Economic performance of organic dairy farms during and after conversion

Abstract

Empirical research on the economic performance of organic farming yields conflicting results. The findings vary by region, agricultural sector, economic performance indicators, and methodologies used. This study compares the economic and financial performance of conventional and organic dairy farms by specifically examining the performance trends during the period when conventional farms got converted into organic farms. Using a difference-in-differences estimator within a panel regression model, we analysed the microdata of 1,016 farms in Ille-et-Vilaine (Brittany, France) between 2007 and 2018, including 62 farms that had converted from conventional farming to organic farming during this period. Our results show that, contrary to expectations, the financial performance of the farms did not reduce significantly during the conversion. However, after the two-year conversion period, the profitability and return on assets of organic farms exceeded those of conventional farms. Considering the limited sample size of organic farms analysed in this study, these findings require further validation.

1. Introduction

Social science literature on the conversion to organic farming has largely focused on the incentives and obstacles influencing the adoption. In this study, as an addition to recent reviews on attitudinal (Thompson et al. 2023) and supportive factors (Sapbamrer and Thammachai 2021) affecting the conversion, economic and financial incentives have been studied as key determinants affecting the decision to convert to organic farming. Some farmers view conversion to organic farming as an opportunity to secure farm sustainability, while others expect it to yield a higher profitability. Although many studies have investigated the drivers of conversion, robust analysis of the impact of the conversion on the farms’ economic performance remains methodologically challenging. To understand the consequences of conversion decisions better, some studies have compared the economic performance of organic and conventional farms.

In the European Union (EU), organic farming has attained considerable development, with organic farmland doubling between 2010 and 2020, and reaching almost 10 per cent of the total agricultural area in 2023.¹ These expansions have facilitated the collection of enough samples for observation, allowing for increasingly sophisticated quantitative analyses (Uematsu and Mishra 2012; Gillespie and Nehring 2013; Lakner and Breustedt 2017; Froehlich, Melo, and Sampaio 2018; Lambotte et al. 2023; Martín-García et al. 2024). Regardless, findings regarding the economic performance of organic farming versus conventional farming remain inconsistent. The findings are strongly dependent on the economic performance indicators and methodologies used to avoid comparison bias between farm types.

While existing literature has studied various agricultural sectors, the current study has focused primarily on dairy farming. Dairy farms that convert to organic farming must fulfil several requirements, which apply to both livestock and crop production farm units. When entire farms are converted at one time, including the lands and the herds, the conversion should be completed within two years. During this period, products must be sold through the conventional market channels. Some studies have found that farms that have completely converted to organic farming achieve higher incomes, primarily because of organic price premium and reduced variable input costs (McBride and Green 2009; Delbridge et al. 2013; Patil et al. 2014; Dedieu et al. 2017). Conversely, other studies have emphasized that high conversion costs and high labour costs severely undermine farm competitiveness (Uematsu and Mishra 2012). Lambotte et al. (2023) found that the gross margin per annual work unit of organic French dairy farms is comparable to that of conventional farms, while the gross margin per litre of milk is higher in organic farms. However, they note that this observation conceals structural differences: organic farms prioritize achieving high margins per litre of milk produced with lower volumes, unlike the conventional model, which relies on high production volumes with lower unit margins.

Research on the performance of organic and conventional farms has also highlighted the financial challenges that organic farmers may face during the conversion period. However, fewer studies have addressed this issue due to the difficulty of obtaining farm data before, during, and after conversion. In dairy farming, financial constraints are generally experienced during the investment phase of the conversion. To offset the decline in milk yield per cow during the conversion period, farmers must either purchase or rear additional cows. They must also acquire new equipment to manage larger herds (Müsshoff and Hirschauer 2008). Conversion also necessitates reorganization of the farmland to increase forage and/or feed. Kerselaers et al. (2007) report that conversion can halve the economic potential of dairy farms due to heightened risks and reduced cash flow (CF). Läpple (2010) found that economic pressures may even push farmers to abandon organic farming. Other studies suggest that the learning curve associated with managing a fully converted organic farm extends beyond the legally mandated two-year conversion period (Sipiläinen and Oude Lansink 2005; Lakner and Breustedt 2017).

