Economic performance and efficiency determinants of crop-producing farms in Norway

1. Introduction

The United Nation has predicted that current world population of 7.6bn is expected to reach 8.6bn in 2030, 9.8bn in 2050 and 11.2bn in 2100 (United Nations, 2017). Finding a way to meet the growing demand for food represents a major challenge for farmers, policymakers and agricultural researchers. These global challenges also affect Norway. Increasing the quantity and quality of food in response to growing food demand requires improvement in the performance of the agricultural sector. Agricultural production growth in Norway is a topic of continuing interest to researchers and policymakers who aim to improve the economic efficiency and economic sustainability of primary agriculture. Good farm management and agronomic practices, in combination with efficient input use, are the best ways to improve farm productivity. In this context, agronomic practices are steps farms incorporate into their farm management to produce crop output. These steps include land preparation, time of sowing, crop rotation, use of new crop varieties, pest control, etc. While some practices may be designed to increase output, others may be directed to reducing labor time or costs.

Farms use different farm management/agronomic practices and resource combinations to produce crops. Thus, we can expect differences in performance between farms, with some less efficient than others. Farm inefficiency can be defined as the extent to which farmers are using more resources to produce a given level of output than the resources used by the best practice farmers.

In recent years, several studies have used “benchmarking” techniques to examine the efficiency in the agricultural sector (e.g. Kumbhakar and Lien, 2010; Koesling et al., 2008; Odeck, 2007; Flaten et al., 2010; Lien et al., 2010; Kumbhakar et al., 2014; Sipilainen et al., 2013). “Benchmarking” is defined as a measurement of the quality of an organization’s policies, products, programs, strategies, etc., in comparison with standard measurements, or with similar measurements of its peers, see for detail Kumbhakar and Lovell (2000), and Coelli et al. (2005). The objectives of benchmarking are: to analyze how the more successful organizations achieve their high-performance levels; to determine what and where improvements are called for; and to use this information to improve performance. Measuring such efficiency gaps and identifying the causes can be useful to both farm managers, taxpayers, consumers and to policymakers and planners seeking to help farmers to improve their performance (Singh et al., 2016).

An investigator may not only be interested in the level of economic performance but also might want to know which factors (exogenous variables) affect the level of farm performances (Smith et al., 2006; Parmeter and Kumbhakar, 2014).

The ability of a farm manager to convert inputs into outputs via a given technology is often influenced by “exogenous variables” that characterize the environment in which production takes place (Coelli et al., 2005) (different names have been used in the economics literature for exogenous variables, such as environmental variables, z-variables and determinants of inefficiency). Thus, the accurate measurement of the economic performance of the crop farms demands an understanding of differences in the working environment. The environmental factors include farm-specific factors, such as management skill, institutional constraints and attitude to risk, or innovations that are unmeasured but can be partially represented by observable variables such as age, experience, participation in farm improvement programs and education. Environmental variables can be expected to provide farmers with various types of opportunities and challenges, which ultimately affect their level of farm performance.

In recent years, several studies have used benchmarking techniques to examine the level and determinants of inefficiency in farming. For instance, Latruffe et al. (2004) analyzed the technical efficiency and its determinants for individual farms specialized in crop and livestock production in Poland. They reported that, on average, livestock farms are more technically efficient than crop farms. Moreover, soil quality and the degree of market integration proved to be important determinants of technical efficiency. Curtiss (2000) investigated technical efficiency and competitiveness for Czech crop production for the years 1996–1998 and reported that there was a high correlation between technical efficiency and competitiveness. Moreover, market conditions, transaction costs in marketing and bargaining power were positive determinants for competitiveness. Zhu and Lansink (2008) analyzed the impact of the CAP reforms on the technical efficiency of three crop-producing EU countries in the period 1995–2004. They reported that the mean technical efficiency was 75 percent in the Netherlands, 70 percent in Sweden and 59 percent in Germany. The ratio of crop subsidy to the total subsidy had a positive impact on the technical efficiency in the Netherlands and Sweden, but no significant impact in Germany.

