analisis de factores que afectan la adopcion detecnologia
Cien. Inv. Agr. 37(2):101-109. 2010 www.rcia.uc.cl research paper An analysis of factors affecting the adoption of economic and productive data recording methods of Chilean farmers Alejandra Engler,1 and Roger Toledo2 1 Departamento de Economía Agraria, Facultad de Ciencias Agrarias, Universidad de Talca, Casilla 747 Talca, Chile.2 Departamento de Economía Agraria, Instituto de Investigaciones Agropecuarias INIA Quilamapu, Casilla 426 Chillán, Chile. Abstract A. Engler, and R. Toledo. 2010. An analysis of factors affecting the adoption of economic and productive data recording methods of Chilean farmers. Cien. Inv. Agr. 37(2): 101- 109. Integration of the Chilean domestic economy into international markets has created the need to incorporate more technology, information, and management tools and to generate better entrepreneurial skills at the farm level. These changes require the development of strategic capabilities and farmer changes in attitude. The goal for farmers is to be more prepared for the decision-making process and to have adequate evaluation and control systems to face the complexity of the farm business. The literature suggests that using management tools positively correlates with profits and concludes that management skills are positively related to farmers’ well-being. Survey information on 211 farmers from central and southern Chile was used to estimate a probit model where the dependent variable was record-keeping by farmers. The results show that the farmers’ educational level, age, membership a Technological Transfer Group, land leasing, and the farmers’ own perception of their aversion to risk are statistically significant variables in the model. The model goodness of fit is 0.41, and the model has good predictive power for groups of farmers. Key words: Record-keeping, management skills, probability models.Introduction agri-food exporters. However, this goal de- mands developing a more competitive and inclu- sive agriculture along with the agri-food supplyPast decades, the Chilean agricultural sector (Barrera, 2005). To meet this challenge, largehas evolved into a more specialized business by and small farmers will need to incorporate moreincreasing product and export value. In the last technology, information, and management toolsdecade alone, the agricultural PIB increased 1.3 and to develop better entrepreneurial skills, alltimes, while the export value grew 2.7 times of which imply developing strategic capabilities(ODEPA, 2009). The development of the sec- and a change of attitude (McElwee et al., 2006).tor has allowed the country to set the short-runobjective of becoming a member of the top ten Some of the required skills can be achieved by developing effective management strategies,Received July 23, 2009. Accepted August 21, 2009. making informed decisions, and having bet-Corresponding author: firstname.lastname@example.org ter evaluation and control systems for the farm
102 ciencia e investigación agrariabusiness. Controlling economic and produc- of Cer Los Lagos and CEAgroChile, personaltive results of the agricultural activity leads to communication).efficient business management. El-Osta et al.(2007) even conclude that management skills After a decade, only a small portion of the po-are positively related to the well-being of farm- tential farmer population is made up of MC cli-ers. Moreover, there are studies that reveal that ents or associates. The need today is to developcomputer use on the farm is positively cor- and increase farmers´ management abilities, andrelated to profitability, indicating that the use a relevant concern is the capacity farmers haveof information in decision making increases to process information for decision making.business performance (Gelb et al. 2001). Nut- Keeping economic and financial records are thehall (2004) concludes that even though he first steps in promoting modern managementcould not find a relationship between computer tools in the farming business. Record-keepinguse and profitability, computer use does allow is essential in building up management skillsfarmers to methodologically collect, enter, and and making informed decisions. The objectiveinterpret data, all of which has a synergistic of this paper is twofold: first, to identify howeffect on mental decisions. The primary use of popular the practice of keeping digital econom-computers on the farm is record-keeping. Ac- ic and/or productive records is in the Chileancording to Batte (2005), 89.1% of a sample of agricultural sector, and second, to determinefarmers in Ohio, USA indicated that their pri- which factors favor a higher probability of keep-mary use of computers was for keeping finan- ing such records. Understanding