Farmland valuation
The analysis of farmland value drivers in Poland and Romania
By
Marnix M. van Deursen
1DD MSc. International Financial Management
Supervisor (IFM): Dr. W. Westerman
Second supervisor (IFM): Dr. ing. N. Brunia
Supervisor Rabo Farm: Alexander Muhring
Date: 09-01-2015
Word count: 17,951
Abstract: This research is conducted during a five month internship at the Dutch fund and asset manager Rabo Farm and analyses the value drivers of farmland in Poland and Romania. Literature identifies the relevancy of size, compactness, geology, meteorology, credit availability, urban pressure, neighbouring farms and subsidies in relation to farmland values. The empirical analysis is based on cross-sectional data from 2013 and consists of 1,136 land transactions in Poland. General multiple regression analysis shows the significant impact of land quality, size, urban pressure and farm density on farmland values. Interpretation of the results provides more understanding of the exact value appreciation of farmland related to the tested value drivers. Qualitative research focused on Romania shows that land quality, urban pressure and farm density have less impact and size has more impact compared to Poland. The size, compactness and land quality can actively be managed by investors such as Rabo Farm by acquiring neighbouring land, consolidation and land quality improvements.
Keywords: Land Value, Farmland value, Agricultural Land, Farmland Jel classifications: Q15, Q51
Page | 1 TABLE OF CONTENTS
1. INTRODUCTION ... 2
1.1EXPLAINING THE GLOBAL TREND OF FARMLAND INVESTMENTS ... 2
1.2COMPANY PROFILE RABO FARM ... 3
1.3.COUNTRY PERSPECTIVES ... 4 1.4VALUATION METHODS ... 5 1.5PROBLEM STATEMENT ... 6 1.6RESEARCH OBJECTIVES ... 7 2. THEORETICAL BACKGROUND ... 9 2.1HISTORY ... 9 2.2FARMLAND VALUATION ... 13 2.3AGRICULTURAL PERSPECTIVE ... 13 2.4NON-AGRICULTURAL PERSPECTIVE ... 15
2.5VALUATION ACCORDING TO LEASE PRICES ... 20
3. METHODOLOGY...21 3.1MODEL SELECTION ... 21 3.2DATA ... 22 3.3ESTIMATION PROCEDURE ... 26 4. ESTIMATION RESULTS ...27 4.1INTERPRETATION ... 29
5. DISCUSSION AND COMPARISON ...32
5.1AGRICULTURE IN ROMANIA ... 33
6. CONCLUSION ...37
6.1IDENTIFIED VALUE DRIVERS ... 38
6.2IMPLICATIONS ... 39
6.3LIMITATIONS AND FURTHER RESEARCH ... 39
BIBLIOGRAPHY ... 40
Page | 2
1. Introduction
This introduction explains the global trend of farmland investments by evaluating the possible motives of the investors. Next it illustrates the company profile of Rabo Farm and explains their role as an investor and research partner. The third part elaborates on Rabo Farm’s activities in Poland and Romania by introducing the dynamics in farmland investments in both countries briefly. The final part describes the problem statement, research question and objectives. The research relevancy is further explained through an international, management and finance perspective.
1.1 Explaining the global trend of farmland investments
Over the last few years, various investors ranging from financial institutions to high net worth individuals have shown a growing interest in farmland investments. Research and study reports show the existence of five motives that explain the global trend of investing in this traditionally not very exiting asset class (Stookey and Lapérouse, 2009; Iowa State University, 2014). The first motive is based on long term fundamental attractiveness caused by a growing world population, changing nutrition preferences and a rising welfare in developing countries. The Food and Agriculture Organization (FAO) of the United Nations expects an incremental land demand of at least 75 million hectares in 2015. Taken into account the 24 million hectares brought into cultivation in the period 1995-2005 it can be expected that the existing deficit will increase rapidly (Alexandratos and Bruinsma, 2012). The second motive relates to the attractive historic returns that provide a reliable track record for potential investors. Data analysis shows that farmland returns often exceed the returns of equities (Bloomberg, 2014; Reuters, 2014). 2 The third motive relates to the
mix of short-term income with capital appreciation.
Despite the long-term orientation of farmland investments it also provides a stable short-term income which is attractive for organizations that require annual cash payments (e.g. pension funds). The fourth motive refers to uncorrelated returns with
the equity market that make farmland an attractive
2 Appendix 1 presents a graphical comparison between farmland and equity returns.
0% 2% 4% 6% 8% 10% 1997 2001 2005 2009 2013 CPI (EU) Cereals
Page | 3 alternative or addition to an existing portfolio. For example, farmland investment can be used for efficient risk diversification purposes (Lins et al., 1992).3 The last motive relates to an effective inflation
hedge because commodity prices correlate positively with the consumer pricing index. Eurostat data
is used to assess the correlation between CPI (European Union) and cereal prices for the period 1997-2013 (figure 1). The correlation coefficient is 0.75 which means that in years of high inflation, farmers receive more income and are able to pay more rent. This automatically increases the direct return of the investor and the asset becomes more valuable.
