farmers (See Figure 1). This number makes the dependent variable for the Poisson regression model. Similarly, four irrigation technologies denoting three major irrigation technologies and one with all other remaining irrigation technologies are created for the fractional regression analysis (See Figure 2). The fraction of total areas (a continuous variable with values between 0 and 1) under each of these irrigation technologies makes dependent variables in a multivariate fractional regression model. Farm size is considered an important variable determining the adoption of technologies (Martin et al. 2008). Producer age is an important variable determining technology adoption (Paudel, Mishra and Segarra 2012). Younger producers are more likely to adopt a new technology than the older because they are more handy with new technologies and the new
9
technologies would help them to reduce time spent on farming (Feder et al. 1985). It is also expected that producers with college education (bachelor or graduate) are likely to adopt a new technology because of human capital and sound skills acquired during education (Caswell et al. 2001). Adopting an irrigation technology particularly drip or trickle needs heavy investment therefore wealthier producers are likely to adopt these modern water saving irrigation technologies. Regarding higher cotton yield, we expect that it is a possible indicator of land quality and has positive impact on modern irrigation technologies. The computer use is important to keep financial record in the farm. Farmers who use computers for financial processing are more likely to be successful (Mishra, El-Osta and Johnson 1999). It is expected that cover crop preserves moisture (Lu et al. 2000); therefore, producers who plant cover crops are less likely to invest money to adopt water saving irrigation technologies in cotton farming. McCunn and Huffman (2000) divide various states into different ecological regions in order to analyze agriculture total factor productivity differences due to investment in research and development. We have adopted their ecological region definition in our study. We generate dummy variables for each region: Appalachia, Southeast, Delta States, Southern Plains, Northern Plains, and Mountain regions. Substantial number of farmers located in Southern Plains (Texas and Oklahoma) has adopted water saving irrigation technologies. Since Southern Plains frequently face water shortage, farmers in the region are pioneer in adopting modern irrigation technologies in cotton production.
Results Three separate models are estimated based on data and methods described above. The definition of each variable used in the models and their descriptive statistics are presented in Table 1.
10
Farmersβ decision to adopt drip or trickle water saving irrigation technologies The estimated IRRIGATED YIELD coefficient is positive and statistically significant at 1% level which indicates that yield from the irrigated portion of land is an important reason for choosing water saving irrigation technologies for cotton farming. It is generally true that more yield generates more income to the farmers, so they search for efficient water saving irrigation technologies for maximizing farm production and profit. Marginal effects of IRRIGATED YIELD is positive which indicates that adoption of water saving irrigation technologies is increased by 0.06 percent if irrigated yield increases by one pound. Farm location is another variable that affects the adoption of water saving irrigation technologies. SOUTHERNPLAINS comprising of Texas and Oklahoma states have positive and statistically significant coefficient at one percent level. This result indicates that cotton producers from this region realize a higher profit (from yield) from adopting water saving irrigation technologies. The positive marginal effect implies that cotton farmers in SOUTHERNPLAINS increase water saving technology adoption by 191% as compared to cotton farmers in other regions.
Farmersβ decision to adopt multiple irrigation technologies Here, five explanatory variables are statistically significant at 1-10 percent level of significance. The coefficient of the IRRIGATED YIELD is positive and significant at one percent level. The marginal effects of IRRIGATED YIELD for irrigation intensity is positive which shows 0.06 percent more chances of increase intensity as irrigated yield increases by one percent. The coefficient of landholding is positive and statistically significant at 10 percent levels. The marginal effects indicate that intensity of irrigation technology adoption increases by 0.0028
11
percent as land holding increases by one acre. Education shows negative but statistically significant coefficient at a 10 percent level. This result implies the inverse relationship between education and irrigation technology intensity. Its marginal effects indicate that an additional level of education decreases irrigation intensity by three percent. Another important determinant for selecting irrigation intensity for cotton farming is computer use which shows positive and significant coefficient at a five percent level. Additionally, this result implies 17 percent more intensity of irrigation if a farm household uses a computer in a decision-making process. Its marginal effects indicate that those producers who use computers for cotton farm management have 11 percent more irrigation technologies than who do not use computers. Similarly, location specific dummy explanatory variable shows a positive relationship with intensity, and it is statistically significant at one percent level. It indicates that Texas and Oklahoma producers adopt more irrigation technologies (56.6 percent) as compared to cotton farmers from other regions. Its marginal effects imply that cotton farmers from this region likely to adopt 37 percent more irrigation technologies than who are not from this region.
