1 World Development Vol. xx, No. x, pp. xxx xxx, 2011 Ó 2011 Elsevier Ltd. All rights reserved X/$ - see front matter doi: /j.worlddev The Impact of Fair Trade Certification for Coffee Farmers in Peru RUERD RUBEN Radboud Unversity Nijmegen, The Netherlands and RICARDO FORT * Grupo de Analysis para el Desarrollo, Lima, Peru Summary. Smallholder farmers producing for Fair Trade market outlets are usually considered to benefit from better prices and stable market outlets. However, many empirical studies verifying this impact suffer from strong selection bias. This study uses a balanced sample of Fair Trade farmers and likewise nonfair Trade producers of organic and conventional coffee from Peru to compare the net effects on production, income and expenditures, wealth and investments, and attitudes and perceptions. After careful matching, we find only modest direct income and production effects, but significant changes in organization, input use, wealth and assets, and risk attitudes. Moreover, important differences between farmers with early and more recent FT affiliation are registered. Ó 2011 Elsevier Ltd. All rights reserved. Key words Fair Trade, impact, matching, coffee, Peru 1. INTRODUCTION The effects of Fair Trade (FT) certification for coffee producers and their organizations have been analyzed in several recent studies. Different studies provide, however, opposing results. Detailed case studies from coffee cooperatives in Costa Rica (Ronchi, 2002), Nicaragua (Bacon, 2005; Bacon, Méndez, Flores Gómez, Stuart, & Díaz Flores, 2008) and Mexico (Calo & Wise, 2005; Jaffee, 2007; Milford, 2004) state that FT strengthened producer organizations and suggest that FT standards improved returns to smallholder producers, positively affected their quality of life and reinforced the strength of local organizations. Other studies reveal that Fair Trade initiatives improved the wellbeing of small-scale coffee farmers and their families, particularly due to better access to credit facilities and external funds, as well as through training and improved product management (Murray, Raynolds, & Taylor, 2003; Taylor, 2005). FT farmers were also found successful in expanding their production, experienced greater satisfaction with the prices obtained for their crop, and reached improvements in food consumption and living conditions that resulted in a significant drop in child mortality (Becchetti & Costantino, 2008). Recent studies that used more quantitative survey approaches for comparing FT farmers with nonparticipants reached far more critical conclusions. FT organic coffee production in Nicaragua (Valkila, 2009; Valkila & Nygren, 2009) and Mexico (Barham, Callenes, Gitter, Lewis, & Weber, 2011) achieved only slightly better yields but required considerable higher labor efforts. Therefore, the increase in farm income proved to be modest, and many farmers remained in poverty despite being connected to FT organic markets (see also: Bacon et al., 2008). Moreover, off-farm employment and migration opportunities generated more income than coffee production. Arnould, Plastina, and Ball (2009, 2007) argue that economic results of FT participation are unassailable, but effects on education and health are highly uneven. Mendez et al. (2010) conclude that FT farmers may reach higher gross coffee revenues, but net contributions to improved livelihoods remains limited since many farmers could not sell their entire production at guaranteed prices. Therefore, perceived changes in the quality of life show no significant difference between farmers participating in conventional and alternative trade networks. Results of focus group interviews (Raynolds, 2002) suggest that income from coffee sales to FT markets is often insufficient to offset other adverse factors that provoke the perceived fall in life quality (e.g., higher input costs and steadily increasing consumer prices, gasoline and communication costs). The assessment of the net impact of FT has become more important with the recent proliferation of coffee certification schemes based on competitive civic and corporate standards (Bitzer, Francken, & Glasbergen, 2008; Conroy, 2001; Giovannucci & Ponte, 2005). Farmers increasingly get options to sell (part of) their production under labels of responsible trade (Utz-Certified), sustainable trade (Rainforest Alliance) or to deliver directly to commercial coffee companies like Starbucks (C.A.F.E. Practices) and Nestle (Nespresso AAA) that aim to guarantee sustainable sourcing under private labels. Consequently, the original rationale for establishing FT as an alternative marketing system that enables smallholder cooperatives to escape from existing monopolistic trade relationships is challenged. Whereas FT initially played an important role in opening-up market outlets and supporting smallholder organization as key elements in the strategy for combating rural poverty and environmental degradation, the current development of the coffee market focussing on quality competition requires reliable supply chain relationships and contractual procedures for improving farm management * Field research for this study has been funded with a contribution from Solidaridad, The Netherlands. Earlier versions of this paper were presented at seminars at the World Bank, the German Development Bank, the Nijmegen Institute for Socio-Cultural Sciences and Wageningen University. We gratefully acknowledge the collaboration of the directorates and farmers of the coffee cooperatives in Junin department of Central Peru. Final revision accepted: July 12,
