1
Impact of the abolition of the EU milk quota on
the Dutch dairy industry
JEL classification: Q18, D3
Author: Inge Klaver
Student number: 10220666
Date: 29-06-2015
Bachelor thesis: Economics and Finance Supervisor: M.O. Hoyer Msc
This paper examines the possible effects of the recent abolition of the EU milk quota on the dairy industry in the Netherlands and investigates farm size inequality for all provinces for the period 2000-2012. A literature review is conducted to illustrate the complexity of the matter and evaluate the expectations about the impact of the quota abolition. Subsequently, an own analysis is undertaken by calculating Gini coefficients and running a fixed effects model with several potential explanatory variables. The farm size inequality is found to have decreased slightly for the country as a whole but surely differs across provinces, with herds in Zeeland being most
unequally distributed and in Flevoland most equal. The milk quota proves as an important positive driver for farm size inequality in the Netherlands and the time trend is negative. Results on the province level is ambiguous and requires further research.
2 Statement of Originality
This document is written by Student Inge Klaver who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
3
CONTENTS
1. INTRODUCTION ... 4
2. LITERATURE REVIEW ... 5
2.1 Background information ... 5
2.2 Milk production and prices ... 7
2.3.1 Cost function structure analyses ... 7
2.4 Farm structures ... 8
3. DATA & METHODOLOGY ... 10
3.2 Methodology ... 12 3.2.1 Gini coefficients ... 12 3.2.2 The model ... 13 3.2.2.1. Motivation model ... 13 3.2.2.2 Model specifications ... 13 3.2.3 Variables ... 14 4. RESULTS ... 15 4.1 Gini coefficients ... 15 4.2 Model estimates ... 17 4.2.1 National level ... 17 4.2.2 Province level ... 18 4.3 Discussion of results ... 20 5. CONCLUSION ... 21 6. BIBLIOGRAPHY ... 22
4
1. INTRODUCTION
As of April 1st of this year, the EU milk quota that has been in place for over 30 years is completely abolished. Initially this quota was introduced as an additional policy tool in 1984 by the European Commission (EC) at a time when dairy production in the European Union (EU) increasingly exceeded demand. The EC was constantly required to intervene in the market and buy dairy products which were then kept in stock by the EC and built up to the infamous ‘milk lakes and butter mountains’ (Corron et al., 2007).
The quota system for milk imposes a maximum of farms-specific milk deliveries and penalizes production above the quota. By now, the milk market has stabilized significantly and other tools of the EU’s Common Agricultural Policy (CAP) seem to have targeted dairy market objectives more specifically. This has made the EC to decide during the 2003 policy reforms to gradually increase the quotas on milk and eventually abolish the regime in 2015.
This gradual increase has occurred over the past years in small steps at a time. In 2008 the quotas were decided to increase by 2% that year, and after that first step the quotas were
increased by 1% for five consecutive years. Moreover, there was a fat correction factor
adjustment meaning another 1% increase in allowance effectively (EC, 2012, p.6). This approach was called the “soft-landing strategy”.
Since the introduction of the milk quota in 1984 the dairy industry in the Netherlands has been through several developments. Prominently, the number of dairy farms has decreased significantly. According to the Central Bureau for Statistics (CBS), an average 7 out of 10 dairy farms have sold their cows since 1984 until 2014. At the same time, the average size of dairy farms has increased, in fact almost doubled.
Several studies have been conducted to investigate the possible impact of the milk quota abolition on the dairy industry both in the EU as a whole or the Netherlands in particular. This impact is of interest because of the vast size of the dairy industry in the Netherlands and the active stance of government for protection of the industry. Moreover
In this paper, the impact of the EU milk quota abolition will be evaluated in a literature review and the effects of the quota on dairy farm size inequality in particular is assessed through calculation of the Gini coefficient for different periods and regions in the Netherlands. With
5 these results a fixed effects model will be estimated to explore the drivers of farm size inequality across the country.
2. LITERATURE REVIEW
A great amount of research has been done on the subject of the impact of agricultural
policies, and on the milk quota in particular also. This work is of importance to both the industry and policymakers. Effects on production developments (Jongeneel et al., 2012), output and price (Lips and Rieder, 2005), entrepreneurial behaviour of farmers (Bergevoet et al., 2004), structural farm change (Jongeneel and Tonini, 2008)(Piet et al, 2011), herd-size change (Huettel and Jongeneel, 2011), environmental impacts (Kempen et al, 2011) etc. have been studied on a regional level, member state level and EU-wide. Different theories and models have been used for these analyses such as a CGE analysis (Lips and Rieder, 2005), partial equilibrium model (Bouamra-Mechemache et al, 2008) or a Markov chain model (Huettel and Jongeneel, 2011). This proves that the matter of agricultural policy implementation and effects is quite complex and many factors are contributing to all developments that can be interpreted and evaluated in several ways.
