Careers in arm’s-length contracting: evidence from the Chilean wine-grape market

Abstract

1. Introduction

Imperfect performance measurement creates a transaction cost in the market for labour, and this friction is seen as a fundamental reason for the existence of firms. Costly monitoring and authority mechanisms are organisational responses to associated incentive problems, and these mechanisms effectively define the firm as an organizational unit. In this context, long-run ‘career’ and ‘promotion-based’ incentives provide additional sources of motivation for employees to work diligently and to acquire new skills. In this paper, we investigate the presence of these forms of long-run incentive provision in the context of market contracting between firms and independent suppliers. We find evidence that organisational strategies used by firms to address imperfect performance measurement with respect to their employees, such as career and promotion-based incentives, are also used in arm’s-length relationships within the context of agricultural contracts.

Holmström (1982) identifies implicit incentives embedded in the market for managerial talent—‘career incentives’—as a substitute for direct on-going performance incentives. Similarly, Prendergast (1993) examines the use of promotion ladders as a source of motivation for workers to invest in non-contractible human capital. In both cases, the authors use employment contracts to motivate their theory, and subsequent empirical work similarly focuses on the existence and use of these strategies among firms and their employees. For example, Gibbons and Murphy (1992) test the career concerns model using evidence from chief executive officers’ compensation and stock market performance. Consistent with theory, they find that pay for performance is stronger in years close to retirement. Chevalier and Ellison (1999) study the relationship between termination, performance and seniority for a sample of mutual fund managers. They find that the likelihood of retention of a manager’s job increases with her performance. They also find that the likelihood of termination is more performance-sensitive for younger managers.

The purpose of this paper is to ask whether similar incentives exist for farmers in contracts with their buyers. In particular, we provide evidence for implicit market-based incentives within arm’s-length contracts between independent wine-grape farmers and wineries in Chile. We observe pay-for-performance provisions that depend on measurable characteristics of grape quality, and we test for the presence of complementary implicit incentives that arise from competition in the market for contract farmers. To motivate and guide our empirical analysis, we develop a conceptual framework where wineries use variable contract durations as a mechanism for provision of intertemporal incentives. This mechanism, which operates within and across organisational boundaries, provides effort incentives to farmers and information to wineries about farmer abilities.

Our conceptual framework draws heavily on formal models of career and promotion-based incentives developed within the literature on employment relations, but we apply it to arm’s-length relationships across the farmer–intermediary organisational boundary. We hypothesise that wineries desire supply relationships with farmers who are capable of reliably producing high-quality grapes. Upon finding such a farmer, wineries then make specific investments that can only be protected by commitment to a long-term relationship. Farmers benefit from participation in a long-term relationship and undertake effort early in their careers to signal their ability. This is valuable to the extent that wineries use a sequence of single-period contracts to learn farmers’ types. We see support for this type of progression in our data but also suggest other possible interpretations.

In particular, our conceptual framework suggests long-term contracts are associated with investment in training, potential for superior-quality production (high ability) and high performance under short-term contracts. These predictions guide our empirical analysis in which we estimate the probability of currently having a long-term contract (LTC) or transitioning to one on variables representing ability, performance and investment in high-quality production.

We find evidence suggesting that contracts in the Chilean wine-grape market are structured like a promotion ladder and that more capable and experienced farmers move up the rungs of this ladder over time. The probability of accessing an LTC is associated with past performance and farmer ability. Our findings are consistent with wineries competing for high-ability high-quality producers. These results suggest the existence of implicit market-based incentives that effectively represent ‘careers in farming’ in the sense that farmers must invest and become more able, educated and experienced to increase the likelihood of being selected for an LTC.

This may be a specific example of a more general phenomenon in arm’s-length relationships, where downstream buyers experiment with upstream suppliers through a sequence of repeated short-term contracts. The prospect of promotion to an LTC provides incentives for quality improvement that complement short-term pay-for-performance incentives. Similarly, market knowledge of supplier performance generates competition for the better suppliers, providing long-term incentives to improve quality.

In what follows, we begin by discussing related literature. We then describe the production environment and contracts observed in the Chilean wine-grape market. In the subsequent sections, we present our conceptual framework, followed by our estimation strategy and results. The final section concludes.

2. Related literature

Our empirical setting fits well the prototypical principal-agent setting with buyers using explicit pay-for-performance contingencies to motivate unobserved actions in contracts with independent producing agents (e.g. Mirrlees, 1999; Holmström, 1979; Shavell, 1979). There is a substantial literature on agricultural contracts that uses this framework to examine contract design (e.g. Allen and Lueck, 1992, 1993; Hueth and Ligon, 1999, 2002) and the effect of contracts on performance (e.g. Goodhue et al., 2003; Paul, Nehring and Banker, 2004), access to credit and the overall welfare of market participants (e.g. Bellemare, 2012; MacDonald, 2006; Miyata, Minot and Hu, 2009).