Kumbhakar et al. (2014) estimated technical efficiency and its determinants for Norwegian grain farms for the period 2004–2008. They reported that resources were sub-optimally used. Off-farm activity, direct subsidy and entrepreneurial orientation were found to have negative effects on technical efficiency, while farm managers with more experience were likely to be more efficient than those with fewer years of farm experience. Lien et al. (2018) estimated technical efficiency of the Norwegian crop-producing farms observed from 1993 to 2014 and reported that the mean technical efficiency was 0.82–0.88. Moreover, off-farm work was found to decrease technical efficiency while direct subsidy payments had a positive and statistically significant effect on technical efficiency.

Previous studies of agricultural efficiency have given useful insights into farm performance. However, the contribution of farm management practices such as crop rotation to the economic performance of crop farms remains unclear. This paper contributes to the literature in a number of ways. First, we hypothesize that:

Second, unlike previous studies, which commonly estimated the technical efficiency of farms:

A farm achieves technical efficiency when it is able to minimize the use of inputs to produce a given amount of output so that no inputs are wasted. Allocative efficiency is achieved when the farm is able to use its inputs according to their respective relative prices. Measuring such CE gaps and identifying determinants can be useful to farm managers, policymakers, planners and advisers seeking to help farmers to improve their performance.

The rest of the paper is organized as follows: Section 2 discusses the nature of Norwegian agriculture. Section 3 addresses the approach to measuring farm performance, while Section 4 discusses model specification. Section 5 includes a discussion of the data and definitions of variables used in the cost function. Section 6 includes the results and discussion thereof. Section 7 covers conclusions and implications.

2. The nature of agricultural production in Norway

Norwegian farms are usually small and family-operated. Only 3.3 percent of the total Norwegian land area is farmland (Statistics Norway, 2013). Owing to the topography of the country, fields are often small, scattered and difficult to cultivate, which contributes to the high costs of agricultural production. With a relatively long winter in most parts of the country and a short growing season (five months on average), growing fodder, mainly grass, has the comparative advantage in most parts of the country. On the other hand, long summer days, with sufficient rainfall, are beneficial for crop production. Moreover, the cool climate limits the spread of pests and diseases (Steinshamn et al., 2016).

The primary objectives of the Norwegian agricultural and food policies, as set out in the White paper No. 11 (2016 –2017) are: long-term food self-sufficiency along with protection of the environment; creating more added value; and maintenance of small-scale farming in all regions. To achieve these objectives the government supports the farmers. The two main support instruments used by the government are border protection measures to limit or exclude competing imports, and budgetary payments (subsidies). The budgetary support includes market price support, special tax rules for agriculture, area payments, investment grants and grants for research and extension services. Market price support for most commodities, in the form of wholesale target prices, is provided. These target prices and most budgetary payments are negotiated on an annual July/June basis between the government and representatives of the two farmers’ unions (the Norwegian Farmers and Smallholders Union and the Norwegian Farmers Union), resulting in an Agricultural Agreement. Despite the various support measures, the value of support payments has been decreasing in real terms.

As in most developed countries, farming has become highly mechanized and the number of farms has been declining, with production becoming concentrated on fewer farms. According to Statistics Norway, the number of farms was 96,000 in 1991 and had declined to 42,000 in 2015. Moreover, 2.3 percent fewer farms were registered in 2016 compared to 2015. The number of farms growing only crops decreased by 29 percent over the years 2006–2016 (Statistics Norway, 2013). However, according to the NIBIO (2016) farm account survey report, the average size of farm holdings in 2015 was 37 ha, which was an increase of 8 percent compared to 2014. Moreover, the area of rented farmland has been increasing over time, and the average area of rented land per farm reported in 2015 was 17 ha (NIBIO, 2016).