Farmersβ decision to adopt different proportion of land under each irrigation technology Marginal effects from the multivariate fractional regression model are shown in Table 5. Results show that one year increase in farmerβs age increases the proportion of land allocated to furrow irrigation by 0.03 percent. This result is supported by the fact that older farmers are not willing to adopt a new or improved technology because they are at the end of their planning horizon. Cotton farmers in the Southern Plains allocate 14.55 percent less land under furrow irrigation system. This finding is clear because most of the farmers use water conserving technologies in the Southern Plains where water is in short supply, and evaporation loss due to the sun exposure
12
is quite high in furrow irrigation. Cotton farmers who use cover crops allocate 0.36 percent more land in center pivot irrigation method. Similarly, farmers who use more information sources are more likely to use a center pivot irrigation technology. One more use of information source increases the proportion of land allocated to center pivot irrigation technology by two percent. In contrast, cotton farmers who are located in the Southern Plains allocate 14 percent of less land under a center pivot irrigation technology. Cover crop has a negative impact on the allocation of land under water-saving irrigation technologies. The logic may be that if farmers are using drip or trickle irrigation technology, they may not need/want to use moisture conserving additional cover crop expenses in the farming system. Cover crop planting decreases the proportion of land allocated to drip/trickle irrigation technologies by 0.08%. Higher irrigated yield increases the proportion of land allocated to drip/trickle irrigation technologies. For each pound increase in irrigated yield, the proportion of land allocated to drip/trickle irrigation increases by 0.002%. Education also increases the proportion of land allocated to drip/trickle irrigation. Those farmers who are educated are more conscious of natural resources and are likely to choose water conserving irrigation technologies. Farmers in the Southern Plains allocate 29% more land under sub-surface drip/trickle irrigation compared to farmersβ in the other regions.
Conclusions The 2013 Southern Cotton Farm Survey data was used for this research. Three objectives were considered; the first objective was to find factors that affect water-saving irrigation technology adoption which was modeled using a probit model. The second objective was to identify the 13
impacts of selected explanatory variables on the adoption intensity of irrigation technologies. We modeled this problem using a Poisson model. Similarly, a multivariate fractional regression model was used to determine variables that affect the allocation of land under different irrigation technologies. These objectives were explored by using different explanatory variables, viz., AGE, COVERCROP, INFORMATION, IRRIGATED YIELD, LANDHOLDING, EDUCATION, INCOME, AGSHARE, COMPUTER USE, and SOUTHERNPLAINS. Two explanatory variables SOUTHERNPLAINS and IRRIGATED YIELD were statistically significant in the probit model. Because of Southern Plainsβ weather situation and water shortage, farmers generally try to use more efficient irrigation technologies. In the intensity analysis, we found irrigated yield, landholding, education, computer use, and Southern Plains to be significant in the model. Only one variable education showed negative sign. As education level increases, farmers can process information more effectively and decide to adopt technologies that meet their profit and environmental goals. It is likely that they reduce the number of irrigation technologies to only that can generate sufficient profit for them. If the land has various slopes, they may land-level and use only furrow irrigation instead of both furrow and center pivot irrigation technologies. Cotton farmers who use computers for the farm management and from the Southern Plains are more likely to adopt multiple irrigation technologies. Again, this has to do with connecting fieldβs topography and implementing irrigation system accordingly. Results from the multinomial fractional regression model indicated that older farmers are more likely to allocate a higher proportion of land under the furrow irrigation technology. Cover crop increases the proportion of land under center pivot irrigation system but decreases the land under the drip/trickle irrigation system. Higher irrigated yield enhances farmersβ chance of allocating a 14
higher proportion of land under drip/trickle irrigation technology but decreases the chance of allocating land under the other irrigation technologies (other than furrow, center pivot, and drip irrigation technologies). Education increases the chance of allocating a higher proportion of land under drip/trickle irrigation technologies. Farmers in the Southern Plains allocate more land under drip/trickle irrigation but less land on other irrigation technologies. This paper provided knowledge on factors affecting land allocation under different irrigation technologies. To improve more effective irrigation technology adoption by farmers, emphasis should be placed on education and Southern Plains region. Encouraging farmers to adopt cover crop may help to preserve soil moisture, but it may have an unintended consequence of reducing the proportion of land allocated to drip/trickle irrigation technologies. The cost of technology, price of water, and physical feature of land are also important factors in irrigation technology adoption. Unfortunately, our data limitation prevented us from identifying the role of these variables in the irrigation technology adoption decision.
