Research context: the banana sector in the Dominican Republic
Banana production plays an important role in the national economy of the Dominican Republic, because of its contribution to both the gross domestic product and the Dominican diet—either as a fruit (mature) or as part of a meal (green). The Dominican banana sector is characterised by its strong and dynamic positioning in FT and organic certification. The Dominican Republic is one of the world’s leading exporters of organic and FT bananas, together with Ecuador and Peru (FAO 2018b). An estimated 27,000 hectares are dedicated to the cultivation of bananas, 16,000 of which are exported mainly to Europe and the United States (Espinal 2015).
It is estimated that the sector employs 32,000 workers (ILO 2015). Of these workers, 53% are permanent workers and 47% temporary. Many of the producers do not have permanent workers (almost 30%). The main planting areas are in Montecristi (38%), Valverde (31%) and Azua (27%) (Espinal 2015). The banana production sector in the Dominican Republic contains an estimated 1815 farms, of which almost 60% are small producers with less than four hectares, representing only 12% of the total banana production area. The farms with more than four hectares, as well as those belonging to a plantation, represent around 7% of the total number of farms, and represent 46% of the total banana production area (BAM 2016).
While the banana sector provides a relatively stable flow of income to a large number of workers throughout the year (ILO 2015), the sector faces several challenges with respect to labour rights including the large portion of undocumented migrant workers, constraints on freedom of association, and salaries that are below a living wage.
An estimated 60–75% of wageworkers are migrant workers from Haiti, almost all undocumented (ILO 2015). The percentage of undocumented migrant workers is even higher in Montecristi and Valverde, given the large plantations in the region and the proximity to the Haitian border. This is one of the key challenges the sector faces in terms of workers’ rights. One of the consequences of this informality is that migrant workers (on plantations as well as on small-scale farms) do not have access to social security. While the government, as well as national and international organisations, have been undertaking new efforts recently to document migrant workers, these issues are far from resolved (ILO 2017b; BAM 2015).
A second challenge relates to the lack of worker representation. While the banana production sector is organised at the level of producers, this is not the case at the level of workers (ILO 2017b). In general, few companies have collective bargaining pacts in the Dominican Republic, nor are there any significant unions in the banana sector. Due to several formal and informal restrictions (Smith 2010), few plantations and/or workers are members. On some plantations wageworkers are represented in various wageworker committees, including FT premium committees, but these only have limited influence or bargaining power (ILO 2015).
Another key challenge is that wages are not high enough for workers to enjoy a decent living. Wages and secondary benefits in the banana sector are higher than in other agricultural sectors and above the minimum wage (ILO 2017b). However, ideally a fair wage should be judged according to the living wage definition.Footnote 1 That wages do not yet provide a decent living standard, especially for migrant workers, is evident from the low living standards. For example, based on data from 370 workers, ILO (2015) showed that only 11% of Haitians had access to water (vs. 50% of producers). Also, only 7% of temporary workers had access to healthcare (ILO 2015).
Data and sampling
For our research focus, the relationship between FT certification and labour conditions for wageworkers, we collected data on plantations in Montecristi and Valverde. Most of the plantations, and thus workers, are located in these two regions. There are approximately 22 plantations operating in the northeast region, close to the border with Haiti. To analyse the relationship between FT certification and labour conditions for wageworkers we used data that we collected in the banana sector in 2015 with funding from FT International and one of its member organisations—the FT Foundation in the UK.Footnote 2 Data were collected between February and May 2015. Fieldwork started with interviews with plantation management and other stakeholders; also to develop the appropriate sampling frame. In March and April, a team of local enumerators, trained by one of the authors of this paper, collected data among wageworkers. In May in-depth interviews were conducted with 12 wageworkers (six on FT plantations) and results were discussed in a validation workshop with plantations’ management and wageworker representatives of FT and non-FT plantations.Footnote 3
The data collection occurred at the request of FT to monitor progress in the implementation of FT’s revised hired labour standards in some key banana-producing countries. FT aimed to assess and analyse the difference that FT certification has made across economic, social and voice-related indicators thus far. Next, we discuss how we constructed a realistic counterfactual: what would have happened to the hired labour conditions if the plantation had not received FT certification? This study compares the situation of wageworkers from FT-certified plantations to the situation of wageworkers from otherwise similar but non-FT-certified banana plantations.