The structure of farms in Norway has been regulated by the Norwegian Concession Act. Norwegian farmers face extensive farm policies which have effects on their choices of the size and scale of the farm. For example, the Norwegian Agricultural Authority manages the quantities produced of milk, meats, vegetables, potatoes, fruits and berries (Knutsen, 2007). Limited access to land and capital restricts productivity changes, as does policy regulation in the form of quantitative restrictions on milk supply.

Norwegian agriculture is so heavily subsidized that, without support, it would not be competitive with imports. There is a threat that Norway may be obliged by international pressures to cut back on border protection and on output-related subsidies. If that happened, it would force a dramatic and painful shift toward a more competitive agriculture. Therefore, there is a case to be made to take urgent steps to improve the productivity and management of farming.

3. Approaches to measuring farm performance

There are two main benchmarking methods in the literature to measuring the performance of farms: a parametric (econometric) approach, such as a stochastic frontier analysis (SFA); and a non-parametric approach, such as data envelopment analysis (DEA). In both methods, the basis for performance measurement is the radial contraction/expansion connecting inefficient observed points representing individual firms with reference points on the efficient frontier (There are other approaches too, for instance, Bayesian stochastic frontier (SF) (Koop and Steel, 2001), semi-parametric (Simar and Wilson, 2007) and stochastic DEA (Huang and Li, 2001), but these are not commonly used in empirical studies).

Based on a sample of producers, both the two main approaches involve estimating a “best practice” frontier for a specific industry or sample of firms. Each approach has its pros and cons. For details, see Coelli et al. (2005), Parmeter and Kumbhakar (2014) and Kumbhakar et al. (2015). The treatment of measurement error is the critical distinction between parametric and non-parametric approaches. SFA models can accommodate stochastic noise, such as measurement errors due to weather, disease and pest infestation that are likely to be significant in farming. The DEA approach is sensitive to outliers since the measurement error is ignored (e.g. Coelli et al., 2005; Barnes et al., 2009). Since farms in our study are sensitive to external random shocks, we have chosen the SFA approach to evaluate the CE scores and determinants of inefficiency.

Depending on the nature of the data set at hand, there are two classes of SFA models. If we have only one observation per farm, then a cross-sectional model has to be chosen to estimate the performance of each individual farm. A data set that consists of different farms that are observed at different time periods is called a panel data set. A panel data set contains more information, and therefore allows us to separate unobserved heterogeneity (farm-specific effects) from inefficiency.

Since the introduction of the SFA model by Aigner et al. (1977) and Meeusen and Van den Broeck (1977), there has been considerable research to extend the basic model, creating a new subfield in econometrics. Different SFA models have been developed based on different assumptions about the temporal behavior of the inefficiency. Reviews of much of this research are provided by Kumbhakar and Lovell (2000), Coelli et al. (2005), Greene (2008) and Kumbhakar et al. (2015).

In addition to estimating CE for each crop farm, it is useful to learn about factors that affect cost inefficiency between and within farms (Lien et al., 2018). The first SFA models that dealt with modeling the impact of exogenous variables (z-variables) on the level of inefficiency between and within farms are those of Kumbhakar et al. (1991) and Reifschneider and Stevenson (1991). In these cross-sectional models, the variance of the inefficiency term was specified as a function of a set of exogenous variables (z-variables). Later, Wang (2002) introduced a further generalization in which both the mean and variance of the inefficiency are functions of exogenous variables. For detailed reviews see Parmeter and Kumbhakar (2014) and Lien et al. (2018).

Among panel data models, the inefficiency specification used by Battese and Coelli (1995), commonly known as BC95, is frequently used in the empirical analysis of performance studies (Parmeter and Kumbhakar, 2014). Using the BC95 model, a researcher not only estimates the efficiency score, but also can investigate determinants of firm inefficiency (exogenous variables) in a single-step procedure. It is also possible to estimate the determinants of inefficiency in two steps, but that method creates biased results (Wang and Schmidt, 2002). In the single-step procedure, the parameters of inefficiency and the determinants are estimated together via maximum likelihood.