15
References Byerlee, D., and E.H. De Polanco. 1986. βFarmersβ stepwise adoption of technological packages: Evidence from the Mexican Altiplano.β American Journal of Agricultural Economics 68(3):519β527. Caswell, M., K. Fuglie, C. Ingram, S. Jans, and C. Kascak. 2001. βAdoption of Agricultural Production Practices.β Economic Research Service/USDA, AER-792. Dillman, D.A. 1978. Mail and telephone surveys. Wiley Interscience. Dinar, A., and D. Yaron. 1990. βInfluence of quality and scarcity of inputs on the adoption of modern irrigation technologies.β Western Journal of Agricultural Economics:224β233. Feder, G., R. E. Just, and D. Zilberman. 1985. "Adoption of agricultural innovations in developing countries: A survey." Economic Development and Cultural Change 33(2): 255298. Green, G., D. Sunding, D. Zilberman, and D. Parker. 1996. βExplaining irrigation technology choices: a microparameter approach.β American Journal of Agricultural Economics 78(4):1064β1072. Koundouri, P., C. Nauges, and V. Tzouvelekas. 2006. βTechnology adoption under production uncertainty: theory and application to irrigation technology.β American Journal of Agricultural Economics 88(3):657β670. Leathers, H.D., and M. Smale. 1991. βA Bayesian approach to explaining sequential adoption of components of a technological package.β American Journal of Agricultural Economics 73(3):734β742. Lichtenberg, E. 1989. βLand quality, irrigation development, and cropping patterns in the northern high plains.β American Journal of Agricultural Economics 71(1):187β194. Lichtenberg, Erik, John Majsztrik, and Monica Saavoss. 2015. "Grower demand for sensorβ controlled irrigation." Water Resources Research 51(1): 341-358. Lu, Y.-C., K.B. Watkins, J.R. Teasdale, and A.A. Abdul-Baki. 2000. βCover crops in sustainable food production.β Food Reviews International 16(2):121β157. Martin, S.W., R.K. Roberts, S.L. Larkin, J.A. Larson, K.W. Paxton, B.C. English, M.C. Marra, J.M. Reeves, and others. 2008. βA binary logit estimation of factors affecting adoption of GPS guidance systems by cotton producers.β Journal of Agricultural and Applied Economics 40(01):345β355. McCunn, A., and W.E. Huffman. 2000. βConvergence in U.S. productivity growth for agriculture: implications of interstate research spillovers for funding agricultural research.β American Journal of Agricultural Economics:370β388. 16
Mishra, A.K., H.S. El-Osta, and J.D. Johnson. 1999. βFactors contributing to earnings success of cash grain farms.β Journal of Agricultural and Applied Economics 31(03):623β637. Murteira, J.M.R., and J.J.S. Ramalho. 2014. βRegression analysis of multivariate fractional data.β Econometric Reviews (ahead-of-print):1β38. Negri, D.H., and D.H. Brooks. 1990. βDeterminants of irrigation technology choice.β Western Journal of Agricultural Economics:213β223. Papke, L. E., and J. M. Wooldridge. 1996. "Econometric methods for fractional response variables with an application to 401 (k) plan participation rates." Journal of Applied Econometrics 11(6): 619-632. Papke, L.E., and J.M. Wooldridge. 