2 2 WORLD DEVELOPMENT procedures and product handling processes (Petkova, 2006). Moreover, Western consumers claim to be informed better about the real impact of the FT system on rural living conditions and livelihood strategies of local coffee producers. Certification agencies like the Fair Trade Labelling Organization (FLO-Cert) and Transfair USA need to communicate how the FT label influences on the welfare of local farming communities. Market surveys indicate that consumers willingness to pay for FT coffee mainly depends on credence attributes (Didier & Lucie, 2008), but under the current competitive market conditions the lack of sound information on the effectiveness of FT standards could easily confine FT coffee to a mere niche product (Balineau & Dufeu, 2010). A general limitation of many impact studies is that reliable baseline studies conducted before starting of certification are lacking, making a sound assessment of welfare effects after some period of implementation extremely difficult. But even if such studies are available, statistical corrections for differences in farm household and behavioral characteristics between FT farmers and the control groups are frequently limited to simple ANOVA tests. If smaller and poorer farmers become engaged in FT, farmers with similar characteristics should be used as comparison in order to get an unbiased measure of FT impact. Otherwise, if better-off-farmers are selected for FT it becomes difficult to attribute the changes in their welfare only to FT involvement. The principal objective of this article is to present an un-biased assessment of FT impact by using information on a sample of FT coffee producers in the central region of Peru and comparing them with nonft producers with otherwise similar characteristics and attributes. We distinguish between (i) direct effects of FT standards for farm household welfare (i.e., for coffee production, farm management, labor conditions and prices, household income and expenditures) and (ii) other indirect implications of FT supply chain relationships (like quality upgrading, contract compliance, access to finance, organizational participation, gender roles and loyalty). We therefore rely on Propensity Score Matching (PSM) to control for selection bias. The remainder of this article is structured as follows: Section 2 starts with a comprehensive overview of the evidence on the impact of FT derived from several earlier studies. In Section 3 we present the regional context in Peru where the study has taken place. Hereafter, we discuss in Section 4 the criteria for the selection of the treatment and control groups, as well as the sampling strategy used for choosing farmers within each group. We present the characteristics of farmers, dividing them not only by treatment and control but also according to their production system (organic or conventional producers) and highlighting major differences between these types of producers. In Section 5 we outline the propensity score matching technique to balance the sample of farmers to be compared in each group. Section 6 analyzes the main differences between treatment and controls for the selected outcome indicators, and Section 7 reports about the benefits from the use of the FT premium. Finally, we discuss our major findings in Section 7 and outline the implications for future certification initiatives. 2. EVIDENCE FROM FAIR TRADE IMPACT STUDIES The European Fair Trade Association (EFTA) provides an overview of more than 100 so-called FT impact studies 1 that have been realized since 2000, but only few studies rely on representative field samples or use a rigorous comparison with otherwise similar nonft producers. Most of these studies present case-study material based on interviews with farmers and results from focus group meetings at village level. Conclusions of such studies generally emphasize the positive effects on production, sales and the participation in farmers organizations focusing on the process of capitalization from FT premium payments while little attention is given to changes in individual livelihoods and household-level welfare implications (Raynolds, Murray, & Taylor, 2004; Taylor, 2005). General limitations of these studies are that (a) attention is almost exclusively focused on gross production effects without considering input and labor costs, (b) data collection is limited to the results obtained in the certified crop without considering potential substitution effects with other farm household and off-farm activities, and (c) no real comparisons are made over time (comparing with baseline data) or with otherwise identical nonft farmers operating in the same area. Nelson and Pound (2009) provide a concise overview of the evidence base regarding FT impact in different result areas. Their meta review covers multiple dimensions, like economic effects, environmental impact, empowerment, quality of life and wellbeing, and gender and equity issues. Their main methodological conclusion is that many studies address whether producers are getting higher prices for their products and improved access to credit, but there are few studies which attempt to measure changes in income, expenditure or assets for participating households. Empowerment impacts are also explored in many of the studies (...), but few of the studies assess social impacts in any great depth (...) or impacts on producers or workers in conventional market. Some recent path breaking studies have been published that are based on a comparative assessment of FT benefits including some kind of control group. Bacon (2005) compared Fair Trade, organic, and specialty coffees with respect to their potential to reduce small-scale farmers vulnerability in northern Nicaragua. In this region, most of the surveyed farmers grow both food and cash crops. Coffee revenues are mainly used to build houses, send children to school and provide savings and investments for the future. The study supports the conclusion that access to FT certified coffee markets leads to significantly higher (and particularly more stable) prices paid to the farmers and enables them to improve their livelihoods. Certification proved to have an even greater influence on prices than the altitude (which is related to the quality of production). Field research by Valkila (2009) and Valkila and Nygren (2009) focussing on FT organic farmers in Northern Nicaragua are far more critical. Organic coffee production achieves lower yields and requires higher labor efforts. Therefore, the net increase in farm incomes is very modest for low-intensity coffee producers. Farmers thus remain in poverty despite being connected to FT organic markets (see also: Bacon et al., 2008). Evidence from these studies does suggest that participation in alternative trade networks reduces exposure and vulnerability to variations in commodity prices. Similarly, Raynolds (2002) points to the price premium as a critical element offsetting other adverse conditions that affect household welfare. Farmers linked to coffee cooperatives that sell to alternative markets received more stable average prices and also felt more secure regarding their land tenure. However, even then, three quarters of all surveyed farmers reported a decline in their quality of life. A comparative study on the economic and social effects of FT certification based on sub-samples of around 400 coffee producers in three different countries (Peru, Nicaragua and Guatemala) conducted byarnould et al. (2009) concludes that FT membership has a significant positive effect on traded volumes and coffee prices obtained. The effect of FT affiliation