The general BC95 model for our panel data in a cost function form can be specified as:

where ln C_it is the logarithm of actual costs incurred by farm i in time t; f(y_it, w_it; β), the chosen function form (e.g. Cobb-Douglas, Translog); y_it, a vector of outputs; w_it, the vector of input prices; and β , the vector of parameters to be estimated. The component v_it is a symmetric disturbance (error term) capturing the effects of noises that are beyond the control of the farmers. The error term has both positive and negative effects, and satisfies the classical assumptions, i.e., v i t i i d ∼ N ( 0 , σ v 2 ) , v i t ⊥ u i t . Further, u_it is a one-sided non-negative term, accounting for inefficiency. In the BC95 model, u_it is obtained by truncating (at zero) the normal distribution, i.e., u i t ∼ N + ( μ i t , σ u i t 2 ) and μ_it = z_itδ,. z_it is a vector that includes exogenous variables associated with variability in the efficiency score and δ is an unknown parameter to be estimated.

The CE is the ratio of the minimum cost of each farm (exp(f(y_it, w_it; β)+v_it)) to its actual cost (exp(f(y_it, w_it; β)+u_it+v_it)). i.e. CE = exp(−u_it). CE has a value between 0 and 1, with 1 defined a cost-efficient farm. Since only the sum of two error terms (ε_it=u_it+v_it) can be observed, the farm’s cost inefficiency index can be estimated using the conditional mean of the inefficiency term, as proposed by Jondrow et al. (1982), i.e. E[u_it/v_it+u_it].

4. Empirical model specification

Consistent with the farm efficiency literature (e.g. Christensen and Greene, 1976), we estimated a transcendental logarithmic (translog) cost function incorporating Hicks-neutral technology change. We used panel data but, to simplify the notation, we have dropped the subscripts i and t, where i would denote firm i = 1, …, n and t would denote time, t = 1, …, t. Our specification of the cost function C, including the prices of inputs (w_j)j=1, …, J, outputs (y_k) k =1, …, K, and a local wage distribution index (r_m) m=1, …, M, is:

The error terms ε splits into two components, i.e., ε≡v_it+u_i. As discussed in section 3 component u_i captures cost inefficiency with u i t ∼ N + ( μ i t , σ u i t 2 ) . v_it is the error term capturing random shocks and assumed to be symmetric and to satisfy the classical assumptions i.e., v i t i i d ∼ N ( 0 , σ v 2 ) . All Greek letters are parameters to be estimated. The trend variable, t, is included to capture technology change and starts with t = 1 for 1991 and increases by one annually.

Economic theory requires imposition of price homogeneity and symmetry restrictions on the parameters. Symmetric restrictions require ∑ j = 1 J ∝ j = 1 , and ∑ k β k k = ∑ j β j j = ∑ m ∅ m m = 0 . An easy way to impose price homogeneity is to divide the all inputs prices and total cost by an arbitrary chosen input price. Thus, in Equation (2) the left-hand side is re-defined as ln C=ln (C)/(w₁), and all input prices are re-defined as ln w_j=ln (w_j)/(w₁), i.e., we divided all input prices and the total cost by wages before estimating the translog cost function. Given the translog specification in (2), the farm-specific cost inefficiency and marginal effects of the exogenous variables are calculated following the procedures of Jondrow et al. (1982) and of Wang (2002), respectively.

As discussed above in Section 3, we included exogenous determinants of farm CE in our model. The choice of variables in the final model was based on two criteria. First, we considered data availability. Many variables that could affect crop management, such as skill, education level, soil type and slope or aspect of the farmland, were not available in our data set, so could not be included. Second, we considered the literature available on the subject, for example, Latruffe et al. (2004), Bozoğlu, and Ceyhan (2007) and Lien et al. (2018). As a result, we chose the following variables:

1. Crop rotation (z₁) – reflecting the impact of rotation system on crop and forage production, measured as the ratio of non-cereal crops such as root and legume crops to the total cropped area. We expected the farm managers’ decisions to rotate the type of crop grown on the land would make a positive contribution to the performance of their farms. Our expectation is in line with other research findings that crop rotation reduces agriculture’s dependence on external inputs through internal nutrient recycling, maintenance of the long-term productivity of the land (Gebremedhin and Schwab, 1998). Crop rotation can improve the fertility of the soil (Reckling et al., 2015) and can increase yields to a higher degree than pesticide intensity and tillage use (Deike et al., 2008).