of these factorscial and economic records, and 76.7% think will permit policy makers to create incentiveskeeping records is one of the most important to improve farmers´ management abilities andapplications. will allow MCs to focus their marketing strat- egy on farmers who are more willing to adoptGovernmental and private institutions in advanced management practices.Chile have been generating mechanisms tohelp farmers keep and analyze economic To address these objectives, we administeredand productive information. As a result of a comprehensive survey to a sample of farm-these initiatives, several Management Cen- ers from central and southern Chile. The sur-ters (MCs) have been created throughout the vey collected current management and record-country since the late 1990s to implement keeping practices utilized by farmers, as welleconomic record-keeping to support the deci- as farm and farmer characteristics. A probitsion-making process of their associated farm- model was estimated to assess the probabil-ers. Each MC was created to group farmers of ity of implementing a record-keeping system.the same geographic area, having the same or Probability models have been extensively useddifferent agricultural activities. The farmer in adoption literature (technology, quality mea-is visited by a professional of the MC staff sures, standards, and management tools) andonce a month to collect economic information are, therefore, a suitable model for our purposesand to give him/her feedback on the monthly (Ampansah, 1995; Hoag et al., 1999; Batte et al.,expense and income statements prepared by 2005).the MC. Each MC holds an annual workshopto analyze the final economic results of eachcrop. The dynamics of the MCs have allowed Materials and methodsfarmers to (i) know the results of each of theirproductive activities, (ii) improve efficiency Description of sample and data collectionin aspects where they had higher costs com-pared to the group, and (iii) increase profitson the whole. MC managers generally reportthat farmers have increased profits since they Factors that influence the use of computershave been participating in the MC (Managers on the farm or the use of data in the decision-
VOLUME 37 Nº2 MAY - AUGUST 2010 103making process have been analyzed in previ- vary throughout the zone. The most importantous literature (Nuthall, 2004; Batte, 2005; activities are cereals, dairy, and cattle farming.Just et al., 2003). The main conclusions have The three regions accounted for approximatelybeen the relevance of both demographic vari- 33% of the national agricultural production inables and agricultural system factors. Educa- 2008, including 92% of milk, 78% of wheat, andtion, business size, off-farm income, invest- 57% of bovine meat in 2007 (ODEPA, 2009).ments, and regional location have been foundto be particularly significant (Batte, 2005; Justet al., 2003). Byma and Tauer (2007) found Estimation procedurethat, along with higher operator education andlarger farm size, other factors contributing to The dependent variable was constructed as a bi-increased farm operation efficiency were ex- nary variable and assigned the value 1 for farm-tended participation in a farm management ers who kept productive and/or economic recordsprogram and younger farmer age. Using these digital form and 0 for those farmers who did notresults, our study was based on statistical anal- have records of any kind. Based on the approachyses of a survey of a random sample of 400 of rational choice theory (Chai, 2001) in whichfarmers from central and southern Chile, in- individuals are assumed to be purposive and in-cluding the Bio-Bio, La Araucanía, Los Lagos, tentional and in which decisions and actions areand Los Ríos Regions (In 2008, the Los Lagos shaped by rational preferences and constraints,and Los Rios Regions started operating as sep- we argue that keeping records for decision-mak-arate administrative areas. When the survey ing is a choice related to farmer and farm sys-was conducted, both corresponded to the Los tem characteristics. Following this approach, aLagos Region; therefore, the Los Lagos Regionincludes the two new regions in this article). farmer i is likely to adopt records if the utilityIn-person surveys were conducted between of adopting, U i1, is greater than the utility of notMay and August 2006. From the original sam- adopting, U i 0 , as shown in equation 1:ple, only 358 were correctly completed andused in the study. The Bio-Bio, La Araucanía, U = β ′Z i1 + ε i1 (1) i1 and Los Lagos Regions turned out 136, 112, U i 0 = β ´Z i 0 + ε i 0and 110 surveys, respectively. For the study athand, only farmers who keep digital records where Z i1 is a vector of farmer and farm sys-and farmers who do not keep any