3 THE IMPACT OF FAIR TRADE CERTIFICATION FOR COFFEE FARMERS IN PERU 3 on social indicators of health and education are more modest and generally only materialize after long-term FT membership. The sample of FT and nonft farmers is, however, strongly biased in terms of differences in coffee areas (FT farmers operate substantially larger coffee areas). Moreover, the statistical method for estimating FT impact by simply including a FT membership dummy into OLS regressions of traded volume, prices, health and educational attainment is considered insufficient, since no tests are made to verify whether the production functions of FT and nonft producers maintain the same functional form. 2 Another comparative study including 470 FT, organic and noncertified coffee producers in Mexico and Central America (Nicaragua, El Salvador and Guatemala) conducted by Mendez et al. (2010) finds that certification leads to higher unit prices and higher gross coffee revenues, but that average coffee volumes sold at certified process remain low and thus little or no discernable effects on other livelihood-related characteristics (education, health, and migration) could be observed. This study also does not correct for farm size and coffee area differences between FT and nonft farmers and relies only on nonparametric (Kruskal Wallis and Mann Whitney) tests to identify simple statistical differences. No judgements can therefore be made about the absolute level of impact derived from FT involvement. In a similar vein, Becchetti and Costantino (2008) conduct an econometric analysis to verify the impact of Fair Trade affiliation on monetary and nonmonetary measures of wellbeing on a sample of 475 Kenyan farmers. The researchers compared a group of FT herbs farmers with a control sample and rely on stepwise regressions to identify significant differences. They find that FT farmers were more successful in diversifying their production, experienced a significant drop in child mortality, improvements in terms of monthly household food consumption and dietary quality, and greater satisfaction in terms of prices obtained for their crop. 3 Methodological problems related to the small relative contribution of FT to household income and the control-group sample bias explain part of the results. Only the results on nutritional intake and perceived changes in living conditions remain robust after controlling for selection bias. Recent studies have expanded on the type of indicators used for the assessment of FT benefits. In addition to income and consumption effects related to the FT impact on poverty reduction, attention is given to other implications for skills and knowledge (Imhof & Lee, 2007), input use and quality management (Balineau, 2008; Ruben & Zuniga, 2011; Renard, 2005), environmental effects (Neilson & Pritchard, 2007) and benefits for the cooperative organization and bargaining power within the value chain (Petkova, 2006; Raynolds et al., 2004; Shreck, 2005). In addition, potential effects of FT involvement are analytically assessed using simulation approaches and game-theory models (Baumann, Oschinski, & Stähler, 2011; Richardson & Stähler, 2007). Results tend to indicate that FT is potentially welfare-improving, but that due to weak substitutability the benefits of FT farmers can easily lead to a loss for nonft farmers that surrender part of their market share. Our study on FT impact amongst coffee producers in central Peru has the advantage that we can compare coffee farmers operating in a broadly similar agroecological and commercial environment (Ruben, 2008). Whereas differences in farm size and coffee varieties are limited, FT benefits can be related fairly directly to the support received from cooperative certification. This creates important advantages compared to crosscountry studies (as conducted by Arnould et al. (2009) and Mendez et al. (2010)) that otherwise employ rather similar data collection approaches. 3. REGIONAL CONTEXT AND SELECTION OF COFFEE PRODUCERS FOR IMPACT EVALUATION To evaluate the impact of FT on coffee producers we selected three organizations located in the Selva Central of Peru, in the provinces of Chanchamayo and Satipo of the central department of Junín. Since the second half of the 19th century, this area was one of the biggest hacienda s economies in the country, with coffee plantations covering vast amounts of land. After the process of land reform in the 1960 s, a cooperative model for the commercialization of coffee started to be developed. The agrarian structure of the region is characterized by the predominance of small-holder and medium-size family farms. While many of the cooperatives were dissolved during the 1990 s (the so-called Parcellation ), some of them survived the structural reforms of that time, and some others have being recently reactivated. Currently, there are around 10 farmer s organizations active in the region. Most coffee farmers operate farms up to 5 6 hectares and generally devote about 0.5 hectare to food production and might also possess some cattle. Coffee revenues represent on average the main household income, but rural livelihoods also depend on offfarm work and some nonfarm income, whereas revenues from remittances are generally low. We selected three FT cooperatives that are labeled as FT1 (Ubiriki Coperative), FT2 (Pangoa Cooperative) and FT3 (La Florida), the latter being one of the first to receive FT certification (see Table A1 in Appendix for further descriptives). An important characteristic is that all of them have the major part of their associates involved in organic coffee production. In terms of location, FT1 and FT3 both belong to the Province of Chanchamayo, and their members live relatively close to each other. Members of FT2, however, are located in the Province of Satipo which is farther away from principal markets and has less accessible roads. Based on the characteristics of the FT organizations under study