This study aims to enhance the understanding of the economic consequences of converting conventional dairy farms into organic ones. Specifically, it examines whether the economic performance of farms declines during conversion and whether it recovers subsequently, after conversion. These consequences have implications on farmers’ motivation to adopt organic farming. We hypothesize that conversion to organic farming can affect farm profitability, solvency, and liquidity. To test this hypothesis, we compared the economic and financial performance of organic and conventional dairy farms and analysed the performance trends during and after conversion. We conducted our empirical analysis on a sample of dairy farms in Ille-et-Vilaine, a subregion of Brittany, which is the leading dairy-producing region in France, accounting for 20 per cent of the nation's milk production and 33 per cent of the nation's organic milk production. Although organic milk comprises only 2.4 per cent of the total production, its volume increased by 50 per cent between 2017 and 2019. Few economic studies have examined the financial performance of organic dairy farms in France. Using propensity score weighting, Lambotte et al. (2023) compared the environmental performance, measured by greenhouse gas emissions, and the economic performance, assessed by gross margin, of organic and conventional French dairy farms. Most other studies rely on grey literature or agronomic research, which often have methodological limitations (Géniaux et al. 2010; Pavie et al. 2012). These studies typically compare organic and conventional farms based on averages, using samples that are not comparable in terms of location or structure. They often focus on a single year without tracking farms over time or distinguishing between farms going through conversion and those already certified (Devauvre 2024).

In the current study, we used widely recognized economic and financial indicators to analyse the economic and financial performance of farms. These indicators are based on three dimensions: profitability, solvency, and liquidity. Direct comparison of the economic performance of organic and conventional farms is not advisable as it may produce biased estimates. Several studies have identified farm characteristics associated with a higher likelihood of converting to organic farming (Kallas, Serra, and Gil 2010; Latruffe and Nauges 2014). Consequently, the outcomes of both farm types (conventional and organic) may result not only from differences in the production process but also from systematic differences in observed and unobserved characteristics between farms (Froehlich, Melo, and Sampaio 2018), such as herd size and type of feeding system. To estimate the effect of organic farming on economic performance, we applied a difference-in-differences (DID) estimator within a panel regression model. This method controls for individual- and time-specific fixed effects to identify the treatment effect of converting to organic farming. This approach allows us to control for selection bias. Alternative methods also exist to address such biases. Martín-García, Gómez-Limón, and Arriaza (2024) examined the economic performance of Spanish fruit farms using propensity score matching to minimize biases caused by structural differences between conventional and organic farms. We believe that the DID approach has never been used to measure the impact of organic farming on economic performance. This method, which is increasingly used in research, is particularly relevant for organic agriculture, as it provides a dynamic framework to study the improvement in economic performance as farmers go through the two-year conversion process.

The remainder of this paper is structured as follows. Section 2 presents the empirical model and the DID estimator. The different types of profitability, solvency, and liquidity indicators are also described in this section, along with the expected effects of organic conversion on these indicators. Section 3 presents the data and discusses the results. Finally, Section 4 concludes the study with a brief summary and a discussion of the relevance and limitations of the main findings.

2. Methodology

2.1 Empirical framework

This section presents the model used in this study to assess the impact of conversion to organic farming on the economic and financial performance of dairy farms. The analysis is based on a random sample of N dairy farms observed over T time periods. For each farmer (⁠$i\in \{0,\dots ,N\}$⁠) and time period $(t\in \{0,\dots ,T\})$⁠, we observe the outcome $(Y_{i,t})$⁠, corresponding to different measures of economic performance and treatment status $(D_{i,t})$⁠.