1.2 Company profile Rabo Farm
This research is conducted in cooperation with the Dutch fund and asset manager Rabo Farm. This fund and asset manager, which is a subsidiary of the Rabobank Group, serves institutional investors that are willing to invest in a niche asset class that can be defined as arable land and affiliated infrastructure such as farm bases and storage capacity. These investments are made in an asset class located all the way upstream in the food and agriculture value chain. Rabo Farm does not only perform arable land acquisitions but also manages these investments actively so farmers can maximise their production potential. The rationale here is to improve soil quality through improved pH levels, water management (both drainage and irrigation) and basic infrastructure allowing farmers to dry and store their products to be less dependent on commodity prices during the harvest season. The ideal situation should provide benefits for all stakeholders including investors, farmers and local communities. By actively managing and investing into the arable land Rabo Farm believes it can minimise the so called reproduction yield gap. The yield gap can be defined as the difference between the current production capacity per hectare on a certain piece of land and the maximum production level if all conditions on which human kind has influence are optimal (ceteris paribus weather and commodity prices). Rabo Farm’s strategy is focused on the initiation, structuring and management of funds that invest in farms (e.g. arable land, infrastructure, buildings, drainage, irrigation, storage, lime and other fertilizers). The tenor of the fund has a long-term investment horizon set at 10 to 15 years or more. Europe is the primary area of focus and currently most of the capital is invested in Poland and Romania. Rabo Farm offers institutional investors the possibility to invest into arable land as an asset class. The business model is based on long-term fundamental attractiveness due to increasing world population, welfare and food consumption. More of the attractiveness is explained by the decrease of arable hectares through urbanisation, dissertation and rising water levels. The company foresees that demand fundamentals will remain strong, while supply is constrained because of limited resources, thus making
Page | 4 investments in this asset class interesting. In addition, Rabo Farm aims to respond to the global food security challenge, by actively managing its assets with the objective to minimise the yield gap.
Organizational activities are structured in three different disciplines. The first is concerned with fundraising and the relationships with current and potential institutional investors. Rabo Farm currently manages one fund with a size of EUR 315 million (European fund I) which is funded by three institutional investors (i.e. APG, ABP and TIAA-CREF) each participating equally with EUR 100 million together with a 5% commitment participation of Rabobank Group. The first fund is almost fully invested and there are several initiatives in process for the second fund (European fund II). The second organizational discipline focuses on acquisitions and portfolio management. This includes the sourcing, structuring, negotiating and acquisition of arable land from sellers and also the active management afterwards whereby the focus lays on minimising the yield gap. The third discipline is focused on financial control and legal support.
1.3. Country perspectives
This study focuses on agriculture and farmland prices in both Poland and Romania. The main part of European Fund I is invested in these countries and Rabo Farm is expected to maintain its focus on these areas. This part provides general information about the local agricultural sectors of both countries.
Poland
Poland is one of the largest countries in Europe with a surface of 312,650 km² and an estimated total population of about 38 million. The local currency is the Polish Zloty often referred to as PLN. The country is characterized as extremely rural and only 19% of the population lives in urban areas (GUS, 2014). The agricultural landscape shows a great variety in terms of demography, structure and economy. According to the Central Statistical Office, the country accounts for approximately 2.28 million agricultural holdings of which the main part is considered as small (i.e. less than 20 hectares). Furthermore, the total area of agricultural land is currently around 16 million hectares of which 73% arable, 16% meadows, 5% pastures, 2% orchards and 4% of another classification. Currently Poland is of interest to many foreign agricultural investors and land prices show huge increases in recent years. Table 1 provides an overview of agricultural land prices in private turnover (GUS, 2014).
Table 1. Average farmland prices in PLN and average increase in % in Poland for the years 2004-2013. Data retrieved from: GUS, 2014
Year 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013
Average price
per hectare 6.634 PLN 8.244 PLN 9.290 PLN 12.134 PLN 15.388 PLN 17.042 PLN 18.037 PLN 20.004 PLN 25.442 PLN 26.339 PLN
Page | 5 The data presents the price increase from the moment Poland became a member of the EU in 2004. Currently Rabo Farm owns around 23,000 hectares of arable land in Poland and this amount is increasing rapidly. The history of Poland has a significant impact on the current agricultural conditions which is further explained in the literature.
Romania
Romania has an estimated population of around 20 million people and a total surface of 238,931 km². The local currency is the Romanian New Leu referred to as RON. Information provided by the European Commission (2014) states that the agricultural capacity is around 14.8 million hectares of which approximately 10% is used as arable land. The agricultural industry adds around 6% to the total GDP (this was around 12% in 2004) and the agricultural sector employs around three million people. This implies almost 30% of the total Romanian workforce (i.e. in contrast to around 4.5% in other European countries). Moreover, 64% of the Romanian farm production is used for own consumption. Similar to Poland it can therefore be stated that Romania is largely depended on the agricultural sector. According to the European Commission, Romania accounts for 32% of the total number of agricultural holdings in Europe (Poland around 12%) and the rural areas cover more than 87% of the total surface. Furthermore, Romania is characterized by erosion and fragmentation of soil, a severe lack of technology and innovation and many property related lawsuits. On the other hand, an increasing number of companies are attracted by the economic development of the country. Rabo Farm currently owns around 29,000 hectares of farmland mainly in the Eastern and Western part of the country. This is explained by the presence of infrastructure used for the transportation of crops. The city Constanta is located in the East and has access to the Black Sea. The Western part of the country is connected to the rest of Europe by roads. The discussion part at the end of this research will elaborate more on farmland valuation in Romania and compares the country’s agricultural characteristics with Poland.
