IntroductionTop

Diversification via processing and direct marketing of agricultural products is one of the strategies available to farmers in order to maintain a competitive position in the market. This type of diversification, promoted by the European Union (EU) and supported by the new Common Agricultural Policy (CAP) 2014-2020, tends to focus on differentiated products, sometimes certified with protected designation of origin (PDO), and aimed at niche markets.

Specifically, regarding the EU dairy sector, since 2003 a series of reforms to the CAP have led to market forces now being the main determinant of milk prices. As a result of this process, milk prices have fallen due to cuts in intervention prices, becoming more in line with world prices. Due to the fall in the price of raw milk, certain dairy farms have adopted a strategy of entering the business of processing the raw milk in order to achieve higher margins.

Although many studies have focused on the profi-tability and efficiency of milk production in dairy farms (e.g. Tauer, 2001; Cabrera et al., 2010; Casasnovas-Oliva & Aldanondo-Ochoa, 2014), a gap exists with respect to the study of the profitability of processing activities (Becker et al., 2007; Bouma et al., 2014). This is a relatively under-investigated field of research due to the difficulty of obtaining data and the problem of separating the value of transformation activities from the milk production results.

The objective of this paper was to investigate the factors that explain the profitability of value-added products in dairy farms. For this purpose we used a unique database from a group of Spanish farms involved in diversification via the elaboration and sale of dairy products.

We used an approach inspired by hedonic price models, where a product is considered as a bundle of attributes and its observed price is a linear combination of these, with the weights of the attributes representing their implicit prices (Rosen, 1974). There is a broad literature about hedonic models applied to different food products (Loureiro & McCluskey, 2000; Troncoso & Aguirre, 2006; Costanigro & McCluskey, 2011), including dairy products (Gillmeister et al., 1996; Smith et al., 2009; Carlucci et al., 2013; Loke et al., 2015; Bimbo et al., 2016), but none of these studies has focused on value-added products processed and sold by dairy farms. To the best of our knowledge this paper is the first to identify the separate effect of each attribute on the margin (income minus variable cost), which allows evaluating the profitability achievable by adopting different strategies (Carlucci et al., 2013).

The product attributes were selected based on both, the most-used attributes reported in the agri-food marke-ting literature on consumer preferences (e.g. Jiménez-Guerrero et al., 2012) and the opinion of 25 Spanish ex-perts possessing in-depth knowledge of the dairy sector. Previous research has found a positive and significant relationship between the consumer’s purchase decision and attributes such as certified PDO (Fandos & Flavián, 2006), organic labelling (Gil et al., 2000), or presentati-on format (Draskovic, 2010). More specifically, studies about dairy products, such as cheese, have found that the main attributes affecting preferences for this product are price, texture, size of the sales unit, PDO certification, ripeness (Tendero & Bernabéu, 2005) and color and packaging design (Eldesouky & Mesias, 2014).

However, it should be noted that attributes of agri-food products refer not only to the physical properties of the product (intrinsic qualities) but also to the conditions under which the latter is produced, distributed and retailed (extrinsic qualities) (Kirwan, 2006). Direct marketing strategies (DMS) include, for example, direct retailing to end consumers, restaurants or grocery stores. This strategy allows a farm operator to capture a larger share of the consumers’ food income budget by eliminating the intermediary in the supply chain (Detre et al., 2011). Indeed, previous research has suggested that farmers involved in DMS are more likely to achieve higher income levels (Govindasamy et al., 1999; Balogh et al., 2016).

In accordance with the above, in the present work we considered three product categories (milk, cheese, yogurt) and nine attributes of each product, namely milk-type composition, yield, organic labeling, PDO certification, maturation period, expiration period, size of the sales unit, returnable packaging and distribution channel.