However, we also observe the significant use of long-term contracts, and anecdotal evidence (obtained during our field work described below) suggests that farmers move from single-period to long-term contracts as wineries learn and gain confidence in farmers’ abilities. This behaviour, to the extent it shows up in our data, is consistent with the notion that long-term contracting itself represents a kind of promotion for farmers. If such a promotion increases a farmer’s value to other wineries, it may also generate implicit market-based incentives for exerting effort that complement incentives provided through pay-for-performance contracts. We are not aware of related theoretical or empirical work on agricultural contracts that focuses on these phenomena.

There are several classes of models that provide theoretical justification for promotion-like mechanisms as the outcome of optimal contracting between firms and employees. Gibbons and Waldman (1999) survey early versions of these models, noting that they can be classified into those that consider promotion as a job assignment mechanism and those that consider promotion as an incentive device for current effort or for long-term investment in human capital. These models build on earlier seminal work by Becker (1962) and Hashimoto (1981) who examine incentives for firms and workers to invest in general and firm-specific human capital. In the conceptual framework and empirical analysis presented below, we focus exclusively on promotion as an incentive device. We argue that farmers make firm-specific investments, and we describe interactions between promotion and pay-for-performance incentives that provide motivation to make these investments.

In career-concern models, market participants learn symmetrically about a worker’s ability over time by observing each period’s output, and firms compete for high-ability workers. This competition generates incentives for workers to exert effort because the market draws inference about ability from the observed output. Career concerns were first proposed by Fama (1980), who suggested that the moral hazard problem that managers face inside of firms is resolved by market forces, so that implicit incentives are sufficient for them to exert effort without the need for explicit contracts. However, Holmström (1982) developed a model of symmetric uncertainty regarding worker (manager) ability and showed that without explicit contracts, managerial behaviour is inefficient. Indeed, managers exert too much effort early in their careers when their ability is unknown, but as the market learns about their ability, they have less incentive to exert effort. Subsequently, Gibbons and Murphy (1992) added explicit contracts to the Holmström (1982) model and showed that optimal incentive contracts with career concerns can induce efficient behaviour by agents. This is the case with linear contracts that include a performance-contingent payment. This payment is lower early in the agent’s career when implicit incentives from career concerns to exert effort are high and market knowledge about the agent’s ability is diffuse; it is higher later in the agent’s career when career concerns are less relevant and ability is revealed to the market.

Promotion-based incentives motivate agents to undertake costly effort or to invest in relationship-specific human capital based on the prospect of higher wages within a firm. In contrast, ‘career-based’ incentives arise from bidding by outside firms for talented workers. Whilst early models analysed career incentives separate from promotion-based incentives, several recent papers have sought to integrate the two. Waldman (2013) contrasts ‘market-based promotions’ with the traditional tournament-based promotion model, focusing on how these two types of mechanisms might be identified empirically.¹ His general conclusion is that much of the existing evidence on promotion incentives is consistent with both kinds of models. Prasad and Tran (2013) develop a model that identifies job flexibility and complexity as a potentially measurable set of job attributes that might be used to identify the presence of market-based promotion incentives. The authors show that when jobs require employees to make a relatively large number of non-contractible firm-specific investments, firms can offer to invest in general training in return for employees making specific investments. When promotion is contingent on general and specific training and is observable in the labour market, firms must pay a market premium to keep workers in the firm. In this way, promotion programmes, combined with public observability of job status and an external labour market, combine to generate incentives that are a blend of traditional promotion and career incentives.

Theory provides a variety of motivations for long-term contracts. One view sees them as a form of ex ante protection against potential future hold-up of relationship-specific investments when contracts are incomplete (e.g. Klein, Crawford and Alchian, 1978; Williamson, 1985; Grossman and Hart, 1986). Alternatively, an optimal complete contract will generally include dynamic incentives when information about current-period unobservable actions is contained in future-period outcomes (e.g. Lambert, 1983; Rogerson, 1985; Chiappori et al., 1994). The evidence we examine in this paper suggests that long-term contracts serve a dual purpose as protection for investments that wineries make and as inducement for farmers to perform well under short-term contracts.

Below, we develop a conceptual framework that incorporates market-based promotion incentives. We use our framework to formulate several hypotheses relating farmer and firm characteristics in current and past period interactions to the likelihood they are in an LTC in the current period. Our analysis focuses on the dual roles that long-term contracts can fulfil as a source of protection for buyers with respect to specific investments they make in sellers or employees and as an inducement offered to suppliers who contribute specific investment and on-going work effort. Furthermore, we demonstrate how market-based incentives can interact with performance incentives in the context of arm’s-length contracts. Our main contribution is to provide empirical evidence for implicit market-based incentives within arm’s-length contracts in the context of agricultural contracts, specifically between independent wine-grape farmers and wineries in Chile. Before presenting our analyses and results, we briefly describe the contracting environment.

3. Contracting environment

Our data come from two wine valleys in Chile. We conducted preliminary interviews with key industry participants in each valley during 2010 and then administered a formal survey to wine-grape farmers in the following year (2011–2012 crop year). We report on our survey data below, but first we briefly describe relevant features of the production and marketing environment.