1.4 Valuation methods
The value determination of the land is particularly important to investors assessing their investment opportunities. Basically there are two distinctive techniques used for the valuation of agricultural land. The first is focused on expected income and expenses of farming the land. The second method is based on transaction prices of comparable assets.
Discounted cash flows
The valuation technique determines the land price based on the production potential of the land (expected production level at assumed prices) and the costs related to that production level (farm inputs, cost of capital, transport costs, etc.). This method is better known as the discounted cash flow
Page | 6 applied in rural areas where there is either no functional market or any other comparable sales data. This method is used in practice, however, the variables that serve as input are hard to determine and are often inaccurately based on gut feelings (e.g. discount rate, harvest prospects, timeframe, etc.). This explains why appraisers and investors often tend to base their valuation on comparison techniques.
Direct comparison
The second method is commonly used for land valuation in developed countries such as the Western part of Europe. This method determines an economic value in comparison with similar transactions of similar plots under similar conditions that have been sold in the recent past. This valuation technique largely depends on comparable sales information, whereby the recording of reliable data in an accessible database is pivotal. The limitations of the direct comparison approach are evident. The first limitation relates to the limited number of transactions in less developed countries. This problem becomes clear during the analysis of appraisal reports used in Romania. These reports compare the appraised subject with asking prices of comparable farms. This often leads to misleading appraisal prices, as what sellers ask is not necessarily what they receive. The second issue relates to the difficulty of obtaining relevant sales data. The documentation in Poland is still limited although very advanced compared to Romania where the state department of agriculture started to collect sales data as of 2014, while the accessibility to this data is very restricted. Another problem relates to the reliability of sales information since prices can be distorted because the transaction also include standing crop, machinery and existing stocks of produce and farm inputs. The lack of a clear breakdown of the price contributes to lesser transparency. More reasons for misleading price information can be caused by creative payment mechanisms to avoid taxes. Furthermore, buyers can have different motives for paying more than the outcome of the comparison method. This occurs when farmers pay a premium for land neighbouring their own land in contrast to land prices paid for land located on the other side of the village. Higher prices are also paid when buyers have special development plans in terms of commercial usage (residential permits, wind turbine parks, etc.). To conclude, issues with regard to land valuation often relate to the availability of different valuation techniques and the several problems mentioned.
1.5 Problem statement
Page | 7 issue is that value drivers in Poland and Romania are still not clearly identified. Therefore their existence and impact is often unclear
The second problem relates to the selection of those farm bases that generate the highest return on investment. There are many different farms for sale in both Romania and Poland and from that the most profitable options should be selected. The acquisitions need to be selected carefully since the main part of the investor’s return is generated through the value appreciation of land. The identification of value drivers and the weight thereof would provide more insight in price development and investment management decision making.
The third problem is more specific and relates to Rabo Farm’s activities in West Poland. The company is unsure about which part of the Western provinces has the best profit potential. Farmland prices tend to differ significantly within this region and prices in the North and North-West are generally lower than in the West and South-West. Decisions need to be made regarding the area in which the fund will invest, and therefore more insight is required about farmland price variability within the country.
The fourth problem relates to questions asked by potential investors in Rabo Farm’s funds. These institutional investors would like to understand the forecasting of future value appreciation of land in order to make appropriate investments decisions. This is important because pension funds search for investment opportunities that combine a solid return with a low risk profile.
To solve these problems, this study will first focus on farmland value determinants that are described by academic literature. Additional information about value drivers is provided by experts that actively participate in the agricultural business in Poland or Romania. This information is used for a cross-country comparison to identify possible differences. The empirical part is focused on Poland and combines sales data with academic findings and information provided by several experts.
1.6 Research objectives
Page | 8 also relates to one of the problems addressed by Rabo Farm as the company needs to determine which provinces are the most profitable. Eventually, the ultimate goal of this research is to provide clear and straightforward advice that is easy to interpret. This allows Rabo Farm to select the best investment opportunities based on the presence of a certain set of characteristics without depending on comparable sales data. Clarification of the problems and objectives finally results in the following research question:
What are the most important value drivers of farmland in Poland and Romania and how can these drivers be influenced to reach the full potential of investments?
Relevancy
The relevancy of this research is illustrated through an international, financial and management perspective. The international role of this research is explained by the cross border activities of Rabo Farm as a research partner. The study attempts to combine academic knowledge with field research in Poland and Romania. The empirical analysis is based on the Polish land market and aims to test the identified value drivers. Farmland valuation in Romania is evaluated according to experiences of farmers, appraisers and investors. Eventually, in the discussion part, a comparison is made between both countries to address the relevant differences. This careful approach is favourable since most prior research about farmland valuation is based on the US land market. The generalizability of these findings to the European, and especially the Polish and Romanian, agricultural land market is somewhat under suspect. All markets possess different characteristics in terms of economic development, demographics and regulation that all could have a considerable influence on farmland prices. For example, land in Poland can only be acquired by local farmers and investors which indicate a unique set of supply and demand. It can therefore be stated that known value drivers need to be reviewed first before applied in other countries.