The treatment variable is binary, indicating whether a farm has converted to organic farming or not. The treatment status $(D_{i,t}$⁠) equals 1 if farmer i is treated during period t and 0 if otherwise. The adoption plan is staggered as farmers convert to organic farming at different times. In our sample, farmers who had converted to organic farming did not switch back to conventional farming, which simplified our framework. Some farmers in our sample chose not to convert to organic farming. We assumed that dynamic treatment effects exist, because from the moment $G_i$ when farmer i starts converting, he enters a conversion period lasting for two years before being officially certified as organic. The impact of the treatment, if it exists, may likely evolve during this conversion period before stabilizing a few years after the start of conversion. Farmers may begin adjusting their practices in anticipation of conversion, which may also impact the results.

We used two different approaches to estimate the effects of conversion to organic farming on the economic and financial performance indicators of dairy farms.

2.1.1 Dynamic two-way fixed-effects model

Under certain conditions, the DID method is derived directly from the framework of individual and time-fixed effect models used in panel data econometrics. These models are called two-way fixed-effects (TWFE) models. This model produces unbiased results under several conditions (De Chaisemartin and D'Haultfoeuille 2023). First, the parallel trend assumption must be verified. The outcome trajectories of the dependent variable for the treated and untreated groups must follow a similar pattern before the introduction of the treatment. Thus, in the absence of treatment, the difference between the groups should remain stable over time. Second, the treatment effect must be homogeneous (i.e. it must remain the same for all individuals and over time).

Unlike static TWFE models, dynamic TWFE models account for treatment effects across multiple periods before and after intervention. They capture delayed responses and gradual changes in treatment effects. These models use relative treatment indicators to track different points in time before and after treatment adoption. Let $G_i$ be the year in which the treatment began for individual i. The relative treatment indicator for year t is $D_{i,t}^k$⁠, which is equal to 1 if individual i began treatment k years ago with $k=t-G_i$⁠. These indicators estimate the dynamic and temporal effects of the treatment, allowing us to observe how the impact evolves over time before and after the intervention. In a staggered model with a binary treatment, the dynamic TWFE model is expressed using Equation (1):

with $Y_{i,t}$ representing the economic performance of the farm, ${\eta }_i$ being the fixed effect for individual i, ${\delta }_t$ being the fixed effect for year t, and $D_{i,t}^k$ being the relative treatment indicator for each period relating to event $k\equiv \{-K,\dots ,K\}$⁠. We estimated the dynamic average treatment on the treated (ATT) for a specific time horizon k (e.g. k periods after treatment) directly by the coefficient ${\beta }_k$⁠. For $k\ge 0$⁠, ${\beta }_k$estimates the cumulative effect of the $k+1$ treatment periods. For $k\le -2$⁠, ${\beta }_k$ is a placebo coefficient testing the hypothesis of parallel trends. We had to exclude some periods to avoid multicollinearity problems (Borusyak and Jaravel 2017) as ${\sum }_{k\in [-K,K]}^{}{D}_{it}^k=1$and because there is a linear relationship between fixed effects and relative period indicators. Excluding the relative periods close to the initial treatment is a common practice. Normalization to the pretreatment period is also common; hence, $D_{it}^{-1}$ was deleted. To account for treatment effects stabilizing after conversion, we followed the recommendation of Schmidheiny and Siegloch (2023), who propose replacing the first $1\{t-G_i=K\}$ and last $1\{t-G_i=-K\}$ period indicators with one, indicating that there are at least K periods left before adoption$1\{t-G_i\le -K\}$⁠, and that adoption occurred at least K periods ago $1\{t-G_i\ge K\}$⁠.

The dynamic TWFE models used in literature have limitations identical to those of the classic static TWFE models. If the parallel trends assumption is violated or treatment effects vary across individuals or over time, the estimated coefficients would be biased. Although the relative time indicators in the dynamic TWFE models can capture some temporal aspects of the treatment effects, they fail to account for the complex and nonlinear manner between units and over time. These models also do not account for interactions between treatment effects and external factors such as climate, economic conditions, or institutional differences.