2008. βPanel data methods for fractional response variables with an application to test pass rates.β Journal of Econometrics 145(1):121β133. Paudel, K.P., A.K. Mishra, and E. Segarra. 2012. βAdoption and Nonadoption of Precision Farming Technologies by Cotton Farmers.β Papers presented at the Southern Agricultural Economics Association Meeting, Dallas, Texas. Peterson, J. M., and Y. Ding. "Economic adjustments to groundwater depletion in the high plains: Do water-saving irrigation systems save water?."American Journal of Agricultural Economics 87.1 (2005): 147-159. Pfeiffer, L., and C.-Y.C. Lin. 2014. βDoes efficient irrigation technology lead to reduced groundwater extraction? Empirical evidence.β Journal of Environmental Economics and Management 67(2):189β208. Shrestha, R.B., and C. Gopalakrishnan. 1993. βAdoption and diffusion of drip irrigation technology: an econometric analysis.β Economic Development and Cultural Change:407β 418. Weber, J.G. 2012. βSocial learning and technology adoption: the case of coffee pruning in Peru.β Agricultural Economics 43(s1):73β84. Weinheimer, Justin, P. Johnson, D. Mitchell, J.Johnson, and R. Kellison. 2013. "Texas High Plains Initiative for Strategic and Innovative Irrigation Management and Conservation." Journal of Contemporary Water Research & Education 151(1): 43-49.
17
Table 1. Descriptive Statistics. Variable
Variables definition
Obs.
Mean
St.Dev.
Min
Max
PCENTER
Proportion of cotton land allocated to center pivot irrigation technology
832
0.659
0.422
0
1
PRDRIP
Proportion of cotton land allocated to drip or trickle irrigation technologies
832
0.073
0.221
0
1
PRFURROW
Proportion of cotton land allocated to furrow irrigation technology
832
0.204
0.362
0
1
PROTHER
Proportion of cotton land allocated to other than other than furrow, center pivot or water saving irrigation technologies
832
0.063
0.222
0
1
WATERSAVING
Number of farmers adopting either drip or trickle irrigation methods =1 if yes; otherwise = 0
1812
0.074
0.262
0
1
AGE
Producerβs age in year
1783
55.486
13.399
18
98
AGSHARE
Share of income from agriculture source
1607
73.479
28.342
0
100
COVERCROP
Whether farmer has planted cover crops, =1 for yes; 0 otherwise
1812
20.750
35.089
0
100
COMPUTERUSE
Computer use for cotton production
1731
0.561
0.496
0
1
EDUCATION
Producers final education level
1780
3.374
1.259
1
6
DEPENDENT
INDEPENDENT
18
LANDHOLDING
Total land holding by producers
1812
661.582
1154.129
0
11000
INFORMATION
Source of information used
1718
2.091
1.428
0
8
IRRIGATED YIELD
Average irrigation cotton yield (pounds/acre)
1812
357.162
550.206
0
2573
INCOME
Household income from both farm
1657
3.028
1.495
1
6
1812
0.348
0.476
0
1
And non-farm sources SOUTHERNPLAINS If a farm is located in TX or OK =1, otherwise =0
19
Table 2. Irrigation technology adopted by cotton farmers in the study region. Irrigation technologies
Number of cotton farmers
Percentage of total farmers
Forrow
259
23.06
Flood
24
2.14
Center Pivot
633
56.37
Hand Move
12
1.07
Solid Set / Fixed
4
0.36
Linear Move
15
1.34
Big or Travelling Gun
28
2.49
Slide Roll
11