and relying on personal interviews with their representatives, we decided to select members of three other neighboring organizations as the control group (labeled as C1, C2 and C3). Most of them are younger organizations that have just recently started organic production and sales. C3 (Tahuantisuyo Cooperative) is the only one with FT certification since last year, but their FT sales did hardly occur and they have not being able to spend the FT premium yet due to a judicial problem. The other two groups just started their FT application process at the time of the survey. In terms of location, C3 and C1 are located in the Province of Chanchamayo, while C2 members are living in the neighboring Province of Satipo. This control sample enables us to select farmers that exhibit the likelihood of becoming FT producers but did not yet receive the perceived benefits. We could count with databases containing information on total farm sizes, size of coffee plantations, and year of organic certification (or transition) for all members of these organizations. Using this basic information, we initially randomly selected 60 coffee producers of each organization: 30 organic farmers and 30 nonorganic farmers. The total sample consisted of 360 households, including 180 FT farmers and 180 nonft producers. Sampling based on homogeneous areas was done first for the FT organizations, and after that we selected producers in the same area range as control groups. This setup of the sample allows us not only to estimates
4 4 WORLD DEVELOPMENT impacts of FT versus nonft producers, but also to make a separate comparison for organic and conventional farmers, as well as potential differences between recent and mature FT farmers. We decided to separate coffee farmers in our sample by the type of production mode that they undertake into organic and conventional producers. 4 Table 1 shows some basic characteristics and pre-treatment variables for both groups of farmers and the comparison between treatment and controls. FT farmers in both groups have on average an older and less educated head of the households, are already living for a long time in their localities, have parcels further away from the district s capital, and are participating more in civic organizations compared to producers of the control group. Most importantly, land holdings of FT farmers are on average smaller than the ones for farmers in the control group, in particular for organic producers. These variables, as well as other factors that might be influencing the expected outcomes from FT or influencing the probability of getting the FT certification, have to be taken into account in order to construct a reliable counterfactual for measuring the real FT impact on households livelihoods (see Section 4). The questionnaire used to assess the FT effects is based on a detailed reconstruction of several key socio-economic indicators at the farm-households level. We start to assess the farm-household portfolio of production activities, recording yields and prices of major agricultural activities that enable the calculation of gross and net profit margins. This is aggregated to household level income by including other nonfarm and off-farm activities (including remittances), providing information on the level and composition of total household income and possible trade-offs or complementarities between FT production and other household activities. Full household revenues are calculated from the expenditure budget, based on the standard procedures used in rural poverty studies (Deaton, 1997). This also permits disaggregation into several spending categories (e.g., consumption, transport, education, housing, health care, etc.) to identify possible differences in income use between FT and nonft households. 5 Hereafter, attention is given to the available capital resources (fixed assets, value of cattle, household durables, savings) and possibilities for borrowing (credit access) that provide an indication of relative household wealth. Current and past investments for house upgrading or land improvements are specifically registered to account for changes in the wealth status. Finally, farmers are asked to indicate their subjective appreciation of past and expected future welfare perceptions, for example, whether they consider that their position improved, remained the same or has deteriorated. In addition to these standard welfare indicators, it is considered particularly important to detect behavioral responses, as well as changes at the level of farmers organization that may result from FT involvement. We therefore included some additional dimensions of FT involvement, like attitudinal response related changes in risk attitudes, willingness to invest and self esteem, as well as institutional effects related to the satisfaction with cooperative service provision, the degree of identification with the cooperative firm (Ashforth, Harrison, & Corley, 2008), and the perceived bargaining power of the cooperative. Moreover, we included questions regarding changing intrahousehold relationships and gender roles (particularly concerning key household decisions made alone or together) and on short- and long-run investment in sustainable land use practices. Finally, the survey explicitly considered spillover effects of FT engagements towards other aspects of rural livelihood strategies (i.e., changes in other crops, adjustments in labor use within/outside the household) and externalities of Fair Trade that might accrue to all households in the region (related to FT premium use and changes on regional coffee or labor markets). Combining all these aspects into an Table 1. Characteristics of coffee producers by group Organic Conventional Fair Trade (N = 91) Control(N = 63) t-test (p > t ) Fair Trade (N = 75) Control(N = 91) t-test (p > t ) Mean Mean Mean Mean Household characteristics Age head of household ** ** Education head of household ** Family size Years of residence *** ** Land Area coffee (hectares) *** Area other crops (hectares) Total productive area (hectares) *** Accessibility Time from parcel to capital (minutes) *** Time from house to parcel (minutes) Time from house to capital (minutes) Initial wealth Value household assets until Value agricultural assets until , Participation Member of Organizations before *** ** * Significant at 10%. Significant at 5%. Significant at 1%.