3.1. Wine-grape quality

Wine quality depends not only on the physical and environmental conditions of a given farm, but also on the actions of grape farmers and wineries. The contracts we investigate are, at least in part, designed to motivate costly and difficult-to-monitor actions that wine-grape farmers can take to improve wine quality. There are several ways a winery can attempt to predict potential wine quality attainable from a given vineyard. Important field and vineyard characteristics include balance between crop load and canopy capacity, uniformity of ripening, sun exposure, disease and observable berry characteristics such as size, colour and taste. Relevant post-harvest chemical measurements of the berries include total soluble solids, acidity and total anthocyanins (to test for colour in red berries). For a given terroir², farmers can influence grape quality with their cultivation and growing practices. These include irrigation, insect and disease control, pruning, training system and thinning (Winkler et al., 1974).

3.2. Characteristics of contracts

In general, wineries in the Chilean wine-grape market do not produce enough grapes to cope with international demand, so it is common to buy from third parties (Mora, 2019). The percentage of grapes bought by each winery compared with own supply varies by winery. Similar is the case of the percentage of grapes bought through long-term contracts versus other contracts. For example, Lima (2015) indicates that in 2013 around 34 per cent of the grapes Concha y Toro winery bought were under long-term contracts whilst for San Pedro winery this percentage was 15 per cent that year.

Contracts vary depending on length, price, product quality requirements and technical assistance provided by buyers. There is not much information in the literature regarding types of contracts present in the Chilean wine-grape market. According to Lima (2015), there are three types of contracts between buyers and wine-grape producers in the Chilean wine industry: an LTC, a one-year contract and the spot market. The LTC is generally signed with farmers whose land has exceptional conditions to produce high-quality grapes. In this contract, conditions are established for vine management during grape growth, such as maximum yields and handling requirements for harvest, and a high price is set per kilogram of grapes. These contracts are used increasingly in the Chilean wine-grape market (Mora, 2019). Lima (2015) also indicates that the one-year contract is signed between small or medium grape farmers and brokers or wineries. Farmers in the one-year contract cultivate both traditional and fine varieties. This contract is generally offered between the grape growing period and the harvest, and it is not renewed yearly. This contract is also signed with a base and a maximum price. The spot market arises during harvest time. In this market, there is no grape quality classification. The farmer sells to a broker or winery by kilogram, regardless of probable alcohol degree.

We complement this information with the information obtained from our random sample of 184 wine-grape farmers located in the Colchagua and Maule valleys. Grapes are mostly sold through written contracts between wine-grape farmers and wineries. In our sample, we observe that 80 per cent of current contracts are written. Of those, most farmers (71 per cent) signed a contract only with one buyer. Contracts normally consider more than one plot with farmers reporting cultivating on average 2.5 plots.

Contract provisions vary, but commonly they refer to a minimum or a specific sugar content, disease, pest, chemical residue and foreign material (e.g. stems and leaves) tolerances. Additional provisions can include yield-per-hectare target or a yield upper bound, harvest timing, sales exclusivity, thinning grape clusters, maximum decay percentage or no decay at all, colour, acidity or pH and payment scheme. The payment scheme normally consists of a fixed price per kilogram of grape produced and instalment payments. It is uncommon to observe payment per area. In locations where the contract is signed early in the season, wineries may establish a base price and modify that price later in the season based on the actual market price.

When offering a contract to a farmer, wineries consider terroir, but also other characteristics such as crop management practices and farmer characteristics, especially when offering long-term contracts as we will see below. When purchasing via contract, we observe in our sample that bonuses for high quality production are almost exclusively associated with long-term contracting. Five per cent of farmers reported being offered a quality bonus and 24 per cent of farmers reported that there was some quality classification of their grape at contract signature.

In this research, we acknowledge the definitions provided by Lima (2015) and provide further ones based on our fieldwork findings. So, for the purpose of this paper and based on our field data, we distinguish among relationships governed by an LTC, a one-period contract (OPC), which would be equivalent to the one-year contract specified above, a repeated one-period contract (ROPC), which is not considered in the definitions above, and no contract (NC), which is similar to the spot market defined above.

LTCs are signed for several years with an average duration of 6 years and specify prices by zone, variety and trellising/training system, among other characteristics. Early field interviews revealed that this type of contract is usually signed after an OPC has been signed for the previous season, which is suggestive of the presence of a promotion ladder. Furthermore, we were told that LTCs are usually associated with higher-quality grapes, have more demanding quality requirements, provide more training to farmers, involve more monitoring of farmer production practices by wineries and pay prices that are typically higher than the prices paid in other types of contracts.³ Approximately 5–10 per cent of hectares in Chile are sold under LTCs.⁴

The OPC is signed for one production season. It specifies a fixed price per kilogram, by zone and variety. In this type of contract, farmers usually receive some technical recommendations from buyers. The ROPC repeats the OPC for two or more periods, in other words, this contract is signed each season with the same buyer, and the contractual conditions are similar to OPCs.