The following section will present a review of the existing literature on the subject after providing some background information on the introduction and abolition of the milk quota and developments and adjustments in the period in between.
2.1 Background information
Agricultural and dairy policy in the EU in the 1980’s already consisted of import tariffs, export subsidies and intervention buying (Huettel and Jongeneel, 2011, p.498). However, these procedures did not prevent overproduction from occurring and other measures had to be taken to stabilize the market. 1984 was the year that the EU milk quota was introduced. The level was set at the amount of milk production in the EU in 1981 plus 1 per cent (Jongeneel and Tonini, 2008, p.3). For the Netherlands, this implied a rather sharp reduction since milk production in 1984
6 was already higher than in 1981. Under the quota regime, each producer got a farm-specific quota and producing more than this was fined. The so-called “super-levy” per kilogram was considerably higher than the milk price so this would impose extra costs on the farmer. Milk quotas were tied to land in the Netherlands, so tradability was limited and changing milk production cumbersome (Boere et al., 2014, p.1). This was partly resolved when quotas were allowed to be permanently transferred through temporary lease of land from 1989-1990.
Intervention prices for butter and skimmed milk powder (SMP) are still in place, but the amount of intervention purchases has decreased drastically since the 1980’s (Bouamra-Mechemache et al., 2008, p.474). Nonetheless, they still play a role in equilibrating the milk market.
Another number of adjustments to the CAP was put in place after the 2003 EU
agricultural policy reforms. The intervention prices for SMP and butter were reduced with 25 percent and 15 percent in total respectively over the years 2004 till 2008 (Samson et al., 2012, p.3). The reduced lower bound caused more volatility in the prices as is normal for commodity prices. To compensate for this liberalising measure the dairy premium was introduced and distributed to milk producers based on production volume. From 2007 onwards though, these were replaced by Single Farm Payments in the Netherlands which were not volume-related but provided to selected farmers to ensure income (Samson et al., 2012, p.3). A period to note for this study is the years 2007-2008, since there was a price spike in 2007 after which milk prices fell significantly again and producers’ income also declined rapidly. After this period of severe price volatility the situation recovered in the second half of 2009 (EC, 2010, p.3).
Governments and farmers have been anticipating on the quota abolition that happened 1st of april 2015 for a while. Without milk quotas, it could be that farmers start producing limitless, but there are already efforts in the Dutch government to implement new measures to keep this within certain boundaries to minimize negative externalities from large-scale milk production. The state secretary of economic affairs Dijksma’s idea is to keep milk production levels tied to land. She has presented a proposal to peg pasture levels to land holdings of a farm, with as motivation that farmers should be able to process their own cows’ pasture and that allowing upscaling only with land increase would stimulate grazing (Dijksma, 2015). It is highly probable that this policy will find support and be implemented before 2016 in the Netherlands.
7 2.2 Milk production and prices
The most easily observable effect of milk quota in the milk market is the effect on production and prices, but since the abolition has only been executed since a few weeks,
researchers have only been able to make some predictions about the future production and price developments.
Lips and Rieder (2005) use an applied general equilibrium model for an analysis at a member country level but take into account bilateral trade flows. They adjust the model to serve for their purpose by implementing the quota rents as additional factor payments. The possible effects of the abolition of the quota on production and prices is estimated with this model. Their prediction of the EU-wide effect is a 22% decline of the raw milk price and a 3% increase in output. This effect is heterogeneous across countries though. For example, there is nearly no price change in Portugal and a 33 percent price decrease in Ireland) with the Netherlands expected to increase production by more than 10%.
Bouamra-Mechemache et al. (2008) in their more recent report use a different approach, they namely analyse the different possible effects of different quota abolition scenarios in a spatial and dynamic equilibrium model of the world dairy industry. They predict that the ‘soft-landing’ strategy as has been performed would yield a price decrease by 8.6 percent and production increase by 3.2 percent.
Even though the findings of these papers differ in the actual numbers they predict, they agree on an effect of a moderate price decrease and production increase.