The financial perspective relates to implications of this research with respect to land investments. Also the hedonic pricing model is evaluated in relation to real estate and farmland valuation. Moreover, the research is conducted in cooperation with a financial institution and the implications concern the future income of business activities. Furthermore, the agricultural perspective of this research is eventually translated and capitalized into hectare value variability. These findings serve as input in calculations that provide insight in the profitability of investments (such as IRR calculations).
Page | 9 results. This also included the collection of all relevant characteristics in a certain area that could influence farmland prices. The third phase develops activities that maximally utilize the characteristics of the land and region in order to maximize the investor’s IRR.
The next part of this research provides the theoretical background and addresses all relevant elements that concern the valuation of agricultural land. The methodology explains the model selection, data and estimation procedure with regard to farmland transactions in Poland. The results show the outcome of the model estimation combined with clear guidance regarding the interpretation. The discussion part is important as it explains agriculture in Romania and compares the agricultural conditions with the characteristics of Poland. This analysis is of great value as it contributes to the limited research done in this country. The conclusion combines both the empirical findings regarding Poland and the qualitative findings from the discussion part concerning Romania. Eventually some arguments are raised related to the limitations of this study and ideas for future research.
2. Theoretical background
In this study, both the synonyms ‘agricultural land’ and ‘farmland’ are used. Both terms can be defined as: “land including arable land, land under permanent crops and land under permanent meadows and pastures” (OECD, 1997). This study is primarily focused on arable land used for cereal farming. Rabo Farm and most other investors, which invest in land in order to lease it out for agricultural production, are especially interested in land producing annual crops (crops that are replanted after each harvest). Evaluating the efficiency of the land market in Poland is necessary to decide whether transaction prices are a fair representation of market values. This condition is crucial for the true identification of value drivers (Folland and Hough, 1991). There are several historical events that deserve special attention as these have shaped Poland in terms of agriculture. An important distinction can be made between the land market up to 1989 and after 1989.
2.1 History
Page | 10 areas were traditionally divided among the
Prussian aristocracy, and characterised as larger areas that were centrally farmed by a large number of peasants that were loyal to the local aristocrat. The communists, after killing many aristocrats, decided to keep the level of efficiency driven by the economies of scale. They turned them into state-owned farms, managed by well-educated farm managers and peasants originally living in the confiscated parts in the East, which were moved to perform the hard labour. After the fall of communism, the Agency of State Treasury Fund for Land Ownership was established in order to restructure the state-owned land in the early 1990s. This process implies the selling and leasing
of land to joint ventures and individual farmers. This contributes to the great variety within Poland in terms of ownership and size of agricultural holdings.
The agricultural census 2010 shows that the total number of farms is respectively 3,424 in the public sector and 1,445,254 in the private sector (GUS, 2014). Banski (2011) states that private farm holdings are extremely small compared to public farms (i.e. about 70% of private farms can be found in the one to eight hectare category, compared to an average size of approximately 600 hectares in the public sector). Moreover, the small farms can mainly be found in the South-East of the country and the large state-owned farms are located in the West (formerly Germany). These large legal entities are characterized by large blocks that are farmed together historically, although neglected over many decades. This historic perspective illustrates the differences within Poland and explains Rabo Farm’s concentration on the Northern and Western part (bolded and underlined in image 1). The scattered land in the South-Eastern part is simply not appropriate for intensive large-scale farming (too many small plots with different ownership). As a consequence, this research is primarily focused on the provinces in the Western part since all significant land transactions (i.e. five hectares and more) originate from the Western provinces. The differences within the country are further explained by Jerzy Ptaszynski (Research and Analysis Director, AMRON), an expert in the field of agriculture and farmland valuation.
Province
Page | 11 Mr Ptaszynski states that Poland has significant differences in farmland prices within the country. He further argues that the Western part, including the North-West and South-West, are attractive rural areas which are normally characterized by higher hectare prices. This is mainly explained by the presence of small fragmented holdings in the East compared to large and integral holdings in the West which are more suitable for intensive cultivation. This seems therefore even more important than soil quality in explaining farmland prices. This is illustrated by realizing that the best land quality, located in the province Lubelskie, is also one of the cheapest. This study supports the argumentation of Mr Ptaszynski by addressing farmland price differences based on data retrieved from the Central Statistical Office (see Table 2). Most land transactions took place in the Northern, Western and Southern regions. This research will therefore especially focus on the eight bolded provinces (i.e. Lubuskie, Dolnośląskie, Opolskie, Kujawsko-Pomorskie, Wielkopolskie, Zachodniopomorskie, Pomorskie and Warmińsko-Mazurskie) presented in table 2. As earlier explained, these provinces have more appropriate conditions for intensive farming and are therefore of more interest to Rabo Farm.