At the time of data collection, 14 of these plantations were FT certified, one was in the application process, and seven plantations were not certified.Footnote 4 This region is suitable for organic banana production, and plantations can be characterised by whether they produce organically or not. Organic plantations usually have a different production system and cost structure than conventional ones (i.e. more labour, and fewer chemical fertilisers and pesticides), and they also provide different markets and are confronted with different price fluctuations. All these differences can affect the productivity and profitability of the plantation and hence workers’ wages and conditions. Also, similar types of workers may have different loads and specific tasks in organic and conventional systems, so it is better to compare workers within the same production system. Another important difference among the plantations in this region of the Dominican Republic is their size in terms of workforce. Over half of the plantations in the region have fewer than 100 workers, five have between 100 and 200 workers, and 5 have more than 200 workers (based on primary data and data provided by FT, see Fig. 2).
To obtain a representative sample of plantations we selected FT-certified plantations and for the counterfactual analysis comparable non-certified plantations. First, we deliberately selected a representative sample of five FT-certified wageworker plantations out of the total of 14 FT-certified plantations. The selection was based on three key defining features of the banana hired labour sector in the Dominican Republic as described above: whether a plantation is FT certified,Footnote 5 whether the plantation produces organically, and the size of the plantation in terms of the number of wageworkers employed. All plantations became FT certified between 2007 and 2009. Second, we deliberately selected six comparable albeit non-certified plantations in the same areas, based on the same characteristics: organic versus non-organic production and plantation size in terms of the number of workers. The sample of plantations is well-balanced with respect to organic status, but non-certified plantations are generally somewhat larger in terms of number of workers than the certified ones. Market relations of plantations with their buyers do not differ strongly in terms of the stability of markets.
To obtain a representative sample of wageworkers from the selected plantations we randomly selected 369 wageworkers from the selected plantations; 161 from FT-certified plantations and 208 from non-certified plantations.Footnote 6 Wageworkers were selected randomly from wageworker lists provided by plantation managers, which resulted in a representative sample. For plantations with more than 200 workers we selected around 10% of their workers; for plantations with 100 to 200 workers 25%; and for plantations with less than 100 wageworkers we selected close to 50% of workers.
Table 1 compares wageworkers from FT-certified plantations to those from non-FT-certified plantations. Wageworkers are comparable in several aspects. Twelve percent of the respondents are female. Due to the proximity of the study areas to Haiti and the large immigration of Haitians into the Dominican Republic, more than 60% of wageworkers in both types of banana plantations are Haitians. Migrants’ original households (the ones they left behind) consist of 1.7 members on average. Only 4% of the respondents in both groups own the land they work on.
Table 1 Descriptive statistics of the wageworkers interviewed Wageworkers at FT plantations are significantly different from wageworkers at non-FT plantations in terms of various other characteristics. Compared to wageworkers at non-certified plantations, wageworkers at FT plantations have been employed longer at the plantation they work on, as well as in the banana sector in general, and have been living in the area for a longer time. They are slightly older, their (current) households are smaller, they are more often married, and their households tend to rely more often on other income sources than agricultural wage labour. The first two differences are especially interesting. In qualitative interviews during our first visit respondents told us that wageworkers in non-FT plantations are always looking out for job opportunities at FT plantations; they prefer to work at FT plantations because of more and/or better benefits. Plantation managers told us that workers, including Haitians workers, tend to stay longer on FT plantations than in non-FT plantations—probably for the same reasons. This may be related to the fact that FT policies make it mandatory for managers to help wageworkers from Haiti to obtain all the required papers to formalise their working status in the Dominican Republic.
Aside from these differences there may be other unobserved differences between certified and non-certified plantations, or the type of wageworkers they attract, that could bias our analysis if these same factors also influence wageworkers’ well-being or voice. A first potential bias would be that the FT plantations are more efficient, which could be an additional factor influencing worker well-being, aside from certification (e.g. through improved profits). We consider this bias to be unlikely. All plantations included in our sample were already exporting to European markets—including those that are not FT certified. Also, a recent study by the ILO (2017b) shows that productivity is not necessarily related to being certified or not. Some of the more efficient plantations, in terms of production and profitability, offer economic and social benefits that are in fact equal or higher than the ones available at some FT plantations. A second potential bias is that wageworkers who work on FT plantations are on average more motivated—given the prospect of better working conditions promoted under FT certification—or related to this, more ambitious or hard working. We control for these differences as much as possible by considering the variables listed in Table 1, including years of employment on the plantation.
Indicators
As argued before the study focuses on three key areas of benefits: economic, social and voice-related. The sub-themes covered under the economic benefits include wages, sense of job security, non-wage (in-kind) benefits and proxies for standard of living. These benefits relate to access to sanitation, healthcare and food at the workplace; transport to the plantation; housing and/or access to water and electricity; but also schooling for children, recreation and sports, and access to education for adults. Another key theme for FT that could be subsumed under economic benefits is living wage.1 The living standard sub-theme is strongly related to the living wage discussion, and includes savings (provision for unexpected events), poverty levels and food access (two measures of a decent standard of living).