5 THE IMPACT OF FAIR TRADE CERTIFICATION FOR COFFEE FARMERS IN PERU 5 integrated approach permits us to generate simultaneously insights in the effectiveness and constraints of targeting efficiency (i.e., inclusion of the poor and possible leakages to nonpoor), as well as the broad socio-economic implication of FT at different scale levels (e.g., individuals, households, farmers organization and region). 4. MATCHING PRODUCERS TO ANALYZE FAIR TRADE IMPACT The main challenge for impact evaluation resides in the ability to answer the question: What would have happened to a participating household if they would not have joined in the FT scheme? This hypothetical situation known as the counterfactual cannot be observed empirically, whereas just taking the mean outcome of nonparticipants is likely to generate selection bias (Bourguignon, 1999; White & Bamberger, 2008). We rely on a matching approach as a solution for this problem (Heckman, Ichimura, & Todd, 1997; Rosenbaum & Rubin, 1983; Rubin, 1974; Rubin & Thomas, 1996; Smith, 1997). Its basic principle is to identify a group of nonparticipant who are similar to the participants in all relevant pretreatment characteristics. Differences in outcomes between the treatment and the control group can then be attributed to the program. 6 Since identifying all relevant characteristics tends to be complicated, Rosenbaum and Rubin (1983) suggest to use so-called balancing scores, that is, a function of relevant observable characteristics that gives a probability distribution of participating in the program but that are themselves independent of the participation in the program. The propensity score matching procedure for obtaining un-biased groups of FT and non FT farm-households thus needs to rely on intrinsic characteristics that are themselves not influenced by FT participation (Heckman et al., 1997). In order to balance the sample of FT farmers with the controls, we first estimate the probability of having FT certification for the organic and conventional groups, based on a set of exogenous characteristics and pretreatment variables (see Table 2). 7 The probability model for organic farmers reveals a significant effect in terms of the years of residence in the locality, the number of civic organizations in which members of the households were participating before the year 2000, and in the size of their coffee plantation. Only the latter effect is maintained for conventional farmers that also show a significant role of the head of household s age. Even though the explanatory power of the models is considered limited, the inclusion of other variables that could also influence the participation decision and the outcome variables did not improve this model. Based on the results of these regressions we proceed by estimating the propensity score (predicted probability of having FT certification) for organic and conventional coffee producers to identify the regions of common-support (Caliendo & Kopeing, 2005; Rosenbaum & Rubin, 1983). These regions are set after eliminating the observations in the control group with a p-score lower than the minimum p-score in the treatment group, and the observations in the treatment group with a p-score higher than the maximum p-score in the control group. 8 The matching estimation is subsequently performed only on common-support observations. The final distribution of the sample after matching and stratified by type of production system and treatment status resulted in 82% of the organic farmers and 95% of the conventional farmers within the common domain. 9 We relied on three different matching algorithms in order to check the robustness of the results to the method applied: (a) kernel matching that uses weighted averages of all individuals in the control group to construct the counterfactual outcome, (b) one-to-one matching that chooses for each treatment observation the observation in the control group that is closest in terms of propensity score, and (c) nearest neighbor matching that relies on the weighted average of the three closest neighbors in terms of propensity scores for each treatment observation. Given the size of our sample (in particular of the small control group size) and the strong restrictions placed on the common-support, we mainly use the results from the Kernel matching to discuss our findings. We relied on two additional procedures to assess the reliability of our model. First, we repeated the same Probit regressions only considering the observations that remain valid after matching. No significant differences appear anymore between the two groups that influence the likelihood of a particular choice. Second, we plotted the distribution of the estimated propensity scores before and after the matching. 10 Table 2. Factors influencing the likelihood of ft participation (Probit analysis) Organic S.E. Sign Conventional S.E. Sign Age head of household (0.013) (0.012) * Education head of household (0.041) (0.035) Family size (0.060) (0.046) Area coffee (0.037) *** (0.027) ** Area other crops (0.093) (0.092) Time parcel to capital (0.002) (0.002) Value agricultural assets until (0.000) (0.000) Organization membership before year (0.168) *** (0.133) Years residing in locality (0.013) *** (0.010) Constant (0.833) (0.724) * Number of observations = 151 Number of observations = 164 LR chi 2 (9) = LR chi 2 (9) = 21.4 Prob > chi 2 = Prob > chi 2 = Pseudo R 2 = Pseudo R 2 = Note: Standard errors in parentheses. * Significant at 10%; ** Significant at 5%; *** Significant at 1%.