Restrictions
Certain regulatory aspects deserve special attention in explaining the current agricultural land market in Poland. One important aspect is the restriction for foreigners to acquire land, directly or through foreign or local companies. Dadak (2004) argues that these rules are in force since 1920 and imply that foreigners are only allowed to buy small plots, however since the entering of Poland into the EU all land purchase by foreigners (also through legal entities) is prohibited. Exception to the rule needs to be obtained at the Ministry of Interior; however this is rarely given for political reasons. This restriction is expected to have significant consequences for the prices of land, as there is no real free market. These regulations prohibit acquisitions of land in terms of asset deals; however the law does not prohibit the purchase of shares of a company that already owns land. This acquisition method is defined as a share deal and identification of these transactional structures is important because it
Geographical location Avg. Price (ha./PLN) Zachodniopomorskie North-West 20.173 Pomorskie North-West 27.891 Warmińsko-Mazurskie North 21.871 Podlaskie North-East 26.078 Lubelskie East 18.140 Mazowieckie East-Central 28.301 Łodzkie Central 25.811 Podkarpackie South-East 16.459 Świętokrzyskie South-East 19.382 Małopolskie South-East 22.459 Śląskie South-East 25.157 Lubuskie South-West 17.312 Dolnośląskie South-West 27.802 Opolskie South-West 31.283 Kujawsko-Pomorskie West-Central 36.377 Wielkopolskie West-Central 36.838
Page | 12 explains the way foreign investors are able to invest in Polish farmland4. The complexity of a share deal construction prevents small individual farmers and investors from purchasing land. Large investors such as Rabo Farm use this legally perfected structure frequently5. Critics argue that these restrictions are not in line with the free movement of labour and capital within the European Union’s boundaries. As Poland entered the EU, Poland is limited in its possibilities to preserve the existing restrictions and the liberalization will start, after an exemption period has ended, as of 2016. Dadak (2004) further states that price calculations concerning farmland are mainly based on the assumption that the market remains closed. It is therefore conceivable that the prices of land will strongly increase when foreign investments are no longer restricted. Increased demand can mainly be expected from foreign farmers and smaller investors that do not have the knowledge related to complex purchase structures such as share deals. This illustrates the possible consequences of the liberalization with respect to small land transactions. If the same applies for large land acquisitions remains to be seen after 2016.
Agencja Nieruchomości Rolnych
The organization responsible for the management of public land is the Agencja Nieruchomości Rolnych (ANR) founded in 1992. Their task is to sell off or lease out all the state owned land and therefore their decisions have a significant impact on the agricultural land market in Poland (Hurrelmann, 2008). The ANR is a governmental organisation with strong roots in the Polish socialist parties. At this moment, the ANR still owns 2.68 million hectares, although that they have already sold 2.34 million hectares of their initial holdings. Most of the holdings are located in the Western part of Poland. This means that the ANR will continue to flood the land market with their hectares over the next period. As this EU regulated state divestment scheme has impact on overall market prices, the ANR acts cautious and feeds the market slowly by drip, often through auction processes. Foreigners are discriminated from this divestment program, not only because foreigners and foreign owned companies are not allowed to acquire land, but also by a number of other laws. Some of the most important restrictions and privileges are listed below.
Auction restrictions: This means that when a plot of land is sold, only officially registered farmers that live in that municipality or one of the bordering have access to the auction. Pre-emptive right: The ANR always has the right to step into the buyers obligations for each
and every land transaction that occurs in Poland, whether sold or not sold by the ANR.
4 An asset deal implies the purchase of land alone. A share deal refers to the purchase of shares of a legal entity
owning land.
5 Appendix 2 shows the deal structure (i.e. share deal) that is frequently used by Rabo Farm for the acquisition
Page | 13 Re-purchase right: For all land that the ANR sells, the ANR has a repurchase right for the price
for which they sold the land, for a period of five years.
2.2 Farmland valuation
The academic literature with regard to farmland valuation is mainly based on American research and a long history of data collection. A comprehensive documentation enables researchers to assess a broad set of factors that possibly influence farmland prices. The difference with European research is that the latter are mainly based on a shorter period and a smaller research perspective. This means that American studies are mainly focused on macro level and macroeconomic factors in contrast to European studies that often use a micro level focus and assess more fundamental factors (i.e. agricultural factors such as irrigation and soil quality). This study uses both research perspectives as this provides a more comprehensive illustration of all elements that concern the valuation process.
Transaction prices of arable land
Folland and Hough (1991) state that the potential future yields from agricultural and non-agricultural operations are capitalized into the sales price of agricultural land. Furthermore, it is considered that the best and most profitable use of the land serves as basis of the capitalization of expected future returns (Phipps, 1984). For farms this is usually agricultural income and subsidies related to the land. If non-agricultural operations, such as the potential development of residents have higher expected returns than agricultural activities, then the price of land would be higher per hectare. In such a case, researchers should realize that they need to incorporate measures related to non-agricultural usage (Shi et al., 1997). For this reason, previous studies often make a distinction between agricultural and non-agricultural factors when potential value drivers are identified.
2.3 Agricultural perspective
This research will first explain all factors that concern the valuation of farmland through an agricultural perspective in Poland. A brief description of all important land characteristics is provided including differences in size, soil quality, geology and meteorology. This part further mentions some well-known instruments that improve the agricultural condition of the land such as irrigation and drainage.