When there is NC, transactions occur on a ‘spot market’. For example, brokers put a truck in the main square of a town or city and farmers bring their wine grape and sell it directly to a broker or buyer of a winery. Usually, farmers do not receive technical support if there is NC, and they get the lowest price compared to the other types of contracts. In our sample, we only have one farmer belonging to this contract category. The information we just described for the spot market is anecdotal, based on expert interviews in the field.

3.3. Presence of a promotion ladder and careers

To provide further motivation for the analysis that follows, here we provide further description of the apparent career dynamics present in this market. To do so, first we classify contracts according to requirement level, technical assistance level, quality indicators and price. We measure technical assistance using a binary variable equalling 1 if the farmer reported receiving training from the winery during the current contract. To quantify the ‘level’ for this assistance, we use a 1–5 scale with 1 being low and 5 being high.⁵ For quality indicators, we asked the farmer if his wine-grape variety was graded varietal, reserve, premium, A, B, C or whatever grading system the buyer used, if any (there is no standardised grading system throughout the market). Based on our experience and on information collected in interviews with wine representatives along the value chain, we classify the high-quality grapes as varietal, reserve, premium, super premium, A and B; any others we consider low quality. We also asked if the farmer was offered a bonus for quality production at the time of contract agreement.

We observe that quality, prices, requirements and training are all increasing as we look at the characteristics of farmers and contracts starting with NC and moving to OPC, ROPC and LTC (see Table 1). A bonus for producing quality is only offered at the top of the ladder in an LTC where 29 per cent of farmers are offered a bonus for quality production in the current contract. High-quality grading is more frequent in LTCs than in the other type of contracts. Indeed, 67 per cent of farmers in LTCs had their grape classified as high quality at the beginning of the contract, compared to 13 per cent in ROPCs and 4 per cent in OPCs. On average, prices are also much higher in LTCs compared to the other contracts with an average price of CLP $321 per kilogram of grapes.⁶ However, prices in OPCs and ROPCs do not differ by much. The requirement level is also increasing as farmers progress through the ladder, with a minimum of 1 out of 5 in the NC and an average of 3.6 in the LTC (5 being the most demanding level). This is an indicator that farmers in the higher levels of the ladder must accept greater responsibility.⁷ In addition, training and training level are also increasing as farmers advance up the ladder with 100 per cent of farmers receiving training from the buyer in LTCs, 58 per cent in ROPCs and 29 per cent in OPCs. The one farmer without a contract did not receive training by the buyer.

We also observe that most farmers have repeated relationships with their buyers with 71 per cent in ROPCs. Thirteen per cent of farmers are (or were) in LTCs in the season prior to our survey. Fifteen per cent had a one-season agreement with their buyers, and only one farmer did not sign a contract or have a verbal agreement with his buyer. We also find that 6 per cent of farmers transitioned from a short-term contract to an LTC.

Considering the farmers that transitioned from a short-term contract to an LTC, 27 per cent contracted with a different winery, while the rest contracted with the same winery in previous and current contracts. If we consider all the farmers in the sample, including those that did not transition to another type of contract, 80 per cent contracted with the same winery in previous and current contracts.

We also refer to potential benefits and costs of signing LTCs compared to other types of contracts. Benefits of long-term contracting in the wine-grape market in Chile imply at least higher prices, having a stable buyer and stable payments to the farmer over time. Costs imply greater responsibility in the sense of following the cultivation practices indicated/taught by the winery and being committed with one winery throughout the length of the contract. Since LTCs are associated with greater prices it is common that farmers prefer signing an LTC over not signing it. Indeed, only 4 per cent of farmers in the sample that received a recent LTC offer (considering the last three seasons) did not sign the LTC. The main reason for not doing so is to not want to commit to one single buyer.

The market structure described above suggests the existence of two ‘employment-like’ mechanisms within arm’s-length contracts between farmers and wineries:

1. Promotions. Wineries experiment with short-term contracts, learn farmers’ types and then reward high-ability farmers with farmer-specific investments and long-term contracts.

2. Career Incentives. A farmer’s ability is revealed implicitly when he/she is offered an LTC, and this information diffuses through the market, increasing the surplus he/she can earn in relationships with other wineries.

This description provides a novel rationale for long-term contracting, in the context of agricultural contracts, that is distinct from traditional motivations that focus exclusively on the protection of specific investments or implementation of dynamic incentives through full commitment to long-run contracts. Our objective in what follows is to explore evidence for these mechanisms more systematically. Before doing so, we present a conceptual framework to guide our analysis.

4. Conceptual framework

We focus on three salient features of the contracting environment for wine-grape farmers and wineries as described above:

1. Performance is not fully captured in contractible measures of grapes at the point of sale or transfer from farmer to winery. It takes time for wineries to learn the characteristics of grape quality that manifest in the finished and matured wine product.

2. There is mostly symmetric uncertainty between farmer and winery with regard to potential suitability for high-quality grapes. Wineries perhaps have greater knowledge and experience with professional agronomic support and experience across many vineyards, but farmers know idiosyncratic features of their land and vineyard. Information from both sides is used during early periods of a contract relationship to learn and adapt management practices to local conditions in pursuit of desired characteristics in harvest grapes and ultimately in the final wine product.