6 6 WORLD DEVELOPMENT Table 3. Comparison FT-non FT organic coffee farmers Variable Ps-match kernel Ps-match one to one Ps-match Neighbor(3) Difference S.E. T-stat Difference S.E. T-stat Difference S.E. T-stat Income Salary income * Self-employment income Gross coffee income 211 1, ,565 3, ,343 2, Gross income other crops Total gross household income 359 2, ,209 3, ,099 2, Total net household income 1,707 1, ,813 2, , Coffee production Profit coffee production 878 1, , , Average coffee price Coffee price weighted average Production coffee (beans) 698 1, , , Production coffee (processed) Coffee yield Wealth Household expenditures 1,395 1, ,540 1, , Value of household durables Value of agricultural assets 2, ** 1,713 1, , * Value of animals stock * * *** Credit access * Amount of credit 2, *** 2,738 1, * 2,928 1, *** Value of savings Value household assets since Value agricultural assets since , ** 1, * 1, ** Investments Have land-attached investments Value of land-attached investments Made land-attached investments Made house improvement Investment in new coffee plants Family labor used in coffee Hired labor used in coffee 2,164 1, * 3,734 1, ** 2, Value of hired labor Perception & participation Economic perception (last years) Economic perception (future) *** ** * Satisfaction with price (range 1 10) *** *** ** Satisfaction with technical assistance *** *** ** Satisfaction with trade (range 1 10) *** ** ** Identification index (range 1 5) ** * Force index (range 1 5) *** * ** Membership of organizations (past) Membership of organizations (current) Perceived land selling price 1,768 8, ,914 11, , Perceived land rental price 3,337 1, ** 3,288 1, ** 2,775 1, ** Gender & environment Number decisions by head household * Number decisions by spouse Number decisions by both Number environmental practices Organic fertilizers in coffee ** * * Organic fertilizers in other crops Chemical fertilizers in other crops Risk attitude Risk attitude (range 1 7) * ** Notes: The force index is constructed from set of five statements (5 points Likert scale) regarding the importance of the cooperative for bargaining coffee sales. The Identification index is constructed from five statements regarding the degree of self-identification with the cooperative enterprise. Satisfaction refers to the degree of satisfaction with cooperative service provision (measured on a 1 10 scale from low to high). Risk attitudes are derived from quantitative risk preferences games yielding outcomes that range from 1 (= risk acceptance) to 7 (= risk averse).
7 THE IMPACT OF FAIR TRADE CERTIFICATION FOR COFFEE FARMERS IN PERU 7 Table 4. Comparison FT-non FT conventional coffee farmers Variable Ps-match kernel Ps-match one to one Ps-match neighbor(3) Difference S.E. T-stat Difference S.E. T-stat Difference S.E. T-stat Income Salary income , * Self-employment income , , Gross coffee income 1, , ,196 2, Gross income other crops * Total gross household Income 4, * 3,238 3, ,433 3, Total net household income 3, * 3,254 2, ,677 2, Coffee production Profit coffee production 1, ,106 2, ,033 1, Price average Price weighted average Production coffee (beans) 1, , ,266 1, Production coffee (processed) Coffee yield ** * Wealth Household Expenditures , , Value of household durables ,266 1, Value of agricultural assets 1, , , * Value of animals stock ** ** * Credit access * ** Amount of credit , , Value of savings Value household assets since , * Value agricultural assets since , , * 1, Investments Have land-attached investment Value of land-attached invest Made land-attached investment * Made house improvement ** ** Investment in new coffee plants Family labor used in coffee Hired labor used in coffee 218 1, , , Value of hired labor * Perception & participation Economic perception (last years) Economic perception (future) Satisfaction with price (range 1 10) Satisfaction with technical assistance Satisfaction with trade (range 1 10) Identification index (range 1 5) Force index (range 1 5) * * Membership of organizations (past) Membership of organizations (current) * * Perceived land selling price 108 3, ,765 4, , Perceived land rental price 1,119 1, ,603 1, ,060 1, Gender & environment Number decisions head household * * Number decisions spouse Number decisions both ** * Number of environmental practices Organic fertilizer in coffee Organic fertilizer in other crops * Chemical fertilizer in other crops * Risk attitude Risk attitude (range 1 7) Overall variability is largely reduced and the distribution of the propensity scores reveals a balanced sample composition that is appropriate for subsequent comparisons (see Annex Figs. A1 and A2). 5. RESULTS Based on the before-described matching procedure, we made a detailed comparison between (a) organic FT and nonft