records were tem characteristics associated with choosingconsidered. By using this selection criterion, to keep records, Z i 0 are characteristics asso-the sample was reduced to 211 observations, ciated with not keeping them, β is a vectorin which 159 farmers declared keeping digital of parameters to be estimated, and ε i1 andrecords, and 52 did not keep any records. Theremaining 147 farmers who declared keeping ε i 0 are random error terms assumed to fol- low a normal distribution. Therefore, therecords in notebooks were left out in orderto generate a strong differentiation between probability of adoption can be expressed asfarmers who use and do not use records in their P(U i1 > U i 0 ) = P( β ´Z i1 − β ´Z i 0 > ε i 0 − ε i1 ) o rmanagement. The criterion was selected using * P( y > 0) = P(ε > −α ´ X i ) , where y * * i states thethe argument that the computer allows farm- i iers to methodologically collect, enter, and in- differences in utility, ε i the difference in er- *terpret data (Nuthall, 2004). Furthermore, on- ror terms, and α ´ X i is the systematic compo-farm information will provide the farmer with nent of the equation that includes all farmer-a decision-making tool to evaluate and control and farm-system characteristics referred tohis/her business (Amponsah, 1995). as the index function (Greene, 2003). The above-mentioned equation is known as the la- tent regression model where the variable yi*Agriculture is considered a traditional occupation is not observable and can be linked to yi = 1in the area studied, although farming systems when yi* > 0. For P( yi = 1) = F (α ´ X i ) , F is the
104 ciencia e investigación agrarianormal cumulative distribution function that size (size), income sources (inc), and leased landdepends on α and X i parameters and vari- (lease). A summary of the variable statistics isables. Parameters of the model were estimat- presented in Table 2.ed by maximum likelihood estimators withNLogit 4.0. It can be observed from the descriptive statistics that 25% of the sample did not keep records ofResults and discussion any kind, meaning that 75% kept digital records of his/ her activities. Was a difference between the two groups regarding their participation inData description MCs and TTGs. In both cases, the proportion of farmers who participated in these initiatives was higher in the digital record-keepers. ThisThe dependent variable, digital record-keeping was expected because farmers associated withas explained in the methodology, was a binary an MC are more interested in using informationvariable whose value was 1 when the surveyed and improving management skills, or they atfarmer kept productive and/or economic digital least consider management as a relevant busi-records and 0 when he/she did not. Table 1 pro- ness factor. Technological Transfer Groups,vides a detailed explanation of the independent which were created to increase opportunitiesvariables used in the model. They were divided for sharing knowledge and technology betweeninto the following two groups: (i) farmer char- farmers, were also expected to influence record-acteristics that included age (age), educational keeping since management tools are one of thelevel (edu), own perception of risk-aversion topics discussed and analyzed in the group(relav), membership in a Management Center meetings along with technological topics. The(MC) and Technological Transfer Group (TTG); age and education level variables also exhib-and (ii) farm system characteristics including ited differences between groups. The groupTable 1. Description of independent variables. Code Description Type of variable lease 1 = Farm uses leased land. Binary variable 0 = otherwise TTG 1 = Belongs to a Technological Transfer Group. Binary variable 0 = otherwise MC 1 = Belongs to a Management Center Binary variable 0 = otherwise Age Age of the producer Continuous, measured in years Edu Categorical variable that indicates the level of education from 1 to 11, where Discrete 1 means no education and 11 means a graduate degree. Relav Categorical variable that measures the farmers’ perception of his/her risk- Discrete aversion. The categories are: 1 = the farmer is willing to take more risks than the mean only on a few occasions, 2 = the farmer is willing to take more risks than the mean on some occasions, and 3 = is generally willing to take more risks. Inc Categorical variable that indicates the proportion of the total income that is Discrete derived from agricultural activities: 1 = low proportion, 2 = close to 50%, 3 = high proportion, 4 = 100%. Size Total hectares dedicated to the business. Continuous, measured in hectares