3. Adaptation occurs through training provided to farmers, and this adaptation is costly to both sides of the relationship. Farmers spend time and effort to acquire knowledge specific to the needs of a particular winery, and the winery incurs costs in transferring knowledge that cannot be recovered if the relationship terminates.

In this setting, short-term contracts are used to experiment and test suitability for high-quality production. Both parties seek a long-term relationship, but the winery will only make this commitment if learning in early stages of the relationship generates a favourable signal of suitability for high-quality production. It is in the farmer’s interest to work hard towards generation of a favourable signal. Incentives for high-quality production are implicit in the possibility of a long-term relationship and disappear once the parties commit to an LTC. At this point, the winery must substitute implicit incentives for explicit quality-based performance bonuses, even if performance metrics do not fully capture all the matters for high-quality wine production.

Relative to existing models that provide a rationale for the use of long-term contracts, this conceptual framework emphasizes the notion that such contracts can serve a dual purpose. The firm protects its investment in the agent’s productive capacity, but only after learning that a farmer is capable of superior-quality production. The prospect of an LTC provides the farmer incentive to work hard to generate a favourable signal during early learning stages of the relationship.

Qualitatively, this framework suggests that long-term contracts are associated with investment in training, potential for superior-quality production (high ability), high performance under short-term contracts and greater use of explicit quality incentives under long-term contracts. In what follows, we examine these predictions systematically for a sample of Chilean wine-grape farmers, interpreting characteristics of farmers and wineries and local conditions for grape growing, as features of the contracting environment. Due to the lack of a direct measure of these parameters, we do not make claims of causal inference nor conduct formal hypothesis tests; rather, our aim is to use our conceptual framework to present one possible interpretation of the statistical description presented below. We interpret progression across contract durations as a kind of ‘promotion ladder’. As information of progression disseminates in the market, a farmer becomes more valuable to other wineries, possibly having new contract opportunities. In this sense, we interpret farmers as having ‘careers’.

As indicated in the introduction, this conceptualisation borrows heavily from existing formalisations developed in the literature on employment relations. We do not deny other possible interpretations or explanations for the results presented below.

5. Data and empirics

We surveyed a random sample of 184 wine-grape farmers located in two wine valleys of Chile, Colchagua and Maule, during a 5-month period in the 2011–2012 growing season. The random sample was selected from an agricultural census list of farmers that included names and approximate locations. We had to conduct an exhaustive field search in each location to find the farmers in the list. In cases where we did not find the farmer or where the farmer passed away and the family did not continue cultivating wine grapes, we searched for a random replacement from the original list. Together with a team of enumerators, we collected data on farmer and farm characteristics, characteristics of the vineyard and crop-management practices, characteristics of the current contract, investments made throughout the history of the crop and the timing of those investments. In addition, we asked farmers about the contract prior to their current contract, including contract characteristics such as training level, requirement level, name of buyer, and year and month of contract start and end.

To look for evidence of implicit market-based incentives we include an indicator of past performance and proxies for the farmer’s cultivation ability, as control variables in the above estimations. We measure past performance with a dummy variable for thinning in the previous contract given that thinning is an indicator of high-quality grape production.⁹ We evaluate several measures of farmer ability including the farmer’s experience cultivating wine grapes, his/her specialised agricultural education and past training provided by wineries. We measure experience of the farmer as the number of years cultivating wine grapes. Our measure of education is a dummy variable that equals 1 if the person in charge of the vineyard has a post-secondary agricultural degree and equals 0 otherwise.¹⁰ For past investments in training, we incorporate a dummy variable for investing in training in the previous contract as a control variable. This training was provided by the buyer. Our perception is that the objective of this training is to ‘convince’ and teach the farmer to conduct practices the way the winery desires. This is the reason why we consider this kind of training ‘relationship specific’. Furthermore, we include the training level in the previous contract to see if training intensity matters with respect to the probability of accessing or transitioning to an LTC. We expect these variables to be positive and significant in both regressions—access and transition to LTCs.

We believe the decision of wineries to offer an LTC and to offer a bonus for producing high-quality wine grapes is joint. However, we do not have enough variation among farmers reporting being offered a bonus across the different contracts, so we only include information about bonus offers in the descriptive statistics presented in the next section.

Given that LTCs provide explicit incentives for high-quality production, we expect farmers to exert effort in these contracts. We think of the cultivation practices as proxies for effort in producing quality. As measures of these practices, we use summer pruning and the presence of diseases such as botrytis bunch rot and powdery mildew (henceforth botrytis and mildew) during the current season. We think of a disease as an indicator of the application of chemical products to avoid pests and diseases.¹¹ We also consider farm size as a proxy for the farmer’s wealth. We include this variable based on the possibility that accessing an LTC requires an investment outlay by farmers.