2.3.1 Cost function structure analyses
As opposed to Bouamra-Mechemache et al. (2008) and Lips and Rieder (2005), the 2005 paper from Ooms and Peerlings takes a micro approach to the event of quota abolition. They estimate the production function of milk farmers based on the assumption of profit maximisation. To avoid the problem of endogeneity in their model they use an IV estimation method
Generalised Method of Moments (GMM). In this manner, they focus on the case of the
Netherlands and determine the effects of the 2003 dairy policy reform. Given a price decrease of 15 percent, farmers’ profits are estimated to fall by 15 percent, and given a price fall of 21
8 percent the profits would decrease with 22 percent. It concludes that small dairy farms are most likely to suffer from these setbacks and end up with a negative income.
Following Frahan et al.’s example (2010) of their study on the farm-level supply and income effects from removing milk quotas and reducing producer prices for a panel of Belgian dairy farms, Samson et al. (2012) more recently undertook a comparable study for the case of the Netherlands. They constructed a quadratic cost function to establish the potential impacts of feed price increase, milk price decrease and milk quota abolition on milk production. Since Dutch producers cannot maximise profit by choosing the optimal production level (quota is binding) Samson et al. (2012) constructed a cost minimisation model based on shadow prices and quota rents using farm-specific data allowing them to optimally explore individual farmer’s behaviour. Their conclusion is that the potential increase in the milk production by Dutch farms after quota abolition is large, but partly offset by decreasing margins due to higher costs (feed prices) and lower revenues (milk prices).
2.4 Farm structures
The developments in farm structure in the years since the introduction of the milk quota in 1984 have been rather fundamental. The overall trend is that there are fewer dairy farms, but on average these farms hold much more animals. This trend is not only of interest for the farmers community itself, but is also a concern for policy makers because of its effect on competition, income distribution, regional development and rural employment. There have been several studies in the EU and elsewhere on the changes in dairy farm structure and the share that the EU milk quota has in this phenomena. A few will be discussed here.
Huettel and Jongeneel (2011) investigated the impact of milk quotas on the size structure of dairy herds in Germany and the Netherlands. To this end they used Markov chain models based on the assumption that dairy farmers act in a way to maximize profits over a number of time periods. This model takes into account the herd-size growth, herd exits and the interrelation between herd-size classes. This is important under the quota regime because there is only
possibility to grow in this environment if other herds contract or exit and new quota rights become available. Based on their results on structural change in the pre-quota (1972-1983) and quota period (1984-2006) in the Netherlands, Huettel and Jongeneel made a prediction for the post-quota period. They acknowledge that the environment after the abolition is not directly
9 comparable to the period before 1984, mainly because of lower milk prices and higher volatility, but they state that the trend towards increasing herd size is obvious. A counteracting effect might be the lack of means to make the investment for expansion due to the low milk prices expected. Mishra, El-Osta and Gillespie (2009) used the Gini coefficient concept and a large farm-level dataset to investigate the impact of government payments on income inequality among farm households in nine farming resource regions of the U.S.. The main result was that government payments are an important equalizing factor for income distribution of farms in the U.S..
Piet et al (2011) also used the Gini coefficient to identify farm size inequality but for the different ‘départements’ in France from 1970 until 2007. They found that inequality among farms increased only marginally in France. However, they also found that differences in evolution across regions prevailed. To further explore the main determinants of change in farm size inequality they used a second-stage robust IV-GMM regression with variables relating to farm-support programmes (including a dummy for the milk quota), the profitability of farming and other controls. Their results show that agricultural policies indeed affect farm size inequality in France. Milk quota was found to have decreased inequality, especially in the plains.
In general, the findings are that the quota regime brings about more market rigidity and hampers growth of efficient farms. By combining research on the impact of the introduction of the milk quota and predictions about the impact of the abolition and the latest data a thorough analysis can be done. Still, identifying the main drivers of structural change in agriculture and measuring their impacts is challenging since there are several factors that play a role in farmers’ strategies.
10
3. DATA & METHODOLOGY
The aim of this paper is to analyse developments in farm size inequality in the Netherlands and assess to what extent this is caused by the EU milk quota or other factors. This will be done by firstly calculating Gini coefficients for the provinces through the years and then estimate a fixed effects model using least squares with the Gini coefficients as dependent variable and several explanatory variables including the milk quota.
3.1 Data
The data used in this study derives from the database of the Central Bureau for Statistics (CBS). For the analysis of developments in farm size inequality the number of farms in several size classes in terms of dairy herd size is used. The data available divides the farms into six size classes. These figures are composed by the CBS from the agricultural census which is conducted yearly and is used for research and policy making both nationally and internationally. Farmers are obliged to participate in this census and thus it offers a very high reliability. The reference date for the number of cows per farm is the first of April each year. There are no missing values so the dataset is balanced.