Parcel size, compactness and consolidation
Page | 14 appraisers and defined as the total surface of a parcel or group of adjoining parcels. The appraisers state that the size of a parcel has a positive influence on the price. This is confirmed by studies of Folland and Hough (1991) and Xu, Mittelhammer, and Barkley (1993). In general, larger parcels are more appropriate for intensive farming and increase the efficiency and productivity per hectare. The most important reason is that larger parcels are often cheaper to farm. If farmers have one big plot on which they can work, this saves farm input, fuel and time. Comparable advantages are caused by the compactness of several parcels which explains why many farmers attempt to consolidate their own land and search for expansion and exchange opportunities. This consolidation process is expected to be both time-consuming and costly (e.g. transaction costs related to due diligence, hiring multiple brokers and salary expenses). Consolidated land eventually results in many cost savings and increases the revenue per hectare which is in the end reflected by a higher price. Sklenicka et al. (2013) further state that farmland analysis is only reliable if the number of parcels and the compactness of the land are considered. Image 2 shows the situation before and after consolidation.
Quality, geology and meteorology
Many authors stress the importance of soil quality and soil fertility. Researchers often translate soil quality into different agricultural characteristics such as soil class, soil productivity, nutrient level, pH values, tillable area, natural soil fertility and fertility class (see e.g. Oltmans, Chicoine, and Scott, 1988; Barnard et al., 1997; Huang et al., 2006). The authors unanimously state that the quality of land has a significant positive influence on the sales price. The reasoning behind this relation is that a higher quality will produce more direct yields from the land when all other conditions are held constant. By investing in land quality, one can actively manage the value of the land. Two examples that have a huge impact on land quality are irrigation and drainage (James Little, Agricultural manager Rabo Farm). In this respect, there are very interesting differences between dry countries and countries with sufficient precipitation. For example, most important rural areas in Poland have sufficient precipitation
Page | 15 for growing crops and therefore irrigation is not necessary. Many rural areas in Romania have little rainfall and therefore irrigation systems are of great value. The opposite applies to drainage being of more value in Poland (and other countries with severe precipitation) and not in dry countries such as Romania. The presence of the appropriate system (irrigations or drainage) has a huge impact on land values caused by increased yields or the reduced risk of losing crops due to excessive rainfall. This also indicates the possible influence of meteorological factors on crops yields and farmland prices.
The climate of Poland is mainly temperate throughout the country. Some parts of the country have little precipitation due to their geographical location pertaining to the Baltic Sea. On the one hand, differences occur as a consequence of varying conditions between North and South. These differences are due to the presence of the Sudeten Mountains in the South that cause more and better dispersed rainfall over the year. This results in more appropriate temperatures for growing crops. It is interesting to note that the conditions in the South enable crops to grow one additional month a year. This also allows the production of corn that requires a relatively long growth period. Towards the East, the climate becomes more continental with warm summers and cold winters. The North-East is characterized by an average winter temperature of approximately -6°C. Although rainfall occurs the entire year, the summers have more precipitation than the winters. Moreover, the amount of sun hours every year also explains why the Southern provinces have better conditions for growing crops. Moreover, Poland is characterized by a special phenomenon called Mosaic. This means that due to historical geologically movements (i.e. ice age) the land structure can differ strongly within small areas.
Illustrating the differences in soil, land structure and climate perfectly explains why the majority of agricultural activities take place in the Western part of Poland. Surprisingly, there are not many authors explicitly mentioning meteorological phenomena. On the other hand, capitalizing climatological factors into direct yields would be very difficult since crops require different sets of conditions. It is therefore not surprising that direct yields are often used instead of soil quality and meteorological characteristics (Folland and Hough, 1991; Drescher et al., 2001). Data related to direct yields is often documented and therefore available and easier accessible than soil quality measures and meteorological maps (applies for the US). Other studies prefer measures related to land quality and growth periods as this also explains direct yields (e.g. Huang et al., 2006).
2.4 Non-agricultural perspective
non-Page | 16 agricultural aspects that need to be considered relate to credit availability, subsidies, regulation, infrastructure, the scarcity of arable land and several macroeconomic factors.
Credit availability
Periods involving low real interest rates are often characterized by an increasing attractiveness of real assets, such as farmland, compared to monetary assets (Reinsel and Reinsel, 1979; Devadoss and Manchu 2007). These studies found a positive relationship between farmland prices and monetary assets in periods with low interest rates. On the other hand, a negative relationship was addressed in periods characterized by high interest rates. Devadoss and Manchu (2007) also address the negative relationship between interest rates and agricultural land values. Their macro-economic model confirms that during high interest rates the returns of financial assets are higher compared to agricultural land. This decreases the value of farmland in such a period and demonstrates the negative relationship on both sides. Shalit and Schmitz (1982) also found that farmland prices tend to increase relatively in periods with easier access to funds.
Although there are many providers of financing at different rates and conditions, this study also identifies interest rates related to preferential bank loans which are offered by the state and managed by the ANR. Mr Ptaszyński (Director Research and Market Analysis, AMRON) states that a considerable part responsible for the land price growth in previous years can be assigned to this preferred lending program. An example will illustrate the effect of preferred loans and interest rates on the Polish land market. Until the end of 2013, farmland could be bought from the ANR for a preferred loan whereby only 10% of the asset value had to put up for down payment, while the remainder would be amortised over 15 equal semi-annual instalments, with interest rates of just 2% (whereas commercial banks would not apply advanced rates of above 60%, against 6-8% interest requiring an equity investment of 40%). This offers a significant advantage to local farmers compared to foreign investors. Studies including the effect of interest rates on farmland values always use a broad timeframe. The reason for this could be that land acquisitions are time-consuming and effects of interest rate changes will therefore be delayed visible. It can also be stated that interest rates are similar within the same country and do not explain farmland value variability.