2.1.2 Robust to heterogeneity model

Recent studies suggest that, when the conditions required to make a robust estimation of the treatment effect are not met (e.g. parallel trend and heterogeneous effects between individuals and over time), the TWFE estimator would be inadequate. To get an overview of the recent econometric literature, see De Chaisemartin and D'Haultfoeuille (2023) and Roth et al. (2023). De Chaisemartin and D'Haultfoeuille (2020) showed that when the treatment effect is not constant between groups and/or between years, the parameter estimated from the TWFE would be biased.

A primary issue in this study is that all the farms did not convert to organic farming at the same time. If the conversion had happened simultaneously, based on the parallel trend hypothesis, we would have considered that organic farms followed the same trajectory as conventional farms. This would have allowed us to estimate the output of organic farms had they not converted. However, as there were multiple treatment dates, the TWFE estimator not only compared organic and conventional farms but also evaluated farms undergoing conversion relative to those already converted. If treatment effects are not constant over time, the trajectories of these comparisons may be different from those of the farms that were not converted (heterogeneity over time). Similarly, problems may arise when comparing a farm that had a greater treatment effect with one that had a lower treatment effect (group heterogeneity). This can be problematic if, in the end, the effect of the treatment is more significant after several years of conversion, as may be the case with organic conversion. Consequently, TWFE may yield a negative average treatment effect despite all farms experiencing positive treatment effects.

Several DID estimators that support heterogeneous treatment effects have been proposed in literature (De Chaisemartin and D'Haultfoeuille 2023). This study has adopted the estimator proposed by Callaway and Sant'Anna (2021), which is well suited to our dynamic staggered adoption design with binary treatments. Their framework groups individuals into cohorts of those who began their farm treatment simultaneously. Cohort g groups individuals who began receiving treatment in year g. For each cohort g and relative period k, the conditional treatment effects $\widehat{CAT{T}_{g,k}}$ are estimated in Equation (2):

with $N_{g,k}$ being the number of individuals in cohort g at relative period k and ${\hat{\omega }}_{jt}$ being the weights for the control individuals. The term ${\sum }_{j:G_j\ne g}^{}{\hat{\omega }}_{jt}{Y}_{jt}$ denotes the weighted sum of outcomes observed for individuals j outside cohort g, serving as a control group. The term ${\sum }_{i:G_i=g}^{}(Y_{it}-{\sum }_{j:G_j\ne g}^{}{\hat{\omega }}_{jt}{Y}_{jt})$ thus corresponds to the sum of the differences between the outcomes of individuals in cohort g and those in the control group. The ATT is obtained by aggregating conditional average treatment on the treated (CATT) estimates to obtain a synthetic measure of the average treatment effect across all treated individuals in Equation (3). To obtain a dynamic ATT for all cohorts at a specific time horizon k years after treatment, we calculated a weighted average of the CATT values for all cohorts that have reached this time horizon. Hence, we define the dynamic ATT at k periods after the onset of treatment in Equation (3) as follows:

where ${\pi }_{g,k}$ denotes the weight applied to each cohort g in the aggregation. This weight is based on the proportion of each cohort in the sample or the number of individuals in each cohort. $\widehat{CAT{T}_{g,k}}$ represents the conditional average treatment effect for cohort g observed k periods after treatment, as defined in Equation (2). The standard errors associated with the ATT estimates are designed to be robust. Hence, intracluster correlations and statistical uncertainty in the CATT estimates have been accounted for.

2.2 Economic and financial performance indicators

We have relied on several indicators commonly used in literature to assess the economic and financial health of the farms. Economic indicators measure the profitability of production, while financial indicators evaluate farms’ investment decisions and funding strategies. The calculations of these indicators are presented in  Appendix A.

Farm profitability was assessed using three types of economic indicators—milk gross margin (milk GM), earnings before interest, taxes, depreciation, and amortization (EBITDA), and gross profit margin (GPM). These may be influenced by individual farm characteristics, such as the farmer's age, skills, education or training, or by the type of production system, particularly the livestock feeding system. Additionally, they can be influenced by external factors such as price and weather conditions (yields), herd size, and the existence of economies of scale (Chavas 2001; Wolf et al. 2016).