Data for the years 2000 until 2012 are provided on a regional basis in order to compare the developments across the country. The regions are the twelve provinces of the Netherlands. As regional borders and municipalities shift occasionally these regions do not stay exactly equal over the period estimated. Moreover, the farms are included in the region where their head establishment is in, this may intermittently differ from the location where the actual farming takes place. These variances are minor though and will be assumed insignificant for this study.
11
Year 2000 2006 2012
Total number of farms 35,962 27,087 21,861 Total number of cows 2,839,493 2,544,783 2,678,213 Average farm size (in cows) 78.96 93.95 122.51 Farms with less than 50 cows 12,224 7,425 4,249
as a share of total 34% 27% 19%
Farms between 50 and 200 cows 10,200 17,970 14,495
as a share of total 28% 66% 66%
Farms bigger than 200 cows 1,201 1,692 3,117
as a share of total 3% 6% 14%
Table 1 Source: own calculations based on CBS data
Table 1 and figure 1 and 2 summarize the data and illustrates the overall evolution of farm numbers, cow population and herd size between 2000 and 2012. Number of farms
decreased with almost 40 percent. The total number of cows also declined, but much less, only 5.6 percent over the whole period, with a lower count half-way the period but with no large volatility. Consequently, the average size of Dutch farms in terms of herd size increased
drastically, from 79 in 2000 to 123 in 2012, which is an increase of 55 percent or 3.7 percent per year.
However, while the number of farms has decreased everywhere and the average size increased, it cannot be concluded yet that farm size inequality in the regions changed based on these observations. Thus, further computations and analysis is required.
0 5000 10000 15000 20000 25000 30000 35000 40000 Figure 1
total number of farms total number of cows (x100)
0 5 10 15 20 25 30 35 40 45 1-20 20-50 50-100 100-200 200-500 500-> Sa hre o f fa rm s (%)
Size classes (cows)
Figure 2
2000 2012
12
3.2 Methodology
3.2.1 Gini coefficients
Farm size inequality will be measured by calculating the Gini coefficient, or the
concentration ratio. This can be calculated by constructing a Lorenz curve for size distributions (Gastwirth, 1972). The Lorenz curve is a function of the cumulative proportion of ordered individuals, in this case dairy farms, plotted against the corresponding cumulative proportion of the farm size in total, as illustrated in figure 3. With a sample of n classes the plot will show n points which can be connected to form the Lorenz curve. If all farms had the same size, the plot would represent a straight 45-degree line called the equality line. The degree of deviation of the Lorenz curve from the equality line is the level of inequality in the sample. The Gini coefficient quantifies the amount of inequality and is defined as the ratio between the area enclosed by the equality line and the Lorenz curve (area A) and the total area under the equality line (area A+B). If farm size were equally distributed, the Gini coefficient would be zero and vice versa. Thus, the higher the coefficient the more heterogeneous the distribution is.
In this paper, the integral of the convex hull represented by the plotted points in the graph is calculated not by fitting a curve as sometimes suggested in the literature but by actually
calculating the area with the coordinates of the points. Since the number of points is limited this seems the best way, Piet et al (2011, p.5) chose the same approach in their paper after concluding the results did not differ greatly from fitting a parametric curve.
13 3.2.2 The model
3.2.2.1. Motivation model
The data we have is cross-section/time series data or panel data (provinces over the years 2000-2012). This allows us to control for variables that cannot be observed or measured like historical factors, geographical characteristics or difference in farming practices across provinces; or variables that change over time but not across provinces (i.e. national policies). This is, it accounts for individual heterogeneity.
Given this data set and the objective of our analysis, namely only estimating the impacts of variables that vary over time, we will use a fixed-effects (FE) model to explore the
relationship between the chosen predictors and the Gini coefficients within a province. FE allows us to eliminate unobservable characteristics that differ across entities but not across time (Stock and Watson, 2012, pp.396-398). This way we can avoid omitted variable bias caused by
unobservable province-specific variables. The model allows for different intercepts for the provinces. We assume that these time-invariant unit-specific characteristics are not correlated with other individual characteristics.
A Hausman test is performed to confirm whether the FE model is appropriate for this data set. Moreover, a modified Wald test for groupwise heteroscedasticity in fixed effect regression model is performed to estimate whether it is needed to control for autocorrelation and
heteroscedasticity. On the basis of these tests, the most appropriate model proves to be a fixed effects model with robust standard errors.