8 8 WORLD DEVELOPMENT producers (Table 3) and (b) conventional FT and nonft producers (Table 4). Each comparison analyzes significant differences in the defined performance indicators. In addition, attention is given to one specific variable, namely the influence of the duration of FT involvement (see Annex Tables A2 and A3). At first glance, there are few significant differences in income generating activities between FT and nonft organic coffee producers, whereas in conventional coffee only the income derived from other crops is significantly lower. The results of the matching estimation for the conventional coffee farmers even reveals a negative effect of FT on total gross and net household income, even though the difference is only significant when performing the kernel estimation. This negative effect seems to be driven by the significantly lower coffee yields of FT producers as compared to conventional nonft farmers. Even though FT farmers could receive better average prices, these differences are not strong enough to represent a clear welfare effect. The lack of a price effect can be explained by two factors. First, the market prices for coffee were generally high in the year when the survey was undertaken. Second, not all certified FT production could be sold under FT price conditions. While we did not find significant difference in the level of household expenditures, in terms of wealth effects FT farmers present higher levels of animal stocks and have accumulated significantly more agricultural assets in past years. Moreover, access to credit and amounts of loans has increased substantially due to the collateral value of FT delivery contracts. FT farmers have also pursued more house improvements and land-attached investments compared to their counterparts in nonft cooperatives. This seems to indicate that a gradual process of internal capitalization is taking place. Another important effect of FT is observed in terms of farmer s perception about their future wellbeing and their increased satisfaction with the cooperative services provision. FT farmers are not only more satisfied compared to their nonft counterparts in terms of prices, technical assistance and market management services, but they also feel more identified with their cooperatives and are more convinced of its bargaining power. 11 Far fewer differences are found in the areas of gender and environmental care, where hardly any positive effects are obtained. Although we do not find significant differences in perceptions about the sales value of land, FT organic farmers perceived their land having a higher renting value. This is consistent with the limited regional market for land sales that is only active for land rentals. Despite the fact that all organic producers have certification, we find that FT farmers use more organic fertilizers for coffee production than farmers in nonft cooperatives. This difference might be attributed to a better service provision by FT cooperatives in terms of technical assistance and access to agricultural finance. Organic FT farmers also appear with significantly higher risk acceptance, indicating their improved entrepreneurial capacity. Taking a closer look at the group of FT cooperatives we made a comparison of their relative performance, based on separate matching of farmers in each of the FT cooperative (see Annex Tables A2 and A3). Using the same propensity scores from the Probit estimation for each group, we matched the observations for each of the FT cooperative with the closest observation in the control sample. Interestingly enough, we notice that farmers from the oldest FT-certified cooperative of La Florida are driving down coffee profits, mainly because of the relatively small amount of realized FT sales under certified conditions. On the other end, the negative effect of organic FT on household expenditures as well as the significantly smaller amounts of credit received by conventional FT farmers can be mostly attributed to the performance of farmers in Pangoa and Ubiriki, the two youngest FT cooperatives of the sample. Moreover, the payment system used by these cooperatives gathers all organic coffee from its members and pays them an average price of the different markets where it is sold (normally as organic FT, but sometimes also as conventional FT or nonft). 12 In addition to direct income production and income effects, FT farmers most appreciate the potential income stabilization implications, as indicated by the strong effect on risk attitudes observed in Ubiriki cooperative, the most recent FT affiliate. The relatively limited production and modest income effects of FT can be explained by different reasons. First, given the generally high coffee prices farmers are inclined to sell part of the FT coffee to conventional market outlets. Moreover, adverse weather conditions for coffee growing in the year that the survey was undertaken in turn increased local prices for conventional coffee and reduced the price gap with the FT market. Second, part of the coffee production especially in the large La Florida cooperative is found in transition towards organic farming and adjusting to the new production techniques tends to be accompanied by reduced productivity levels. Since the cooperative management is inclined more towards organic production, conventional farmers receive less technical assistance and commercial services. This also explains the lack of significant differences in the satisfaction perceived by conventional FT farmers regarding prices, technical assistance and trading. Third, a longer period of FT affiliation gives rise to increasing specialization in organic coffee production, reducing income generation derived from other crops and off-farm employment. Whereas labor intensity in organic production is generally higher, hired labor use in organic FT production is substantially below what their nonft counterparts use, and FT farmers tend to rely more on family and exchange labor. 6. THE FAIR TRADE PREMIUM Even though the Fair Trade premium is supposed to be one of the most important benefits from the FT certification, Table 5. Benefits derived from FT premium La Florida Pangoa Total % Benefit Avg. value % Benefit Avg. Value % Benefit Avg. Value Technical assistance Education Credit 68 2, ,244 Health services Infrastructure Total value (soles) 3,508 1,768 3,010 Note: 1 US$ = 3.0 soles.