VOLUME 37 Nº2 MAY - AUGUST 2010 105Table 2. Characteristics of farmers, farm systems, and groups with and without digital record-keeping. Variable Description No record-keeping Digital record-keeping Number of farmers 52 159 25% 75% Farmer characteristics Age Mean age 56 years 47 years Edu Category mean of 5 8 educational level from 1 to 11 (higher category implies higher educational level). Relav Percentage of farmers in a 1.: Will rarely take more risk than the 1. Will take more risk than the risk-aversion category. mean: 54% mean: 35% 2. Will take more risk than the mean on 2. Will take more risk than the some occasions: 31% mean on some occasions: 41% 3. Will generally take more risk than the 3. Will generally take more mean : 15% risk than the mean: 24% TTG Percentage of farmers 12% 40% participating in a Technological Transfer Group. MC Percentage of farmers 8% 45% participating in a Management Center. Farm system characteristics Size Mean size of the 281.9 ha 533 ha productive system Inc Percentage of farmers 1. Low proportion derived from 1. Low proportion derived in each income source agriculture: 13% from agriculture: 14% category. 2. Close to 50% derived from 2. Close to 50% derived from agriculture: 13% agriculture: 9% 3. High proportion derived from 3. High proportion derived agriculture: 23% from agriculture: 36% 4. All income derived from agriculture: 4. All income derived from 50% agriculture: 41% Lease Percentage of farmers who 29% 50% lease land.of record-keepers tended to be younger and to ence in size. Record-keepers tended to havehave achieved a higher educational level. In this larger farms with a mean of 533 hectares, andgroup, the mean educational category was 8 non-record-keepers 281.9 hectares. The in-(technical certification from a professional in- come source distribution was similar in bothstitute), while in the non-record-keeper group, groups, with approximately 75% of farmers re-the mean category was 5 (high school comple- porting that either most or all of their incometion). These results were consistent with Batte’s comes from agricultural activities. However,(2005) findings analyzing computer use by in the non-record-keeper group, the percent-farmers in Ohio, USA. age of farmers whose income was exclusively derived from agriculture was higher. On theWhen analyzing differences in the production other hand, the percentage of farmers whobetween groups, there was an evident differ- leased land was higher in the record-keeper
106 ciencia e investigación agrariagroup. We argue that producers who lease land the same range used for R 2, where a 0.25 valueneed to control better the results of each ac- or lower is considered weak, 0.5 to 0.74 strong,tivity in order to evaluate the convenience of and 0.75 or higher as very strong (Estrella,future leasing. 1998). Moreover, by analyzing the predictive success of the model, it can be observed that it correctly predicted a digital record-keeperFinally, another variable considered in the 92% of the time, and a non–record-keeper 50%study was the farmers’ own perception of their of the time, meaning the model had a high pre-aversion to risk. The question asked how often dictive power. The variables in the model thatthey were willing to take more risk compared were significant included educational level,to other farmers. Twenty-four percent of the age, membership in a TTG, the producer’s rel-record-keepers reported generally taking more ative aversion to risk, and if the farm systemrisks than other farmers, whereas only 15% of had any leased land.the non-record-keeper group made this state-ment. Thus, on average, digital record-keepersdeclared themselves less risk-averse. Previous studies identified age and educational level as relevant factors when predicting the useThe descriptive statistics reveal that there were of information in decision making and com-significant differences between the two groups puter use (Batte, 2005: Amponsah, 1995). Ac-of farmers. The tendency showed record-keep- cording to our results, both variables were high-ers as younger, more educated, less risk-averse, ly significant with less than 1% probability ofand more likely to be leasing land. The prob- rejection. As expected, when educational levelability model in the next section answers the increased, the probability of keeping recordsquestion of how relevant each of these factors was higher. For the age variable, the coefficientwas in determining the use of records. indicates that older farmers were less likely to adopt digital records, a fact consistent with the field observations. In general, new generationsEstimation results of farmers are more conscious of running the farm as a business and of incorporating man- agement tools in their decisions.Estimation results are shown in Table 3. Thecoefficients reported correspond to maximum An interesting feature of the model was