3.2.2.2 Model specifications
The equation for the fixed effects model is:
𝑌𝑖𝑡 = 𝛽1𝑋𝑖𝑡+ 𝛼𝑖 + 𝑢𝑖𝑡 [eq.1]
Where
- α1 (i=1…n) is the unknown intercept for each entity (n entity-specific intercepts).
- Yit is the dependent variable (DV) where i=entity and t=time
- Xit represents one independent variable (IV)
- β1 is the coefficient for that variable
14 In this case, there are 5 different X’s and 5 Beta’s to be estimated, including a time coefficient.
3.2.3 Variables
The variables used in the regression are presented in table 2. All data is retrieved from the databank of the CBS and covers the years 2000-2012.
Variables Abbreviations Definition Unit Level
Time t Year - 1999
Agricultural area ShAgrArea Share of cultural land in total province area
Fraction Province
Average Economic value
AvEcVal Average economic value per farm
1000 € Province
Average house value AvHouVal Average house value 1000 € Province Milk quota lnquota Natural logarithm of percentage
increases in quota
x National
Table 2 Regression variables
Due to a lack of data availability, in order to run a regression with as many relevant variables as possible it was necessary to choose other variables to provide as a proxy for data that was unavailable.
The dependent variable is the Gini coefficient as calculated in the earlier explained manner for our periods and regions.
A time variable is included to account for the influence of technological advancement or changing overall economic conditions and other time-related variables that might have an effect on the dependent variable. Thus time is no causal factor in its own but represents these changes. An important driver for farm size inequality is expected to be profitability. Since we do not have access to figures of costs or margins on the province level we use a proxy for this. Average economic value per agricultural farm in the province over the years 2000-2012 will be an independent variable in the model. The type of farms accounted for here is farms with grazing livestock, which includes more than only milk cows but will do as an estimate.
15 Two variables are included in the model to measure the level of land competition. One is the share of agricultural land in the total province area. The second is average house value in a province, which acts as a proxy for land prices since this was the closest data available on the province level. A lower land price and a higher share of agricultural land in a province are anticipated to indicate less competition in agricultural land markets.
Lastly, the milk quota is included as a variable. The percentage increases of the past few years are taken as a natural logarithm and act as an independent variable.
4. RESULTS
4.1 Gini coefficientsTable 3 shows the results of the Gini coefficient calculations as explained in the previous section for all different provinces and the Netherlands as a whole from the year 2000 till 2012. As stated before, the Gini coefficients are estimated in such a way that the higher the coefficient, the higher the inequality, or the more heterogeneous the distribution. Table 3 shows that the farm size inequality in the Netherlands on average only decreased slightly. The coefficient for 2012 is 0.344, only 0.030 lower than the coefficient of 0.374 in 2000.
Table 3 Years Provinces 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 Groningen 0,367 0,365 0,353 0,341 0,328 0,324 0,339 0,342 0,327 0,329 0,340 0,337 0,328 Friesland 0,304 0,305 0,301 0,290 0,288 0,290 0,300 0,301 0,296 0,299 0,307 0,312 0,307 Drenthe 0,352 0,356 0,338 0,330 0,316 0,325 0,340 0,343 0,318 0,321 0,330 0,327 0,316 Overijssel 0,358 0,361 0,355 0,337 0,335 0,336 0,346 0,352 0,333 0,330 0,338 0,340 0,323 Flevoland 0,299 0,292 0,270 0,279 0,285 0,284 0,290 0,291 0,284 0,285 0,300 0,297 0,306 Gelderland 0,413 0,413 0,406 0,386 0,384 0,386 0,398 0,400 0,370 0,371 0,373 0,376 0,363 Utrecht 0,334 0,338 0,331 0,320 0,314 0,310 0,321 0,336 0,322 0,324 0,331 0,334 0,335 Noord-Holland 0,369 0,378 0,374 0,345 0,341 0,338 0,355 0,364 0,348 0,352 0,362 0,359 0,355 Zuid-Holland 0,358 0,356 0,350 0,326 0,330 0,325 0,338 0,355 0,350 0,343 0,348 0,359 0,355 Zeeland 0,485 0,504 0,478 0,441 0,443 0,447 0,465 0,491 0,466 0,436 0,453 0,432 0,424 Noord-Brabant 0,358 0,364 0,346 0,318 0,317 0,321 0,337 0,342 0,324 0,320 0,325 0,327 0,324 Limburg 0,405 0,403 0,398 0,370 0,361 0,369 0,382 0,394 0,371 0,382 0,376 0,358 0,349 Netherlands total 0,374 0,377 0,368 0,349 0,345 0,346 0,358 0,363 0,346 0,345 0,351 0,352 0,344
16 The findings do differ significantly across provinces, with for example the lowest Gini in 2012 of 0.306 in Flevoland and the highest of 0.424 in Zeeland. The relatively high Gini
coefficients of Zeeland in all years is fairly evident anyways. Moreover, the coefficients of Zeeland are quite volatile over the years. The province with the lowest implied Gini coefficients is Flevoland.