Material and methodsTop

Product profitability measure

We performed an economic analysis of the trans-formation and commercialization of dairy products by valuing the incremental income and costs generated by these activities with respect to the primary production of milk. We used the margin per liter of milk (ML) as the profitability measure of the transformed product. To obtain the ML, we first calculated the margin per unit of product, defined as price minus variable cost (Horngren et al., 2016). There are three main different sources of variable cost: the raw milk, other raw materials (rennet, ferments, salt, rice, fruit, etc.) and packaging. In most cases the milk used is produced on the farm, so that the cost of milk is the price that the farm would have obtained by selling the milk to a processor (the opportunity cost). The cost of the direct materials acquired from external suppliers is valued at the acquisition cost.

To ensure comparability of products with different milk content and size of the sales unit, we divided the margin per unit by the liters of milk in each unit, obtaining the ML which indicates the value added per liter of milk.

Moreover, we can express ML as the difference between income per liter (IL) and variable cost per liter (VCL).

In this way, it will be possible to determine whether the effect of the explanatory variables on the ML is due to their influence on income or to their effect on variable costs.

It should be noted that our approach, focused on the ML analysis, guarantees the comparability of product margins because we avoid allocating indirect and fixed costs (Goldratt, 1990), which always implies some degree of subjectivity.

Sample and data

Although in the Spanish dairy sector most of the production is sold directly to the processing industry, some farms have embarked upon diversification stra-tegies. No specific register exists for dairy farms that perform production and commercialization of dairy products. For this reason, we resorted to identifying them via enquiries to different agents of the dairy sector (cooperatives, advisors, producer organizations, regulatory organisms of the different PDOs and organic agriculture, etc.). Our study centered on the four regions of Northern Spain (Asturias, Cantabria, Galicia and the Basque Country) where 79% of the dairy farmers and 59% of Spanish dairy production were located in 2012 (MAPAMA, 2016). Collaboration was requested from the 80 cases identified, and the participation of 49 farms was obtained, 14 of which are certified as organic with the remainder being conventional.

The study benefited from the collaboration of a group of 25 Spanish experts possessing in-depth knowledge of the dairy sector (farm advisers, scientists, farm union staff, representatives from government, policy-making, supply chains or rural economic development areas, etc.), who provided their opinion with respect to the factors to be considered in evaluating the success of these types of farm diversification initiatives. For these experts, the most important factors were the management ability of farmers and the differentiation of their production within the market via various strategies such as packaging, direct contact with consumers, sales in specialized shops, PDO, an organic label, and the attributes of the local and traditional product. This assessment by the experts was very useful for the elaboration of the questionnaire that was used during 2012 in order to collect the data for this study through face to face interviews.

The data collected, which refer to 2011 can be briefly described as follows:

- The median quantity of processed milk per farm is 280,000 liters. Approximately 50% of the sample transforms more than 40% of the milk produced by the herd, with 24% of the sample requiring external procurement of milk.

- The average sales structure by product type indi-cates that cheese represents 61% of total sales, liquid milk 26%, and yogurt 13%.

- The analysis of distribution channels reflects that direct sales to consumer represents 29.8% of total sales, grocery stores 33.1%, restaurants 23.9%, and large retail distribution 13.2%.

- The average investment dedicated to transformation and commercialization of dairy products amounts to € 258,500 per farm, 76% of which is allocated to assets related with milk transformation and 24% to the commercialization of products.

- The average number of workers per farm is 4.8, 1.7 of which are dedicated to livestock activity and the rest to activities related to the transformation and commercialization of products (representing 64% of employment in these farms).

- Cow milk represents 85.3% of the milk used for the elaboration of the different products, the remainder being either goat or sheep milk.

- Of all the references sold in the farms, 16.6% pos-sessed a PDO label, accounting for 31.8% of the total sales in our sample.

Apart from the price and direct variable cost of each product and fixed costs for transformation and com-mercialization activities in each farm, our database includes several variables that may explain product profitability. We classified these into product types, at-tributes of each product and control variables (Table 1).

Table 2 shows some descriptive statistics of the vari-ables used in the empirical analysis, which correspond to a total of 265 products manufactured and marketed by the 49 collaborating farms. The number of products by farm oscillates between 1 and 16. We considered only three classes of products but we distinguished between natural yogurt and yogurt with fruit. Also, given that the margin per liter may depend on size, we considered different sizes of the same product as different products.