thatlikelihood estimators which indicate the great- belonging to a TTG also influenced the sig-est probability of obtaining the observed value. nificance factor, and these farmers were moreThe coefficient signs determine the direction likely to keep digital records. In this case, TTGof the effect of the variables on the probability usually deal with technological issues becausethat the farmer will keep records. The model this is their primary objective, but many ofwas estimated correcting for heteroscedastic- them have been incorporating the use of man-ity using a variance function for MC and inc. agement tools as a technological topic in theirThe distribution of the variance function was regular meetings, a fact which explains ournot significant for the other variables included results. On the contrary, belonging to an MCin the model. The model performance can be is not a significant variable. We expected thatrated as slightly strong because it presents an MCs would have motivated farmers to improveEstrella criterion value of goodness of fit of management abilities and make them more will-0.41. Estrella measures goodness of fit of bi- ing to adopt management tools. However, MCsnary dependent variable regressions by using a keep records for their members and so it maytransformation of the log-likelihood function. be that these farmers do not feel the need toThe measure is comparable to R 2 in terms of its adopt a record-keeping system. The estimationrange and the underlying relationship with the result showed that are neutral in the probabilitytest statistics, and it is therefore a suitable mea- of keeping digital records, which could be ex-sure for the estimated model. The interpreta- plained by the two hypotheses offsetting eachtion of the interior point can be assimilated to other. As seen in the descriptive statistics, there
VOLUME 37 Nº2 MAY - AUGUST 2010 107Table 3. Estimation results of the probit model and goodness of fit measurements. Variable Coefficient Standard error P-value Probability Index function Constant 2.88** 1.23 0.02 Lease 0.42* 0.25 0.09 TTG 1.92*** 0.74 0.009 MC 0.49 0.62 0.42 Age -0.07*** 0.02 0.001 Edu 0.20** 0.08 0.01 Relav -0.34*** 0.13 0.009 Inc 0.08 0.25 0.74 Size -0.0001 0.00008 0.14 Variance function MC -11.63*** 0.28 <0.001 Inc 0.39*** 0.11 <0.001 Log-likelihood -73.749 Restricted log-likelihood -117.822 Estrella =1-(L/L0)-2L0/n 0.41 McFadden R2 0.37 Number of observations a 211 ˆ ˆ ˆLegend: *p<0.1;**p<0.05; ***p<0.01. McFadden R2 is calculated R 2 McF = 1 − (ln Lu /ln Lr ) as, where Lu is the unrestricted ˆlog-likelihood function and Lr is the restricted log-likelihood function. Count R2 is based on the table of observed andpredicted outcomes and is calculated as R 2 Count = 1 / N − ∑ n jj where n jj represents the correct predictions for outcome j. jwas a higher participation in MCs among the of total sales was significant for computer use asdigital record-keeper group. Nevertheless, from a management tool. Contrary to our hypothesis,the aforementioned results, it can be inferred size was not significant. This result revealed thatthat the possibility of sharing the experience of small and large farmers have the same potentialusing management tools, such as in the TTG, to increase their management ability, and the rel-was an effective way to increase management evant variable in this case would be educationalability among producers. level and age, as already shown. Using the same arguments, we would have expected that as longThe lease variable, which indicated whether the as the agricultural activities represented a high-farmer had leased land and had to pay annual er portion of the total income of the farmer, therent, was also significant and positive, mean- probability of keeping records would increase,ing that producers who needed to pay rent were but the inc variable was not significant.more likely to adopt digital records of their busi-ness performance. This result is understandable Finally, a variable was included in the model toin that farmers who lease land to increase their capture the impact of the farmers’ perceptionbusiness were constantly evaluating the conve- of their own aversion to risk. The hypothesisnience of renting or not. Size of the farm was ex- was that as farmers declared themselves morepected to be relevant in record-keeping. Larger risk-averse, they would be more likely to adoptfarms manage more resources and for this rea- records. The variable was scaled in three levelsson it was expected that they would need more where 1 represented more aversion and 3 lessmanagement tools to support decisions. Batte aversion. The results show that the variable was(2005) found that farm size measured in terms negatively significant, thus our hypothesis.