Figure 4 more clearly displays the development of the implied Gini coefficients of the provinces over the years. It visualizes heterogeneity and helps detect patterns or unusual
observations over time. For all provinces, the levels of the Gini are rather volatile, and the graphs show some mild peaks and dips. It can be seen that all provinces show an increased inequality around the year 2008, which was discussed as a period of high volatility in milk prices and lowered farm income. This might explain the increased inequality since some farms might have proved to be better prepared for such a setback and able to continue their production whereas other farms might have ceased to continue. Since prior research has not covered this crisis period, it cannot be confirmed that this is a common finding.
Figure 4 .3 .4 .5 .3 .4 .5 .3 .4 .5 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015
Drenthe Flevoland Friesland Gelderland
Groningen Limburg Noord-Brabant Noord-Holland
Overijssel Utrecht Zeeland Zuid-Holland
G
in
i
Years
17 4.2 Model estimates
4.2.1 National level
Table 4 presents the estimates of the fixed effects model for the Netherlands as a whole. Specification (1) is the robust fixed effects model with clustered standard errors. This model seems to give the best estimates as it yields the highest R2 together with specification (2) but also has the most significant coefficients. Specification (2) is the regular fixed effects model and specification (3) is the fixed effects model without the time variable. Overall, the results are a bit poor though not surprising, being that the model fails at explaining the variance in Gini
coefficients in the Netherlands very accurately.
Table 4 Specifications Variables (1) (2) (3) t -0,00462 -0,00462 - 0.0016287*** 0.0014261*** - AvEcVal 0,00006 0,00006 -0,0001555 0,00040 0,00018 0,00001777 ShAgrArea 0,01174 0,01174 0,1458688 0,20869 0,16101 0,1607753 AvHouVal 0,00010 0,00010 -0,0000741 0.0000541* 0,00007 0,0000364** lnquota 0,33121 0,33121 0,086555 0.097148*** 0.1289778** 0,1080223 _cons 0,34448 0,34448 0,3152876 0.1203553*** 0.0780779*** 0,0801409*** R2 within 0,21530 0,21530 0,1561 rho 0,92029 0,92029 0,90323
Source: Own estimates
*,**,*** imply significance level at 10, 5, 1 per cent respectively
The variable of average economic value of farms is not significant at any level in any model for the Netherlands. This could mean that it is not a good proxy for farm profitability or that profitability is no driver for farm size inequality, at least not in the Netherlands from 2000-2012. The same goes for the variable of the share of agricultural area, which was supposed to serve as a proxy for competitiveness. The variable of average housing value however is only significant at the 10 percent level in robust model (1) and at the 5 percent level in model (3).
18 Average housing value has only increased steadily over the period so this outcome is to be expected in the absence of a time variable since the coefficient in model (3) is negative just like the time variable in the other two models.
The time variable included in model (1) and (2) is significant at the 1 percent level even if it is not robust. This means that the factors included in this variable, which were overall
economic conditions and technological advancement, and maybe some other time variant factors not included already, have a significant effect on the Gini coefficient in a negative way. This implies that as time goes on, the farm size inequality decreases.
The most interesting variable for this study is the milk quota. This is highly significant in the first model, and positive. This means that as the milk quota increases, the Gini coefficient increases and farm size inequality rises. The coefficient is quite large also at 0.33121. This is an interesting finding and in line with the literature.
4.2.2 Province level
Table 5 shows the estimates of the robust fixed effects model for all provinces. The fit of the model greatly varies across provinces. For Drenthe, the within R2 is as high as 0.97150 but for Utrecht a mere 0.41630 share of the variance of the Gini could be explained by the model. For Utrecht, Zeeland, Groningen and Limburg none of the variables’ effects are found to be significant.
Yet again the variable of average housing value performs quite poor as a driver for the dependent variable in this case. It is only weakly significant at the 10% level for Friesland, Noord-Holland and Overijssel. These provinces are not located in the same part of the Netherlands so that cannot be an explanation for the close outcomes for this variable, nor are they all very urban or rural areas.