We incorporate other variables in the regressions that could be associated with wine-grape quality, especially variables that relate to the terroir. Those variables are soil quality, valley of production and the variety cultivated. We also control for past investments conducted by farmers that are likely to improve quality. We measured soil quality by asking the respondents to assess the land quality and to rate it on a 1–5 scale ranging from poor to excellent. The valleys of production could be Maule or Colchagua, where our survey was conducted. As to the variety cultivated, usually farmers cultivate more than one variety, so we ask them to select one of those varieties in their response to the survey. We classified the selected variety as ‘classic’ versus ‘traditional’. Classic varieties are determined by Chilean wine law and include varieties that have the potential for producing higher-quality wines such as Cabernet Sauvignon, Merlot, Carménère, Syrah, Malbec, Chardonnay, Sauvignon Blanc, Sauvignon Gris and Sauvignon Vert (Ministerio de Agricultura, 2012). ‘Traditional’ varieties were introduced in Chile before classic varieties and are representative of ‘quasi-subsistence’ agriculture. The most representative of these varieties is the País variety (similar to Mission in California). We could not include a dummy variable for País in the estimations, which is considered a low-quality variety in Chile (even though some adventurous wine makers are trying to change this perception), because none of the farmers in LTCs had this as their selected variety.

Past investments that are likely to be related to wine-grape quality include modifying the irrigation system, replacing plants, changing varieties, planting new varieties, extending the vineyard area, constructing a dam by the vineyard and investing in the frame/structure of the trellis system. If the farmer conducted at least one of these investments before the contract agreement, the dummy variable for past investments takes the value of 1; if there was no investment conducted before this date, the dummy takes the value of 0.

In addition, we take into account other controls that may relate to the likelihood of accessing LTCs such as buyer characteristics and distance to buyer. As buyer characteristics, we treat selling in the export market as an indicator of relatively high quality. In the export market, low-quality wine is usually sold bulk and higher-quality wine is bottled. We also consider the quantity exported as an indicator of the size of the buyer. We create these two variables based on data provided to us by the National Customs Service for 2011 wine exports (De la Cerda, 2011). Use of LTCs varies across wineries, but we observe 58 different wineries. As a consequence, we do not have enough degrees of freedom to include a dummy variable for each winery, although we do include dummy variables for the two largest wineries in the sample.

6. Results

We begin by presenting descriptive statistics of the variables used in our logit models. Second, we show the results obtained from estimation of the probability of the current contract being long-term and of transitioning to an LTC. These regressions explore the possibility that implicit incentives are empirically important and examine the reasons for long-term contracting in this specific market contracting environment. We also consider the possibility that some of our control variables may be endogenous. Variables such as buyer type, current farm size, current variety, current soil quality (representing investments in high-quality land) and current diseases affecting the vineyard may potentially be determined after contract agreement (or signature). Given the lack of panel data and the lack of variation of some variables in the sample, more advanced procedures to address endogeneity cannot be used. We examine robustness to potential endogeneity by presenting regression results with and without potentially endogenous variables.

We classify variables used as control variables in the regressions into several categories: crop-management practices, terroir and past investments in quality production and farmer characteristics (see Table 2). Regarding crop-management practices, 13 per cent of farmers conducted thinning of wine-grape clusters and 84 per cent pruned the vineyard during the summer in the previous contract. We asked about these practices in relation to the specific variety selected for the purpose of participating in our survey. In addition, 8 per cent of farmers reported the presence of botrytis in the vineyard in the current season and 18 per cent reported the presence of mildew.

Regarding variables representing terroir, 53 per cent of farmers in the sample are located in the Colchagua valley. Over half our respondents chose a classic variety for the purpose of responding to our survey, and this reflects previous investments in cultivating classic varieties. Furthermore, most farmers classified their soil quality as good with an average report of 3.9 on a 1–5 scale. Wine grapes, in general, have higher quality when they grow somewhat stressed, so we did not ask for soil quality directly. Rather, we asked for specific problems of the soil that should be avoided such as the presence of poor drainage, high salinity and very shallow soils (Winkler et al., 1974). In general, these characteristics make it impossible to produce high-quality grapes. We think of soil quality as an indicator of investments in high-quality land. As to past investments for high-quality production (or improvement), 37 per cent of farmers conducted a major investment in their vineyard before signing the current contract. Respondents reported modification of an irrigation system most frequently.

The variables that we consider as tests for career concerns include an indicator of past performance, specifically, thinning in the previous contract, and variables representing cultivation ability such as years of experience, having specialised agricultural education and receiving training from a vineyard in the past. Thirteen per cent of farmers conducted thinning of grape clusters. Farmers, in general, had many years of experience cultivating wine grapes. The average experience was 23.5 years with a standard deviation of 15.6 years. Twenty-two per cent of the farmers or administrators in charge of the vineyard had some amount of post-secondary agricultural training. This could be either at the associate or professional level. In addition, 55 per cent of farmers received training from their buyer in the previous contract and, on average, they graded the training level with a 2.6 out of 5 with a standard deviation of 1.7.