The sub-themes covered under social benefits include: (awareness of) working conditions on the plantations (working hours, holidays, social security, and occupational health and safety), quality of social dialogue (grievance redressal, relationship to supervisors and trust in relationship) and the FT premium. The FT premium is an additional sum of money which goes to communal funds for workers, to be used to their economic, social or environmental benefit. Examples of such projects reported during field work are helping a community member rebuild a house after it burned down, providing travelling dentist clinics and providing scholarships to children. Many of the indicators related to working conditions are based on ILO concepts and definitions. In the survey we also asked wageworkers about their access to social security as an indicator of social benefits. However, access to social security turned out to be a sensitive issue in the Dominican Republic. Instead of asking wageworkers whether they received a certain security, we asked for details on the type of security received. However, these data turned out to be unreliable, probably because interviewees were often not aware of the details of the type of security they have access to.
The sub-themes captured under voice aimed to create a balance between the general literature and the definition of some empowerment practices (also see Lyall 2014). The sub-themes listed under voice are sense of ownership, membership in workers’ organisations, sense of control and life satisfaction, and changes in individual career prospects through participation in training. FT considers sense of ownership among workers as an important factor in ensuring future sustainability; it is expected to increase as an outcome of the joint decision-making process that Fairtrade aims to implement (for example, in the use of the FT premium). Ownership is based on a conceptual model used by Ruben and Van Schendel (2008) in a study of the banana sector in Ghana and in a study by Van Dyne and Pierce (2004) on psychological ownership and employee attitudes. Membership in worker committees is an important element of empowerment because the use and management of the Fairtrade premium is part of the work of these committees. Finally, life satisfaction captures workers’ satisfaction in relation to income, housing, schooling, vocational training, heath, food, loans and public services.
Empirical model
We empirically test for the contribution of FT certification to the selected indicators based on the worker survey. We estimate the outcome indicators (I) as defined in Table 2 as a function of working on an FT-certified plantation (FT), wageworker characteristics (X), household control variables (Y) and plantation control variables (Z)::
$$I_{n} \; = \;fn\;\left( {FT_{j} ,\;X_{ij} ,\;Y_{ij} ,\;Z_{j} } \right),\;n = \{ 1,\;2,\;3 \ldots \}$$
(1)
The nth indicator is measured for worker i at plantation j in 2015 (see Table 1). To control for potential bias of worker, household or plantation level characteristics (aside from those controlled for in the sampling designs), we incorporate the set of worker covariates (X), household covariates (Y) and plantation covariates (Y) as listed in Table 1. Household and worker covariates relate to wageworker demographics, education, assets, work experience and residency history (see Table 1). We estimate the model using ordinary least squares (OLS). Since treatment is at the plantation level, we cluster the error term (εij) at the plantation level j.
Robustness analysis
To build a strong counterfactual—i.e. “What would have happened to the hired labour conditions if the plantation had not received FT certification?”—the results from model 1 were combined with the results of qualitative data from the in-depth interviews to triangulate, validate and explain differences or lack thereof.
Moreover, to make our empirical results as robust as possible we test for the effect of FT certification on working conditions and wageworker well-being using three alternative model specifications. Only a randomised experiment can fully eliminate biases in the covariables, both observable and unobservable ones. However, several practical factors and ethical considerations may prevent this type of experiment from taking place. Hence, methods have been sought to approximate the robustness of randomised experiments. Pairing methods are among the most used and complex techniques (Stuart 2015). As a robustness test of the traditional OLS regression with control variables and clustering (model 1) we use propensity score matching (PSM) (model 2), and the entropy balancing method, with and without clustered standard errors at the plantation level (models 3 and 4). A result is considered fully robust if the four models show the same results in terms of sign and significance.Footnote 7
The use of the propensity score—defined as the probability of receiving a treatment given observable covariates—has been popularised among researchers from different fields. This is the first alternative model we use to test for robustness of our OLS estimation (model 1).Footnote 8
However, PSM can lead to bias when there is no clear treatment allocation rule or a very large set of covariates. While this method balances the mean values of the observable covariables, it does not balance variance of observable covariates or differences across groups. In addition, when there are no symmetric distributions in the covariates—such as binary covariables, continuous categories and/or biased variables in a distribution—bias can be reduced in some variables, but it can be increased in others. Hence, PSM can lead to biased impact estimates (Diamond and Sekhon 2013).
Hainmueller and Xu (2013) introduced the entropy balancing method as an alternative pairing method. This method relies on a maximum entropy reweighting scheme that calibrates unit weights so that the reweighted treatment and control group satisfy a potentially large set of pre-specified balance conditions that incorporate information about known sample moments (mean, covariance and bias). Entropy balancing thereby carefully adjusts inequalities between workers from certified and non-certified plantations.