The time variable does not indicate such a strong effect as it did for the national level either. For most provinces it is negative still.
As opposed to the national model, the variable share of agricultural area constantly appears with a negative coefficient on the province level (with one insignificant exception). This would imply that the greater the share of agriculture in a province, the smaller the farm size inequality.
19 The milk quota coefficient estimates are significant for the cases of Friesland, Gelderland and Noord-Holland. It is striking that it is a positive effect for Friesland and Noord-Holland and such a strongly negative effect for Gelderland. For Gelderland this would mean that the farm size inequality decreases when the quota increases, which is contrary of the what the literature
suggests for the milk quota effect in general. There are no obvious reasons why Gelderland should be an exception to this mechanism.
Table 5
Province
Variables Drenthe Flevoland Friesland Gelderland Groningen Limburg
t -0.00193 0.00247 -0.00557 -0.00652 -0.00410 -0.00700 0.00241 0.00302 0.00316* 0.00390* 0.10750 0.00691 AvEcVal -0.00140 -0.00068 -0.00047 0.00135 -0.00073 0.00215 0.00061** 0.00031** 0.00065 0.00143 0.00195 0.00304 ShAgrArea -2.30759 -1.48465 -0.73438 -1.09678 -0.43343 1.28144 0.22184*** 0.51641*** 0.35221** 0.44834** 1.53554 1.37231 AvHouVal 0.00001 0.00015 0.00020 -0.00121 0.00011 0.00033 0.00010 0.00016 0.00012* 0.00016 0.00028 0.00025 lnquota 0.20150 0.17097 0.68583 -1.04801 0.62575 -0.30693 0.11673* 0.23608 0.21318*** 0.47548** 0.47574 1.38800 _cons 1.80975 1.03607 0.79765 0.96428 0.61986 -0.49630 0.17200*** 0.23551*** 0.23345*** 0.3856** 0.71205 0.93674 R2 within 0.97150 0.75600 0.70270 0.89850 0.65570 0.66550
Source: Own estimation. *,**,*** implies significance level at 10, 5, 1 per cent respectively
Province
Variables Noord-Brabant Noord-Holland Overijssel Utrecht Zeeland Zuid-Holland
t -0.01172 -0.00945 -0.12704 0.00037 -0.01519 -0.00048 0.00710* 0.00652 0.00519** 0.00650 0.01222 0.00586 AvEcVal 0.00160 -0.00163 0.00175 -0.00256 -0.00011 -0.00605 0.00385 0.00267 0.00166 0.00298 0.00306 0.00256** ShAgrArea -0.40524 -1.67403 -0.98288 -0.25183 -0.71869 -0.96425 0.76806 0.97462* 0.61595 0.93747 1.76513 1.09103 AvHouVal 0.00025 0.00037 0.00033 0.00019 0.00049 0.00037 0.00026 0.00022* 0.00020* 0.00023 0.00036 0.00027 lnquota -0.27129 0.74086 -0.48327 0.49466 0.35380 0.57828 1.60403 0.3927* 0.53823 0.30902 0.99413 0.37654 _cons 0.32978 1.06695 0.56621 0.73052 0.83621 1.44302 0.70851 0.46069** 0.28697** 0.70374 1.04389 0.70759** R2 within 0.52490 0.52850 0.71160 0.41630 0.56320 0.60810
20 4.3 Discussion of results
It is difficult, if not impossible to draw any solid conclusions about reasons for the dispersion in farm size inequality levels across the Dutch provinces from this model. The
estimates from the model are hardly significant and where variables do appear significant seems rather arbitrary. A Hausman test has been performed to confirm the suitability of the fixed effects model for this dataset and was positive about the model. The modified Wald test identified the presence of heteroscedasticity but heteroscedasticity and serial correlation are controlled for by the robust version of the model. Nonetheless, the model has certain shortcomings.
Firstly, in a typical fixed effects model the number of subjects, in this case provinces, substantially exceeds the number of time periods. For our dataset this is not the case as the number of provinces is twelve and the number of periods thirteen years. This might have created some estimation errors. Moreover, in fixed effects estimations one made a trade-off between bias and efficiency. FE methods are less vulnerable to omitted variable bias but usually yield higher standard errors. This protection from omitted variable bias is only for variables that do not change over time. Failing to include relevant variables that do change over time will cause coefficients to be biased. It is certain that there are some variables missing or poorly measured in this study and the variables that were included often were only a proxy, so the model will suffer from some degree of omitted variable bias.