Cheese products represent two thirds of all products in the sample, while the least represented one is yogurt. From Table 2, it is clear that cow milk is by far the most used input in the production processes of value-added products, while sheep milk is the least utilized. Organic production, PDO, and returnable packaging are product characteristics that are not very common in the sample. Sales are almost equally distributed among direct selling, restaurants and grocery stores, with only 13.2% in large retail firms. Family labor represents more than half of total labor. As for geographical location, almost 70% of products are processed in Galicia and Asturias.

Econometric model

As mentioned above, the model used in this study resembles the hedonic price model in which the price of the product is assumed to depend on several product attributes (Rosen, 1974). However, in this study the dependent variables were the margin, income and variable costs per liter. As such, the ML was assumed to depend both on product characteristics and control variables. The equation to be estimated is:

where i indexes products, ranging from 1 to 265, j denotes product types and attributes, and k indicates control variables; x is a set of variables representing pro-duct types and attributes, z are control variables, ß and d are the parameters to be estimated, and finally, e is the error term. Since ML is the difference between income per liter (IL) and variable cost per liter (VCL), equation [1] was also estimated using these two variables as dependent variables (see equations [2] and [3]). In this way, it will be possible to determine whether the effect of the explanatory variables on the ML is due to their influence on income or to their effect on variable costs.

Specifically, the x vector includes a set of dummies that indicate whether the product type is cheese or yogurt (the excluded category is liquid milk) as well as the following product attributes: proportion of sheep milk in the production of the good, proportion of goat milk (the proportion of cow milk is excluded), liters of milk per kg of product (yield), two dummy variables indicating whether the product is organic or is certified with a PDO, days of maturation in the production process, days of product expiration, the size of the sales unit (measured in kg) and a dummy variable indicating if the packaging is returnable or not. Finally, the distribution channel was considered by including the proportion of sales direct to the end consumer, to restaurants, and to grocery stores (the proportion sold to large retail firms is excluded). It is worth noting that the PDO characteristic was only observed for cheese. Although we have not presented it in the results, we have also considered the interactions between the type of product and the variable “organic” but no significant difference was found for the effect of this variable with regard to any of the three types of product.

The control variables included in the analysis were a set of regional dummy variables indicating whether the farm is located in Asturias, Cantabria or the Basque Country (Galicia is the excluded region), the proportion of family labor over total labor and the liters of milk processed in each product.

We did not include farm fixed effects in our estimation since the variable “proportion of family labor over total labor” varies across farms but not across products, and therefore it was perfectly correlated with a farm fixed effect.

ResultsTop

Regression analysis

The three equations were estimated by ordinary least squares. Table 3 shows the estimated parameters. The R-squared shows that the variables included in the empirical analysis explain between 71% and 83% of the variation in the dependent variables, which can be considered as an acceptable explanatory power, given the high degree of heterogeneity across farms and products in the sample.

Compared to liquid milk, any of the other products included in the sample generated a larger ML. In particular, cheese generates larger income and lower costs, while yogurt generates a positive impact on the ML due to a larger effect on income than on variable costs. While the proportions of sheep and goat milk increased both IL and VCL, only sheep milk exerted a positive impact on the ML. Thus, the higher cost of goat milk (compared to cow milk) was almost exactly transferred to IL, thereby generating a null influence on ML. However, the impact of sheep milk content on income was larger than its effect on variable costs, which generates a positive impact on ML. The quantity of milk per kg of product had a negative influence on ML due to its negative impact on income. Therefore, the larger the milk content, the lower the profitability of the product.

Organic production did not generate a larger ML, as the higher IL was offset by the greater cost of materials. Even when the impact on both income and cost of having a PDO was somewhat unclear, its effect on ML was clearly positive, which emphasizes the importance of a quality indicator such as this. The maturation period was not significant in explaining any of the three dependent variables. The expiration period, on the other hand, showed a positive impact on the ML, which was due to its influence on income given that its effect on variable costs was not significant.