We also consider farm size as a proxy for wealth. The average farm has 85 hectares with a standard deviation of 155. Given the dispersion of this variable, we use its natural logarithm in the estimations presented below.

Other controls relate to buyer characteristics and distance to buyer. Seventy-seven per cent of buyers export bottled wine which we consider a proxy for high-quality wine production. As a proxy for size of buyer, we use the quantity of wine exported. The average millions of litres exported by buyers is 27.9 with a standard deviation of 39.5. We also include dummy variables for the two largest wineries with whom farmers contract in the sample. Twenty-eight per cent of farmers contract with the largest winery and 5 per cent contract with the second largest winery. The average distance to buyer is 51 km with a standard deviation of 58.

We present the main estimation results in Tables 3 and 4 which contain regressions of the probability of having an LTC in the current season and of transitioning to one. We investigate how variables representing past performance and farmer cultivation ability relate to these probabilities. Columns in Tables 3 and 4 differ with respect to inclusion of one or both of the last two controls (distance to buyer and dummy for largest buyers). Also, the last column in each of these tables shows regression results without potentially endogenous control variables.

As for implicit market-based incentives, farmer ability is associated with an increase in the probability of having an LTC and of transitioning to an LTC, leaving all else constant. Indeed, experience, specialised education and past training are positive and strongly significant in the regression of having an LTC. One extra year of experience is associated with an increase in the probability of accessing an LTC by 5–9 percentage points; also, having a person in charge of the vineyard who has specialised education in agriculture is associated with an increase in this probability by 13–19 percentage points. Furthermore, training provided by the buyer in the previous contract compared to not receiving this training is associated with an increase in the probability of accessing an LTC by roughly 15 percentage points. As to the regression of transitioning to an LTC, ability is also positive and significant, represented by experience and training in the previous contract for the regressions including buyer dummies. One extra year of experience is associated with an increase in the probability of transitioning to an LTC by 7–10 percentage points; higher training level in the previous contract is associated with an increase in the probability of transitioning to an LTC by 2 percentage points. Since contracting with the same winery in the current and previous contracts and not receiving training predict no transition perfectly, we used training level instead of a training dummy in these regressions.

As a robustness check for the relation of past training with the probability of having an LTC, we use training intensity instead of a binary variable. We find that training level is also positive and significant, but the magnitude of the coefficient is smaller (see Table A2 in Appendix). For example, training level is associated with an increase in the probability of having an LTC by 3 percentage points, which is about five times smaller compared to the magnitude of the training dummy’s coefficient.

There are few other changes in estimated coefficients for the probability of being in an LTC when including the level of training instead of a training dummy. For instance, in the regressions of the probability of having an LTC, the significance of years of experience is lower when we use training intensity, rather than an indicator for training. The magnitude and significance of the rest of the variables remain mostly unchanged, although the overall model explanatory power (measured by McFadden’s pseudo R-squared) falls by approximately 4 percentage points.

Past performance, represented by thinning in the previous contract, also relates to an increase in the likelihood of either accessing an LTC or transitioning to one. Specifically, previous thinning compared to not thinning is associated with an increase in the probability of having an LTC in the current season by 20–27 percentage points, ceteris paribus (see Table 3), and is associated with an increase in the probability of transitioning to an LTC by 15–18 percentage points (see Table 4). In general, this association is strong and is robust to alternative model specifications.

In terms of effort exerted under LTCs, we do not find strong evidence that farmers exert higher levels of effort in LTCs compared to short-term contracts as reflected in the production practices of summer pruning and the control of diseases in the vineyard. Furthermore, we find that having botrytis during the current season significantly associates with an increase in the probability of accessing an LTC by 14–20 percentage points, leaving everything else constant. Unfortunately, we did not ask if farmers controlled these diseases; we only asked if they had problems with them. So, the fact that the dummy for having botrytis is positive and significant in the regression of accessing LTCs may be indicating a ‘careful’ farmer who detected the problem and might have taken care of it, more than being an indicator for the application of chemical products. As to the transition to LTCs, not pruning in the summer, having botrytis and buyer not exporting bottles (low-quality buyer) relate to no transition perfectly, so these variables could not be included in those regressions. We replaced having botrytis by having mildew as a proxy for effort in these regressions and found no significant coefficient.

We also control for previous investments in high-quality production, including the characteristics of the terroir, and we find that the binary variable for investing before the current contract, the dummy for selecting a classic variety, soil quality and the valley of production all are not significant in the probability of accessing an LTC. In the regressions of transitioning to an LTC (Table 4), results differ for this set of variables in that soil quality and the valley of production, in general, relate to an increase in the probability of transition.

Furthermore, we tested for the association of wealth with the probability of accessing an LTC and of transitioning to an LTC. We do not find evidence of a wealth association with the probability of accessing an LTC. However, we do find that our proxy for wealth significantly relates to an increase in the probability of transitioning to an LTC.