0.98
Subsurface Drip
133
11.84
4
0.36
Trickle
20
Table 3. Parameter and marginal effects of variables affecting the adoption of water saving irrigation technologies (Subsurface/drip). Variables Parameters Marginal Effects AGE -0.00191 -0.0001915 (0.00518) (0.00052) COVERCROP -0.00216 -0.0002168 (0.00182) (0.00018) INFORMATION 0.00543 0.0005447 (0.04754) (0.00477) *** IRRIGATED YIELD 0.000566 0.000566*** (0.00010) (0.00097) LANDHOLDING -8.83E-06 -8.86E-07 (0.00005) (0.00000) EDUCATION 0.03897 0.0039113 (0.04999) (0.00499) INCOME 0.00447 0.0004488 (0.03939) (0.00396) AGSHARE 0.00151 0.0001517 (0.00218) (0.00022) COMPUTRE-USE 0.20344 0.0204192 (0.14462) (0.01443) SOUTHERNPLAINS
1.90814*** (0.21359)
Note : * p < 0.10, ** p < 0.05, *** p < 0.01
21
0.1915148*** (0.02077)
Table 4. Factors affecting the number of irrigation technologies adoption (intensity) by cotton farmers. Variables Parameters Marginal Effects AGE 0.00211 0.00139 (0.00265) (0.00176) COVERCROP 0.00110 0.00073 (0.00088) (0.00059) INFORMATION -0.00657 -0.00435 (0.02449) (0.01623) *** IRRIGATED YIELD 0.00953 0.00063*** LANDHOLDING
(0.00005) 0.00002* (0.02671)
EDUCATION INCOME AGSHARE COMPUTERUSE SOUTHERNPLAINS
-0.04588* (0.02204) -0.00149 (0.00125) 0.00179 (0.00125) 0.016736**
(0.00004) 0.00003* (0.00001) -0.030402* (0.01773) -0.00098 (0.01460) 0.00118 (0.00083)
(0.07263)
0.11089** (0.04826)
0.56602*** (0.06665)
0.37504*** (0.04576)
Note : * p < 0.10, ** p < 0.05, *** p < 0.01
22
Table 5. Marginal effects of land allocation under alternative irrigation technologies using a multivariate fractional regression model. Variables Furrow Center Pivot Drip/Trickle Others AGE
0.0003954*** (-0.0010421)
0.0009567 (-0.0012222)
-0.000249 (-0.0006159)
-0.001104 (-0.0006952)
COVERCROP
-0.0026419 (-0.0004674)
0.0036591*** (-0.0004844)
-0.0008643** (-0.0003025)
-0.000153 (-0.0002294)
INFORMATION
-0.005197 (-0.0102155)
0.0199511* (-0.011572)
-0.003262 (-0.0060526)
-0.011492 (-0.0071022)
IRRIGATED YIELD
-1.50E-06 (-0.0000236)
0.00001 (-0.0000276)
0.000024* (-0.0000142)
-0.0000325* (-0.0000172)
EDUCATION
-0.0138445 (-0.0104606) -0.0008077 (-0.0091907) -0.0006432 (-0.0005116) -0.0356439 (-0.0286235)
0.0029101 (-0.0123264) 0.0032363 (-0.0105883) 0.0006728 (-0.0005778) 0.0532155 (-0.0335809)
0.0114475* (-0.0063545) -0.000143 (-0.0053082) 0.000022 (-0.0002521) 0.000227 (-0.0195116)
(-0.000513) (-0.00663) -0.002286 (-0.0057319) -0.000052 (-0.0003066) -0.017798 (-0.0169389)
-0.1455292*** (-0.0250497)
-0.137139*** (-0.0423092)
0.2903408*** (-0.0477109)
-0.007673 (-0.0139231)
INCOME AGSHARE COMPUTER USE SOUTHERNPLAINS
Note: Standard errors in parentheses
* p < 0.10, ** p < 0.05, *** p < 0.01
23
1000 600 400
576
215
200
41 4
0
Frequency
800
976
-1
0
1
2 intensity
3
4
Figure 1. Number of irrigation technologies (intensity) adopted by cotton farmers in the sample of 14 U.S. cotton production states.
24
Nos. of Cotton Farmers
250 200 150 100 50 0 Center Pivot
Furrow
Trickle/Drip
Others
Irrigation methods
Figure 2. Irrigation technologies adopted by the cotton farmers in the United States in 2013.
25