The size of the sales unit exerted a small impact on costs but its effect on the ML was not significant. This seems to indicate that this variable is not an important aspect in determining the profitability obtained from any kind of product.

Returnable packaging had a negative impact on variable costs, as expected, but it failed to generate a significant effect on the ML. Hence, from an economic point of view the use of returnable packaging did not seem to be important.

Looking at product distribution, the most profitable method was direct selling to end consumers, which generates larger income and lower costs than large-scale distribution (which is the category omitted in the estimated equations). The largest impact on income was generated by stores sales, but costs were similar to those associated with large-scale distribution and also, its impact on ML seems to be almost equal than that obtained by direct selling. Restaurant selling also generated a larger ML than large retail firms.

Looking at the geographical localization of the business, producers in Asturias did not generate a larger ML than those located in Galicia. Products made in Cantabria are less profitable than those made in Galicia, which is due to the lower income achieved as there was no difference in variable costs. Products made in the Basque Country, on the other hand, are more profitable due to lower variable costs than those made in Galicia. Family businesses seem to generate larger margins due to the larger income earned given that the proportion of family labor has a positive impact on both variables.

Finally, while the amount of liters of milk - which is related to the amount of milk transformed for each product - could have a negative effect on IL because of possible discounts to stimulate larger volumes of sales, the results show that this variable does not have any significant effect on the dependent variables.

Coverage of fixed costs and break-even point

The profitability of dairy farm diversification ventu-res does not only depend on the ML of the products: a complete analysis of profitability requires taking into account the fixed costs related to processing and mar-keting activities. Unfortunately, we could not obtain information for allocating fixed costs to products. The-refore, in order to explore the farms’ profitability, we analyzed three different types of business according to the set of outputs produced. In particular, we considered ventures that only produce cheese (19 cases), those that only produce fluid milk (6 cases) and those that produce fluid milk, cheese and yogurt (5 cases).

We assumed that, in the short-term, all transformation and commercialization costs are indirect and fixed, so that they are grouped together under the heading “fixed costs”. Labor, the depreciation of fixed assets and other general costs were also considered as fixed costs. Labor cost was determined as the product of the average cost of labor (salary and social costs per worker) and the number of employees, both hired workers and family members.

As investments for the manufacture of products, we considered buildings, facilities, machinery and tools, and as investments for marketing and distribution we considered vehicles, sales outlets and vending machines. The investments in fixed assets were depreciated over their working life (20 years for buildings and 10 years for the remaining elements). In order to perform an economic analysis using current values, the costs of depreciation were determined applying current cost accounting to investments based on their current acquisition value (taking into account the evolution of the Spanish Consumer Price Index).

Table 4 shows for each venture type the average fixed transformation and commercialization costs, the average quantity of milk transformed, the average ML for the mix of products sold and the break-even point, calculating the quantity of milk that must be transformed to cover the transformation costs, the commercialization costs and also the total fixed costs (transformation + commercialization). From Table 4, it was clear that total fixed costs were highest for milk producers, although the three types of ventures considered showed large differences in the fixed cost structure between transformation and commercialization costs. Transformation costs per liter of milk were lowest for businesses producing fluid milk (0.226 €/L) and highest for those producing milk, cheese and yogurt (0.396 €/L), as was to be expected. Fixed commercialization costs per liter in milk ventures (0.284 €/L) were higher than in the other two types of ventures due to strong investments in vending machines.

From our calculations, 91%, 62% and 78% of the milk transformed were required to fully cover total fixed costs in each of the venture types respectively. These results show the importance of the ML generated by the products for the profitability of these ventures and at the same time highlight that a certain volume of production was needed in order to cover fixed costs and generate profits. In the sample studied, the businesses which only produce liquid milk showed the worst results, given that they generate a lower ML and require the greatest volume of production in order to achieve the break-even point. This was due especially to the higher fixed costs of commercialization.