2.2.1 Milk gross margin

The first profitability indicator is the margin per litre of milk produced (milk GM). Research indicates that organic farms typically achieve a higher gross margin per unit of production than conventional farms, mainly because they sell at higher prices and incur lower operating costs (Pavie et al. 2012; Delbridge et al. 2013; Patil et al. 2014; Dedieu et al. 2017). In France, the purchases and external expenses per dairy cow were estimated to be 20 per cent lower in organic dairy farms in 2013 (Dedieu et al. 2017).

2.2.2 Earnings before interest, taxes, depreciations, and amortization

The second profitability indicator is EBITDA, which measures the level of wealth created by the farm per unit area. Compared to the total gross margin, this measure excludes structural expenses, such as wages, and includes operating subsidies. When all operational and structural costs are considered, the results would be mixed. For instance, in 2008, Uematsu and Mishra (2012) analysed 2,689 field crop farms in the United States, of which only 65 were organic. They found that organic farmers do not have significantly higher incomes than conventional farmers owing to their higher labour, insurance, and marketing costs.

2.2.3 Gross profit margin

The third profitability indicator is the ratio of farm net profit to sales, or the GPM or margin rate. This indicator shows the percentage of sales retained by a company after covering all expenses. It reflects a farm's efficiency in controlling its operating costs in relation to its output value.

Following Wolf et al. (2016), we assume that dairy farms undergoing conversion require greater investments than conventional farms, which potentially affects their asset profitability and overall financial performance. Therefore, in addition to profitability indicators, we considered three financial indicators that evaluate the ability of farms to meet their short-term needs (liquidity) and their short-, medium- and long-term financial obligations (solvency)—return on assets (ROA), CF, and debt-to-asset (DA) ratio. These financial indicators are complex to interpret, as their evolution trends are influenced by investment levels and investment-related financing decisions made by the farm owners. They are also influenced by the current and past financial management of the farm, which can distort the reading of their evolution. Additionally, they may be affected by decisions made by farmers or farm partners regarding consumption, savings, and investment as well as finance obtained from banks.

2.2.4 Return on assets

ROA, also called the ratio of financial profitability, is used to assess the profitability of assets. ROA is calculated by dividing operating profit by total assets. This ratio evaluates a farm's ability to generate profits using both fixed and current assets. An ROA below 5 generally indicates low profitability relative to material and financial resources (Wolf et al. 2016). ROA can be influenced by several factors, such as farmers’ personal attributes, farms’ level of diversification, and farm size.

2.2.5 Cash flow

To complete our assessment of the financial performance, we use the CF as a liquidity indicator. CF is calculated as the balance of asset and liability CFs. A positive CF reflects financial flexibility, enabling firms to manage unforeseen events without having to rely on external funding. Conversely, a negative CF indicates short-term borrowing and a shortage of cash. However, accurate interpretation of the CF remains challenging. The CF level is essentially very volatile as it depends largely on market or weather conditions in a given year and on investment decisions, which vary greatly from one farm to another. Kersaelers et al. (2007) found that dairy farm conversion potential is significantly affected by higher risks and lower CF during the conversion period.

2.2.6 Debt-to-asset ratio

The DA ratio is used to measure the solvency of firms. It is the ratio of total debt to total assets of the firm. The impact of DA on farm economic performance is variable. While debt accumulation can be caused by various factors such as farm expansion, farm investments, and increased farm expenses, debt can have a positive as well as negative effect on economic performance (Mugera and Nyambane 2015). In the short term, higher debt levels increase structural costs and reduce profitability. However, in the long run, higher debt can favour innovative investments such as those incurred during conversion to organic farming.

3. Data and estimation results

3.1 Data

We obtained individual accounting data for farms in Ille-et-Vilaine, Brittany, from a regional private management and accounting agency. This database holds private accounting data of organic and conventional dairy farms, for the period ranging from 2007 to 2018. It contains data related to 1,471 dairy farms, 62 of which were converting to organic farming during the said period and 40 of which had converted before 2007 (i.e. before the studied period) (see Table 1). This dataset represents over 50 per cent of the farms in this region, studied for a period of 12 years, from 2007 to 2018. Each farm was observed for periods ranging from 2 to 12 years, resulting in a total of 8,555 observations.