21
5. CONCLUSION
According to the literature discussed in this paper, the milk quota abolition of April 2015 will bring about less market rigidity and stimulate growth of efficient farms. Moreover, the current trend in farming structures of the shifting weight towards less but bigger farms is expected to continue after the abolition because of unlimited production opportunities. In this paper, the farm size inequality of dairy farms in all provinces of the Netherlands was estimated and a regression was run to explore drivers behind changes over time and across regions. This adds to the literature by including a recent dataset and focussing on the Dutch provinces with as instruments the Gini coefficient and a fixed effects model.
For the milk quota effect on the Netherlands overall, a positive effect on the farm size inequality was found. The time trend variable proved to be an important driver and had a negative coefficient. The time trend included all time-variant explanatory factors like
technological advancement and overall economic conditions. Additional research could try to specify this effect more thoroughly.
Unfortunately, not many significant results were to be found on the province level and so it is hard to analyse these and prove some effect or explanation. It is evident that farm size inequality varies among provinces and over time. Further research would be required to find what are the correct explanatory variables behind these developments. Moreover, the role of the quota abolition in this process is yet to prevail.
Further research could possibly gather more information on a farm-specific level to undertake a much more comprehensive study. Moreover, the coming years will be interesting to see how the abolition of the milk quota will actually influence the dairy industry and research can be done by comparing the period before and after the abolition.
22
6. BIBLIOGRAPHY
Bergevoet, R.H.M., Ondersteijn, C.J.M., Saatkamp, H.W., van Woerkum, C.M.J., Huirne, R.B.M., (2004) Entrepreneurial behaviour of Dutch dairy farmers under a milk quota system: goals, objectives and attitudes, Agricultural systems vol.80, 1-20
Boere, E., Peerlings, J., Reinhard, S., Heijman, W., (2014) The dynamics of dairy land use change with respect to the milk quota regime, paper prepared for the EAAE 2014 congress.
Bouamra-Mechemache, Z., Jongeneel, R., Réquillart, V.(2008) Impact of a gradual increase in milk quotas on the EU dairy sector. European Review of Agricultural Economics vol.35 Dijksma, S.A.M., (2015) Aanbieding AMvB grondgebonden groei melkveehouderij, Ministerie
van Economische Zaken, Den Haag
European Commission, (2010). The CAP Towards 2020: Meeting the Food, Natural Resources and Territorial Challenges of the Future. COM(2010) 672 Final, Brussels, Belgium, 18 November.
Frahan, de B.H., Baudry, A., Blander, de R., Polome, P., Howitt, R. (2011). Dairy farms without quotas in Belgium: estimation and simulation with a flexible cost function. European Review of Agricultural Economics 38: 469-495.
Gastwirth, J.L., (1972), The Estimation of the Lorenz Curve and Gini Index, The Review of Economics and Statistics vol.54(3), 306-316
Huettel, S. and Jongeneel, R. (2010) How has the EU milk quota affected patterns of herd-size change? European Review of Agricultural Economics vol.38
Jongeneel, R., Tonini, A., (2008), Dairy quota and farm structural change: A case study on the Netherlands, paper prepared for the 107th EAAE seminar.
Kempen, M., Witzke, P., Dominguez, I.P., Jansson, T., Sckokai, P. (2011) Economic and environmental impact of milk quota reform in Europe. Journal of Policy Modelling 33. Lips M and Rieder P (2005) Abolition of raw milk quotas in the European Union: a CGE
analysis at the member country level. Journal of Agriculture Economics vol.56, 1-17 Mishra, A., El-Osta, H. and Gillespie, J. (2009). Effect of agricultural policy on regional income
23 Corron, N., He, X.Z. and Westerhoff, F. (2007) Butter mountains, milk lakes and optimal price
limiters, Applied Economics Letters, 14:15, 1131-1136
Ooms, D.L., Peerlings, J.H.M., (2005). Effects of EU dairy policy reform for Dutch dairy farming: a primal approach using GMM estimation, European Review of Agricultural Economics vol. 32(4) pp. 517-537.
Piet, L., Latruffe, L., Le Mouël, C., Desjeux, Y., 2012. How do agricultural policies influence farm size inequality? the example of France. European Review of Agricultural
Economics 39 (1), 5-28.
Samson, G., Gardebroek, G., Jongeneel, R. (2012) The cost function structure of Dutch dairy farms: Effects of quota abolition and price volatility. Paper prepared for the 126th EAAE seminar
Stock, J.H., Watson, M.M., (2012) Introduction to Econometrics (3rd ed., Global ed.). London: Pearson Education. 389-413