108 ciencia e investigación agrariaConclusion and was not a relevant variable in the decision to keep information. The possibility of shar- ing experience through TTGs increased theThe present study has provided strong evidence probability of keeping digital records. Finally,that the adoption of management tools, repre- younger farmers with higher levels of educa-sented in this case by digital economic and tion were more willing to adopt digital records.productive information records, was affected Given these results, it can be concluded that anby farmer and production system characteris- effective way to incorporate management toolstics. Results were consistent with the theory throughout Chilean agriculture is by directingof farmers as rational because producers who established groups of farmers to into the uselease land and believe they are more risk- of record keeping tool. What is more, youngeraverse show a higher probability of keeping farmers should be targeted, since they presentinformation to make decisions. On the other a higher rate of adoption of computer tools.hand, size was not significant in the model Resumen A. Engler y R. Toledo. 2010. Análisis de los factores que afectan la adopción de métodos registros productivos y económicos entre productores chilenos. Cien. Inv. Agr. 37(2): 101-109. La integración de la economía chilena a los mercados internacionales ha creado la necesidad de incorporar más tecnología, información, nuevas herramientas de gestión y generar nuevas habilidades empresariales a nivel de predio, lo cual implica el desarrollo de capacidades estratégicas y un cambio en la actitud por parte de los agricultores. El objetivo para el agricultor es estar mejor preparados para la toma de decisiones y contar con sistemas de evaluación y control adecuados a la complejidad del negocio agrícola. La literatura sugiere que el uso de herramientas de gestión tiene una correlación positiva con los beneficios, lo que permite concluir que la capacidad de gestión incrementa las posibilidades de buenos resultados. A partir de una encuesta a 211 agricultores de la zona centro y centro sur de Chile se estimó un modelo probit, donde la variable dependiente corresponde a productores que mantienen registros digitales. Los resultados muestran que el nivel educacional del productor, edad, pertenencia a un Grupo de Transferencia Tecnológica, arriendo de tierra y la percepción personal respecto a aversión al riesgo son variables estadísticamente significativas. La bondad de ajuste del modelo es de 0,41, y presenta un buen nivel de predicción, para agricultores con diferente método de registro. Palabras clave: Registro de datos, capacidad de gestión, modelos probabilístico.References ic, and H. Rojas (eds). Economía del crecimiento y Nueva Agricultura. LOM Ediciones Ltda. San- tiago, Chile. 325 pp.Amponsah, W. 1995. Computer adoption and use of Batte, M. 2005. Changing computer use in agricul- information services by North Carolina commer- ture: evidence of Ohio. Computers and Electron- cial farmers. Journal of Agricultural and Applied ics in Agriculture 47: 1 - 13. Economics 27: 565 - 576. Byma, J., and L. Tauer. 2007. Exploring the role ofBarrera, A. 2005. La nueva institucionalidad pública managerial ability in determining firm efficiency. alimentaria. In: Barrera, A., V. Venegas, T. Tom- Paper presented at the Annual Meeting of the
VOLUME 37 Nº2 MAY - AUGUST 2010 109 American Agricultural Economic Association. Greene, W.H. 2003. Econometric Analysis. 5th ed. July 30, 31. Portland, Oregon, USA. Prentice Hall. Upper Saddle River, New Jersey,Chai, S.K. 2001. Choosing an Identity: A general USA. 1026 pp. model of preferences and belief formation. The Hoag, D., J. Ascough, and M. Frasier. 1999. Farm University of Michigan Press. Ann Arber. 336 computer adoption in the Great Plains. Journal of pp. Agriculture and Applied Economics 31: 57 – 67.El–Osta, H., A. Mishra and M. Morehart. 2007. De- Just, R., S. Wolf, and D. Zilberman. 2003. Principles terminants of economic well-being among U.S. of risk management services relations in agricul- farmers operators households. Agricultural Eco- ture. Agricultural Systems 75: 199- 213. nomics 36: 291-304. McElwee, G., A. Anderson and K. Vesala. 2006.Estrella, A. 1998. A new measure of fit for equa- The strategic famer: a cheese producer with tions with dichotomous dependent variable. cold feet? Journal of Business Strategy 27: 65 Journal of Business and Economic Statistics – 72. 16: 198 – 205. Nuthall, P. 2004. Case studies of the interactions ofGelb, E.M., C. Parker, P. Wagner, and K. Rooskoph. farm profitability and the use of a farm computer. 2001. Why is the ICT adoption rate by farmers Computers and Electronics in Agriculture 42: 19 still so low? In: Summary of EFITA 2001 Con- – 30. ference Questionnaires and Symposium. Pro- ODEPA. 2009. Statistical Databases. Available on- ceedings of the European Conference of Infor- line at: www.odepa.cl Website accessed March, mation Technology in Agriculture. 2009.