As per our chosen methodology, we excluded organic farms that had converted before the study period from the sample. Hence, the treatment group comprised the 62 farms that converted to organic farming from 2008, totalling 609 observations. The control group comprised conventional farms that had not been converted (1,372 conventional farms, out of a total of 7,601 observations).

The small size of our treatment group presented two challenges. First, when the treatment group is small, there is an increased risk that the estimated treatment effects will not be significant, majorly due to higher variance. This lack of significance does not necessarily imply no effect; it only means that the estimated effect is not precise enough to be distinguished from zero with a high level of confidence. If treatment effects are significant despite the small sample, they are likely to be strong enough to be detected. However, the results must be interpreted with caution. The results of this test do not guarantee that the effect would be stable in other samples or contexts. Confidence intervals and robustness checks must be done to validate the reliability of the results.

Second, we could not add control variables in our model, as this could reduce the variability of the main variables, making the relationships between the independent variables and the dependent variables harder to identify and less significant. While the DID model corrects some biases by accounting for initial differences between the treated and untreated groups, adding more control variables can strengthen the analysis, particularly when groups differ in unobserved ways not captured by simple pre- and post-treatment trends. To address this issue, the parallel trend assumption was validated (see the next section).

Table 2 presents the descriptive statistics pertaining to the study. The organic dairy farms in our sample exhibit structural, technical, and economic characteristics consistent with prior research. First, organic farmers add value to their produce by affixing organic certification labels and by selling milk at higher prices. In our sample, organic milk prices were approximately 25 per cent higher than conventional milk prices, equating to an average milk premium of €86 per 1,000 litres of milk. This premium varies depending on market conditions and dairy collectors. Second, organic farming promotes minimal and efficient use of external inputs, particularly feed concentrates, as chemical fertilizers and pesticides are banned. On average, operating expenses, as a percentage of total revenue, were 15 per cent lower in organic farms, primarily because of lower costs of concentrates (40 per cent lower in organic systems than in conventional systems). Third, owing to reduced use of concentrates, organic cattle diets were based on grazed grass or silage, requiring additional grassland. In our sample, organic farms had an average utilized agricultural area (UAA) of 80 ha (against about 67 ha for conventional systems). Moreover, 88 per cent of the main forage area was allocated to grassland (against about 60 per cent for conventional systems). These last two points allow organic farmers to achieve a higher level of food autonomy, which is essential for conversion to organic production and to mitigate exposure to input price volatility (Bouttes, Darnhofer, and Martin 2019). The organic farms that were analysed in this study relied on animal–plant complementarity to manage soil fertility instead of using synthetic fertilizers. Moreover, they used grass in meadows to feed the herd, thus limiting the use of purchased feed. Instead of using agrochemicals to manage disease and parasite issues, the organic farms diversified their crop rotations by introducing longer rotations. The economic and financial performance indicators of the chosen farms seemed to have improved on average before and after conversion, but with substantial individual variability. Notably, organic farms received EU financial support between 2007 and 2018 for both conversion to organic farming and maintaining their organic status (Table 2).

3.2 Parallel trends test

We tested the hypothesis of parallel trends to confirm whether the treatment and control groups would have followed parallel trajectories before treatment. This was done by estimating placebo effects over the pretreatment periods and testing their statistical significance. Using Callaway and Sant'Anna's (2021) approach, we calculated treatment effects on pretreatment data to estimate dummy effects and analyse their significance. The results of pretreatment periods can be presented in two different ways. The first method involves employing a universal base period and is used in event study regressions to test parallel trend hypothesis. The second method involves the use of a variable base period, wherein the immediately preceding period is used as the base period. This approach highlights how differences evolve relative to the most recent periods, allowing easier detection of early shifts that could indicate anticipation effects. Figure 1 presents these two cases.