Urban pressure
Page | 17 activities. These urban activities put pressure on the fixed amount of land in a particular region. Prior research shows that the local population growth is an appropriate indicator of urban pressure (Drescher et al., 2001; Livanis et al., 2006; Isgin and Forster, 2006; Sklenicka et al., 2013). This means that high levels of conversion are usually reported in areas with a substantial population growth. It is also interesting to note that urban pressure generally adds more value to the hectare price of land than agricultural factors. To stress this even more clearly, farmland prices in these areas are often totally explained by the demand for residents and commercial activities (Forster, 2006). Goodwin et al. (2005) showed that adding one individual per square mile will increase the land price with USD 1.85. With respect to population growth they found that a 1% increase in growth rate adds USD 64.00 to the hectare price of farmland. Many researchers found that this relationship is much stronger in areas with a large population. It is therefore plausible that absolute growth is more influential than relative growth. The influence of urban pressure can best be illustrated according to a comparison between the Netherlands and Poland. For example, cropland in the Netherlands is currently sold for more than EUR 50,000per hectare in contrast to approximately EUR 5,220 in Warmińsko-Mazurskie (European Commission, 2014). When considering the agricultural conditions in Warmińsko-Mazurskie (which are even more suitable for intensive farming) it perfectly illustrates the presence of other value drivers.
Some studies use a different approach for testing the same phenomena of urban pressure and potential conversion. These researchers define the potential conversion of land as the distance to the nearest municipality or built-up area (e.g. Guiling et al., 2009; Sklenicka et al., 2013). A disadvantage of this approach is the time-consuming task of measuring the exact distance of a particular parcel to a constructed area. For this reason these studies use small samples which are very much focused on micro level. Other researchers have only identified the proximity to a substantial city or capital city (Huang et al., 2006). These authors stress the importance of urban pressure in relation to cities with a substantial population or exposure (e.g. more than 100,000 inhabitants). In the contrary, some studies do not include any factors related to urban pressure (Pyykkönen, 2004; Devadoss and Manchu, 2007). These researchers emphasize on agricultural factors and argue that urban pressure and non-farm driven demand do not have any influence on land prices in their country (e.g. Finland).
The historic attractiveness of rural areas
Page | 18 allows the farmer to divide the fixed costs (equipment and machinery) over a larger number of hectares. The relative gain per hectare increases and enables farmers to pay a premium for land acquisitions. Another important aspect relates to the infrequent purchase opportunity of neighbouring land. Since farmland of neighbouring farms could be sold only once per generation this adds to the pressure to overbid. This indicates that a larger number of farmers cause a higher land price. This occurrence can even be measured more precisely when focusing on the amount of land available and the exact land type, as for instance farmers producing crops are more interested in arable land than pastures. Areas with a high concentration of farmers also provide an indication of the historic attractiveness of certain regions.
Livestock farms, manure and regulation
An additional suggestion is provided by Gerd Boeckenhoff (Chief Investment Officer, Rabo Farm) and relates to the amount of livestock in a particular region, where manure spreading is regulated. Farmers with livestock are limited in the amount of manure they can spread over one hectare of land. For example, in some areas the hectare price is well explained by the vast amount of pigs (e.g. Noord Brabant, Netherlands). Farmers are required to use a certain amount of land for manure spreading and as a consequence the demand for arable land increases. This suggests that a higher concentration of livestock farms increases farmland prices.
Subsidies and government payments
Page | 19 The most relevant topics that concern the determination of farmland prices are listed in Table 3. Both categories (agricultural and non-agricultural) are included and described in accordance with some specific information regarding agriculture in Poland.
Identified topic Description Hypothesized
influence Literature
Sales price Potential future yields are capitalized into the sales price per hectare. Most profitable use serves as basis for capitalization.
n/a Chicoine, 1981; Phipps, 1984; Folland and Hough, 1991
Size and compactness Total area of purchased parcel or group of adjoining parcels. Larger and more compact parcels can increase efficiency and productivity.
Positive Folland and Hough, 1991; Xu, Mittelhammer, and Barkley, 1993; Sklenicka et al., 2013
Geology, meteorology and land
improvements
Direct crop yields are influenced by geology (soil quality and soil fertility) and meteorology (climate, precipitation, sun hours and growth periods)
Positive Oltmans, Chicoine, and Scott, 1988; Folland and Hough, 1991; Barnard et al., 1997; Drescher et al., 2001; Huang et al., 2006
Credit availability Low interest rates increase the interest in real assets compared to monetary assets and vice versa. Preferred interest rates stimulate land acquisitions among Polish farmers.
Negative Reinsel and Reinsel, 1979; Shalit and Schmitz, 1982; Devadoss and Manchu, 2007; Jerzy Ptaszynski, AMRON
Urban pressure Urbanization puts pressure on the fixed amount of land. Potential conversion is capitalized according to population growth, total population and proximity to a city.
Positive Chicoine, 1981; Phipps, 1984; Folland and Hough, 1991; Shi et al., 1997; Drozd and Johnson, 2004; Livanis et al., 2006; Isgin and Forster, 2006
Neighbouring farms/attractiveness of area
Number of farms in a particular area. More potential buyers increase the demand for available agricultural land. Restrictions with respect to manure spreading require additional hectares.