2. Land tenure (z₂) – the proportion of the total farmland that is rented. Farm managers’ decisions to rent land or not depend on the price of land and other factors. We expected a positive contribution to this variable because we hypothesize that well-performing farms more commonly rent extra land. Deininger and Byerlee (2011) support our hypothesis and stressed that land rental markets can help transfers of land from producers with low levels of productivity and low comparative advantage in agriculture to more efficient farmers.

3. Off-farm activity (z₃) – the ratio of time of owner plus partner allocated to off-farm activity to the total time assigned to the farm. We expected a positive contribution from this variable on the basis that off-farm experience and income are likely to promote better farm management. Our hypothesis is supported by the literature. Off-farm income has been found to relax cash constraints and to allow farmers to spend significantly more on improved farming technologies (e.g. Pfeiffer et al., 2009; Stampini and Davis, 2009).

4. Government support (z₄) – the ratio of government assistance to the total agricultural income. We hypothesized a positive contribution from this variable because government support can motivate farm managers to invest in new technology and may facilitate such investment by easing cash flow constraints. In line with our hypothesis, there is evidence that government support helps to promote better use of economic resources (e.g. Ferjani, 2008; Kumbhakar and Lien, 2010).

5. Data

Our data source is the Norwegian Farm Accountancy Survey collected by the Norwegian Institute of Bioeconomy Research (NIBIO). The survey participants are selected from a list of farmers, randomly drawn from the register of grants kept by the Norwegian Agricultural Agency. The data include production and economic data collected annually by NIBIO from about 1,000 farms in all regions of Norway. The number of participants varies from year to year. Approximately 10 percent of the surveyed farms are replaced per year to incorporate changes in the population of farms in Norway. Participation in the survey is voluntary. There is no limit on the number of years a farm is included in the study. Some of the farmers have participated in more than 20 years, and others have started participating for the first time. To accommodate panel features in estimation, we included only those farms for which at least three consecutive years of data are available.

To assess the efficiency and productivity growth, we needed to be sure that farmers under consideration are comparable. To obtain a homogeneous group, we choose only farms in the two main cropping areas where 98 percent of the cultivated land was located. Figure 1 shows that out of the 286,000 ha of land cultivated for grain and forage production in 2012, 81 percent was located in the Eastern Norway and 17 percent in Central Norway.

The set of data used in this study is a farm-level unbalanced panel data with 3,885 observations from 455 farms specialized in the production of grain and forage crops during the period 1991–2013.

6. Estimation results and discussion

7. Conclusion and implication of the study

The aim of this study was to investigate whether farm management and socioeconomic factors contributed to improving the performance of crop farms in Norway. We found that farm resources were widely used sub-optimally, i.e., there are farms that produced lower outputs from the inputs they used or used more inputs to produce the same output, compared to the best performing farms.

The findings revealed that agronomic and socioeconomic factors affect the cost-efficient level of crop production. A good agronomic practice – crop rotation – was found to have made a positive and statically significant contribution to reducing cost. Moreover, the magnitude of crop rotation contribution was higher in the central region and increased over time in all regions. Off-farm activity positively and significantly enhanced the performance of the farms, and farmers renting land are more cost-efficient than those not doing so. Our analysis also confirms that government support has an association with improved CE, probably by relaxing farmers’ financial and liquidity constraints, enabling them to purchase new technologies that can enhance farm crop output. We also found that farmers renting land is more cost-efficient than those not doing so.