Figure 1 provides the graphical representation of the pretreatment effects based on Callaway and Sant'Anna's approach. The x-axis represents different periods: periods $t=-1$ to $-4$ corresponds to the four years before conversion, $t=0$ corresponds to the year of conversion, and periods $t=1$ to $4$ correspond to the four years following conversion. The dots from t = 0 onwards represent the average treatment effect estimated for différente periods. This data is presented in tabular format in Table 3. Ths dots before t = 0 represent the pretreatment estimates. The figures on the left employ a varying base period, while those on the right use a universal base period. Across all indicators, pretreatment estimates were not significantly different from zero, confirming the parallel trend hypothesis and indicating that there are no anticipation effects.

3.3 Estimation results

We estimated the treatment effects using both the dynamic TWFE model and the robust to heterogeneity (RH) model proposed by Callaway and Sant'Anna (2021). The second approach relaxes the assumption of constant treatment effects across groups and time periods. Table 3 presents the results of the estimates. In the dynamic TWFE model, the average treatment effect $AT{T}_k$corresponds to the ${\beta }_k$parameter. Conversely, in the RH model, it is represented by the weighted sum of the estimated treatment effects $\widehat{CAT{T}_{g,k}}$⁠. The terms $AT{T}_1,AT{T}_2$⁠, and $AT{T}_3$ indicate the treatment effects one, two, and three years after initiation of treatment, respectively. The term $AT{T}_4$ represents the average treatment effect beyond three years of treatment. This enables us to analyse the differences in economic and financial performance between organic and conventional systems after obtaining organic certification. We have extracted three key findings from this estimation.

First, the results are highly consistent across the two models. Significant parameters remain significant in both cases, although slight differences were observed in the orders of magnitude. For milk GM, the estimated average treatment effect four years after conversion to organic farming was an increase of €166.07 per 1,000 litres of milk as per the dynamic TWFE model and €161.41 per 1,000 litres of milk as per the RH model (Table 3). This implies that, on average, organic farms have higher gross margins of €161–€166 per litre of milk after the conversion period, compared to conventional farms. Greater difference was observed between the two models in EBITDA per hectare: four years after conversion, an increase in EBITDA of €0.38 per hectare was recorded in the dynamic TWFE model, as compared to €0.19 per hectare in the RH model (Table 3). These findings suggest that the potential heterogeneity between groups and across years in our sample is insufficient to generate opposite effects with the TWFE model, as highlighted by De Chaisemartin and D'Haultfoeuille (2020). However, failure to account for heterogeneity can lead to significantly biased parameter estimates.

Second, in both models, some financial indicators of organic farms did not significantly differ from those of conventional farms, during or after the conversion period. Neither the DA ratio nor the CF was affected by the conversion to organic farming, during or after conversion. The ROA ratio did not see any change during the two-year conversion period; however, the GPM was significantly positive in years 3 and 4, after conversion, in both the TWFE and RH models (+9.80 and +4.76 in year 4 in TWFE and RH models, respectively) (Table 3). This contradicts our initial expectations for the conversion period. A decline was expected in the financial ratios during the conversion period due to the need to purchase additional animals (to reach the same milk volume with lower yields per cow) and new equipment. Moreover, farm areas needed to be reorganized to produce more fodder and/or feed for the herd, resulting in higher structural expenses and assets (Müsshoff and Hirschauer 2008). Thus, the level of structural expenses and assets could have increased in the denominator of the ROA and DA ratios. We also expected the conversion to have significant effects on CF. Kersaelers et al. (2007) found that the economic potential of conversion could decline by 50 per cent in dairy farms because of higher risks and lower CF during the conversion period. Our more nuanced results in terms of financial ratios can be interpreted in several ways. DA and CF are highly heterogeneous across farms and can be volatile year to year, depending on consumption and investment decisions. The limited sample of organic farms considered for this study could prevent the identification of an effect on these indicators. Another possible assumption is that conversion to organic farming does not affect financial performance indicators during conversion. More likely, the lack of significant findings was attributable to the lack of observations of organic farms or the high heterogeneity of these farms, particularly in terms of their financial performance indicators (Khanal, Mishra, and Honey 2018).