9 THE IMPACT OF FAIR TRADE CERTIFICATION FOR COFFEE FARMERS IN PERU 9 farmers receive limited information about FT premium use. The FT cooperatives invested most of the premium in road improvement, education services (fellowships) and internal loans. More than 10% of the farmers interviewed within the FT cooperatives did not have any knowledge about the existence of the FT premium. Moreover, the number of farmers that claimed to have received any benefit from the use of the premium is less than one fourth of the total sample. The percentage of benefitting farmers increases for the cooperatives with a longer FT certification. Organic farmers appear to benefit most from the FT premium that is used for credit and technical assistance accompanying their transition process. Out of the 31 farmers who received any FT benefit in the cooperatives La Florida and Pangoa, the large majority are organic farmers. Farmers that recognized the receipt of any FT premium benefit were asked to value the received benefits for the last year (Table 5). Adding up the perceived value of the benefits for each household we calculated that the FT premium sums up to 3,000 soles (= 1,000 US$) per household, equal to almost 20% of family income. The perceived total value for farmers in La Florida is twice as high as the value for farmers in Pangoa. This illustrates the cumulative effect of the FT premium favored by longer FT affiliation. Moreover, the improved operations of the FT cooperatives are likely to be funded at least partly from the FT premium. In general, technical assistance and credit are categories where farmers perceived most tangible benefits. Many farmers prefer using the FT premium for individual purposes and tend to undervalue investments made for collective and community services (education, health care, water, and electricity). 7. DISCUSSION AND CONCLUSIONS This study presents important new empirical evidence to understand the effects of FT certification on coffee farmers in the central region of Peru. Using propensity score matching techniques allowed us to construct an adequate counterfactual for the situation of FT farmers prior to their involvement in this supply chain. This permits us to overcome common problems in earlier studies that do not fully and properly correct for selection bias. Given the importance of organic farming in the region, we analyzed the FT impact separately for organic and conventional coffee farmers. Moreover, the FT selected cooperatives have different time spans of involvement with FT certification which enabled an exploration of the importance of the perceived benefits over time (Ruben, Fort, & Zuniga-Arias, 2009). For both groups of (organic and conventional) farmers our results did not show significant effect of FT involvement in terms of higher household income. Net income of conventional FT farmers is partly affected by increased costs of hired labor and reduced revenues derived from other cropping activities and off-farm work. In line with results obtained by Valkila (2009), yield levels of organic FT farmers are slightly higher than their counterparts but no significant difference could be found, whereas a negative and significant yield difference was observed for FT conventional coffee farmers. The lack of a real price difference between FT and nonft producers in both groups seems to be the main limitation for obtaining higher net benefits. The rather limited market for FT sales in the region and the high local coffee prices largely explain this fact. Consequently, FT prices are increasingly considered as a regional floor price offered by local traders to all coffee farmers and thus nonft farmers reap similar benefits as part of an externality effect. Even though, there are some significant difference in household expenditures patterns for FT producers that generally present higher levels of animal stocks, better access to credit and increments in the value of their agricultural assets during last years. In terms of general wellbeing, farmers in older FT cooperatives appear to be better-off than the ones in cooperatives with recent FT involvement. Additionally, FT farmers also invest more in house improvements and land-attached infrastructure than their nonft counterparts. The improvements made in (organic) coffee production reveal another effect of FT in terms of providing more stable income to farmers that enables a gradually shift towards more specialized (organic) farming. 13 These findings largely corroborate FT effects registered in some earlier studies that report fairly modest direct income effects and highly uneven effects on parameters of health, education and migration (Arnould et al., 2009; Mendez et al., 2010; Barham et al., 2011). Similarly, major FT effects are registered in aspects of credit use and internal capitalization. In line with Arnould et al. (2009) we find that FT implications are strongly related to the length of cooperative participation. Since our study captures full household income, we could also register substitution effects in terms of land use and labor, resulting in a stronger FT coffee specialization and a decline in off-farm employment. This is the likely result of the behavioral change in risk attitudes that may be considered as the most important benefit from FT. It also indicated that FT plays a major role in the transition towards more entrepreneurial coffee production practices and attitudes. The lack of many expected effects from FT can at least partially be attributed to the deficient distribution and use of the FT premium as perceived by the farmers. The fact that only a quarter of the total number of interviewed FT producers perceives any tangible benefits from the premium is a clear indication. Moreover, premium investments in social and collective infrastructure benefit FT and nonft farmers alike. In addition, regional markets for coffee and labor are both influenced by the transition towards (organic) coffee production, occasioning generally higher output and input prices. Whereas household-level welfare effects still appear to be limited, FT played an important role in the processes of recovery of the agrarian cooperatives and for the improvement of coffee production. FT farmers proved to be substantially more inclined to make in-depth investments, renting additional land and improving organic fertilizer use. In this respect, FT paved the way for quality upgrading of coffee production. Given the current proliferation of coffee standards and the increasing importance of premium segments, it is likely that FT producers in Junin province become attractive counterparts for delivering coffee under private labels. This is in fact already happening with the oldest FT cooperative La Florida that made the step towards multi-certification and started to deliver also under the Utz-Certified and Starbucks labels. Since sales to FT outlets remain constrained, FT paved the way for acceding more rewarding outlets served by private labels (see Ruben & Zuniga, 2011 for a similar case of competing labels in Northern Nicaragua). Future research should strengthen our insight in the prospects and constraints for efficient and effective coffee certification regimes that tend to be based on a stratification of producers eligible for particular coffee labels related to (a) individual farm-household characteristics, (b) cooperative life-cycle features and (c) regional scale effects. This would permit us to acquire better insight in the key drivers for FT affiliation and the likely interactions between welfare implications and behavioral effects that permit farmers to reap the dynamic benefits of different coffee certification regimes.