We also include other controls in the above estimations. In particular, we control for wineries’ heterogeneity and distance to buyer. We control for the former by including buyer characteristics representing quality of winery and by including dummy variables for contracting with either of the largest two buyers in the market. Buyer quality—proxied by a dummy for exporting bottled wine—does not significantly relate to the probability of accessing LTCs. We could not include quality of buyer in the estimation of the probability of transitioning to LTCs because having a low-quality buyer perfectly relates to no transition to an LTC. We also included dummy variables for contracting with the largest two buyers in the market in the regressions. Columns 1 and 3 of Table 3 do not include buyer dummies whilst Columns 2 and 4 do. In general, the effect of including dummies for the largest two buyers does not systematically change the results.

Distance to buyer significantly associates with an increase in the probability of accessing LTCs by 7–8 percentage points, leaving all else unchanged. In other words, farmers are more likely to get offered LTCs if they are located farther from wineries than closer to them.

Finally, when omitting potentially endogenous covariates, we observe that the results of the probability of accessing LTCs are very similar to those including potentially endogenous variables (see LTC 5 in Table 3). We also observe that the results of the probability of transitioning to LTCs do not change much when omitting potentially endogenous covariates (see Transition 5 in Table 4). However, in general, the results of these latter regressions are less robust to different specifications probably due to the low percentage of farmers that transitioned to an LTC from a short-term contract.

Summing up, the results show that past performance and ability are significantly associated with an increase in the likelihood of accessing or transitioning to an LTC. At the same time, wineries provide a progression of increasingly complex task assignments as part of a progression of contract durations.

7. Conclusion

This paper investigates the presence of implicit market-based incentives in arm’s-length relationships between farmers and wineries in the Chilean wine-grape market. We provide evidence suggesting that long-term contracts in this market serve as motivation to farmers for investing in quality improvement. We estimate variables’ relationship to LTC choice by farmer–winery pairs. We test for the presence of implicit market-based incentives for wine-grape quality improvement using variables that represent past performance and farmer cultivation ability (including experience, education and past training). Results indicate that past performance and ability, including past investments in human capital, are all significantly associated with an increase in the likelihood of observing an LTC.

We conclude that, in addition to explicit performance-based incentives, there are implicit market-based incentives for production of high-quality wine grapes. Our evidence also suggests that only farmers with certain characteristics can access long-term contracts. These are the high-ability farmers, which include the highly experienced, the more educated and trained and those who have performed well in the past by investing in yield-reducing, but quality-improving, practices. In other words, the farmers who do not invest and those who are less able, educated and experienced get excluded from long-term contracts. We also find that wealth, in general, does not relate to access to long-term contracts.

There are several important qualifications to consider in our analysis. We argue that long-term contracts also serve as a protection device for relationship-specific investments by buyers, but we do not examine this possibility directly. We do, however, find that farmer–buyer pairs with greater distance between one another are more likely to use long-term contracts. There is insufficient variation in our data to be able to exploit the panel structure of the data that stems from information on the history of contracts. Nevertheless, to the best of our knowledge, this is the first paper to consider the possibility of implicit career and promotion-based incentives for farmers in the context of agricultural contracts.

In the future, we plan to investigate why some farmers do not invest to increase the likelihood of accessing an LTC given that wealth does not seem important. If there is a quality–quantity trade-off, farmers may not want to invest in production practices that are yield-reducing but that favour quality, such as thinning of grape clusters, because the increase in the price they receive is not sufficiently high to support the yield loss. Perhaps risk aversion or entrepreneurial character can provide some explanatory power. Additionally, it would be useful in future research to investigate the determinants of the price that farmers get and how prices are established across contracts.

Acknowledgements

We thank two anonymous referees for their valuable comments which helped us improve the paper. We also thank Bradford Barham, Michael Carter, Jean-Paul Chavas, and Charles Trevor for helpful comments on an earlier draft of this work. We also benefited from conference and seminar participants at the American Association of Wine Economists Annual Conference, the Annual Meeting of the Sociedad de Economía de Chile, the Departments of Agricultural Economics at Universidad de Talca and at Pontificia Universidad Católica de Chile, and at the Development Economics Workshop, University of Wisconsin-Madison. We are grateful to the Rural Territorial Dynamics Program, implemented by Rimisp-Latin American Center for Rural Development and funded by the International Development Research Center (IDRC, Canada), for providing support during fieldwork in Chile. We thank the Division of International Studies, the Latin American, Caribbean, and Iberian Studies Program (LACIS), and the Graduate School at the University of Wisconsin-Madison, for providing additional financial support.

Funding

Funding for fieldwork data collection in Chile was provided by the Rural Territorial Dynamics Program, implemented by Rimisp-Latin American Center for Rural Development and funded by the International Development Research Center (IDRC, Canada). Additional funding for fieldwork was provided by the Division of International Studies, the Latin American, Caribbean, and Iberian Studies Program (LACIS), and the Graduate School at the University of Wisconsin-Madison. The findings and conclusions in this publication are those of the authors and should not be construed to represent any official USDA or U.S. Government determination or policy. This research was supported in part by the U.S. Department of Agriculture, Economic Research Service.

Supplementary data

Supplementary data are available at ERAE online.

Footnotes

References

Appendix