Third, the results confirm that farms experienced improved economic performance after converting to organic farming. The EBITDA and milk GM indicators increase gradually during the conversion period and are significantly higher after conversion (represented by $AT{T}_4$ in Table 3). On average, the gross margin per litre of milk in organic farms is approximately 50 per cent higher than in conventional farms. This difference can be attributed to lower external input costs, such as mineral fertilizers, pesticides, feed concentrates, and veterinary expenses, in addition to significantly higher milk prices. The EBITDA per hectare was 25 per cent higher, which aligns with our expectations as the EBITDA indicator includes the EU financial support provided for the conversion to and maintenance of organic status. However, profitability, as measured by GPM, does not differ significantly from that of conventional farms during or after conversion. This suggests that increased milk prices and milk GMs in converted farms do not necessarily translate into higher income across all farms.

4. Discussion and conclusions

This study examined the impact of conversion to organic farming on the economic and financial performance of dairy farms, both during and after conversion. Our analysis relied on longitudinal data from a sample of dairy farms observed between 2007 and 2018, a small proportion of which converted to organic during this period. We used a DID estimator with dynamic methods, allowing us to account for variations in conversion dates and potential heterogeneous conversion effects over time. We found no significant changes in farms’ financial performance during the two-year conversion period. However, after conversion, profitability indicators—including gross margin per litre of milk and farm income per hectare—were higher than those of conventional systems. These results highlight the importance of distinguishing between performance during and after the conversion period, as failing to do so may lead to biased conclusions.

Nonetheless, our findings do not provide conclusive evidence of economic and financial fragility during the conversion phase, which could discourage farmers from transitioning to organic. The small sample size raises concerns about the comparability of groups, as the sampling strategy may introduce bias and limit the interpretability of our results (Offerman and Nieberg 2000). This is all the more problematic as we are assuming a high degree of heterogeneity. The literature highlights considerable heterogeneity in farm strategies and motivations for converting to organic farming. Debates on the conventionalization of organic farming suggest that the profile of farms adopting organic practices has evolved over time (Seidel, Heckelei, and Lakner 2019). Some studies propose a bifurcation between two types of organic farms: larger farms oriented towards production and export, which have converted more recently for economic reasons, and smaller farms focused on local markets, motivated by food quality and environmental concerns (Hall and Mogyorody 2001; Flaten et al. 2006; Best 2008; Läpple and Van Rensburg 2011). Given the limited sample of organic farms in this study, explicitly accounting for this heterogeneity remains challenging. It is therefore impossible to determine whether the absence of economic fragility during the conversion period is genuine or a consequence of the limitations of our sample.

These findings have nevertheless implications for both public and private sector policies supporting dairy farmers in their transition to organic farming. The improvements observed in post-conversion economic performance align with most empirical studies, which show that organic farms tend to achieve higher average profitability than conventional ones. This suggests that conversion and maintenance subsidies have contributed to increasing post-conversion income and incentivizing farmers to transition. However, the substantial reduction in maintenance subsidies for organic farms under the French CAP 2023–2027 strategic plan, which limits support to €30 per hectare as an eco-scheme bonus, could jeopardize these gains and hinder the Green Deal's objective of reaching 25 per cent organic agricultural land by 2030.

Future research should collect additional data on organic dairy farms and track them over a longer period to validate our findings. A larger sample would enable better control of farm heterogeneity, particularly in terms of investment strategies and economies of scale. This would also enable a more robust assessment of the economic risks associated with the conversion period and allow us to evaluate the benefits of financial support, given the substantial investments and operational changes required during this transition.

Conflict of interest

None declared.

Funding

This research is funded by the Horizon 2020 programme of the European Union (EU) under grant agreement number 770747 (LIFT project, https://www.lift-h2020.eu).

Data availability

Data are owned by a third party. The data underlying this article were provided by Broceliande Cerfrance by permission. Data will be shared on request to the corresponding author with permission of this third party.

Footnotes

Appendix A. Calculation of economic and financial indicators

References