Positive Pyykkönen, 2004; Gerd Boeckenhoff, Rabo Farm
Subsidies and government payments
Subsidies increase the earnings and therefore the willingness to pay more per hectare.
Positive Weersink et al., 1999; Devadoss and Manchu, 2007; Mishra, Moss, and Erickson, 2008; Jerzy Ptaszynski, AMRON Table 3. Topics concerning farmland valuation identified by academic literature, research reports and experts. The table also includes a short description of the topics, the hypothesized influence on farmland values and the source.
Page | 20 value drivers. Farmland value estimation based on this approach can therefore be very convenient for both farmer and investor.
2.5 Valuation according to lease prices
This approach is explained by a real-life example containing a profit and loss statement which shows the impact of lease prices, yields, subsidies and expenses on farmland prices. This happens because investors often base their investment decisions on the future lease payments of farmers. The lease is determined by the investors according to an estimation of income and related expenses. Farmers will require a certain return on investments (ROI) as compensation for the equity they make available and the huge individual risk they take. Farmers also
require a certain amount of liquidity for working capital purposes, as their expenditures occur over the entire year, while the turnover is only realised at the end of the agricultural year. During years with bad weather conditions and poor harvests or adverse prices, the costs might be higher than the profits. It is therefore important for both investors and farmers, that the lease costs charges can be paid from the revenues realised under normal conditions. A default of the farmer or any other issues related to the lease payments will directly harm both parties. Table 4 shows a small part of an investment expose including the relevant income and expenses. The example shows a possible ROI of 6.4% for the farmer under normal conditions.
The lease payments are eventually used by the investor to estimate their own IRR and determine an appropriate purchasing price. This research participated in one of Rabo Farm’s deals to explain the practical use of farmland value determination based on lease payments. The calculations turn out to be very clear and straightforward. The first step is to determine a required IRR based on management costs, experience with previous investments and return requirements agreed with investors (Rabo Farm requires a direct return of roughly 4.5%). The next step is to agree a future lease price based on the characteristics of the potential acquisition. Rabo Farm has a lease agreement with Nutre Group for a number of farms in Romania and the agreed lease price for this particular farm is EUR 170 per
Income and expenses PLN Costs in (%) Crop Sales 2.274.800 EU Subsidies 372.680 Straw sales 2.647.480 Cost of Sales Seed -116.946 5% Chemicals -298.518 12% Fertilisers -468.380 19% -883.844 Overheads Rent -630.420 25% Lime -56.880 2% Contract Charges -658.533 26% Drying and Storage -152.497 6% Agronomy -5.000 0% Management Fee -47.400 2% Crop Insurance -52.800 2% -1.603.530 Total Income 2.647.480 Total Costs -2.487.374 Profit 160.106 ROI 6,4%
Page | 21 hectare. An easy calculation shows that the maximum purchase price for acquiring the land is therefore EUR 3,778 per hectare (EUR 170 / 0.045), irrespective of the outcome of any valuation.
Brief summary
The theoretical background identifies many different topics that influence farmland valuation and agriculture in general. These subjects are reviewed and translated into value drivers applicable to Poland (and eventually compared to Romania in the discussion part). The most important topics are translated in an accurate measure that can be used for empirical testing. These translations mainly depend on the distinguishing characteristics of the land market in Poland and the availability of the information. The following table lists the relevant topics and the required measurement units.
Identified topics Required measurement unit
Sales price Transaction price in local currency per hectare
Size and compactness Size of one parcel in hectares and/or the size of a group adjoining parcels Geology, meteorology and land improvements Index of soil class, direct yields, sun hours, precipitation, growth period,
irrigation, drainage, hydrology and relief Credit availability Interest rates on loans used by farmers
Urban pressure Population density per squared kilometer, population growth in percentage or proximity to large cities in km
Neighboring farms/attractiveness of area Number of farms per squared kilometer Subsidies and government payments Subsidies/financial support in local currency Lease payments Total lease per hectare in local currency
Table 5. Translation of most important topics that concern the valuation of land, translated into required measurement units.
3. Methodology
This research is conducted during a five month internship at the headquarters of Rabo Farm in the city of Amsterdam. Most information provided in this study is based on internal databases, knowledge of employees and several contacts of the company in Poland and Romania.
3.1 Model selection
Page | 22 set of farmland prices in a particular farmland market with dissimilar values of the characteristics for the vector (z) is given by the function P(z). Several authors stress the importance of exactly specified and reliable data to justify the use of hedonic pricing models (e.g. Huang et al., 2006; Sklenicka et al., 2013). These authors argue that a specific hedonic approach requires solutions of complex partial differential equations in order to characterize market equilibrium/conditions completely. This study uses transactional information provided by the ANR which must be handled carefully due to the sometime suspicious characteristics of the data (e.g. prices can be affected by preferred sales and purchase prices in the public market differ from the private market). Therefore a more general approach of analysing possible value drivers is favourable which is comparable to the approach used by Sklenicka et al. (2013). General linear modelling provides insight in the general effects of identified value drivers and offers the opportunity for additional detailed economic evaluation by the use of a hedonic approach.
3.2 Data
Source and availability
Information about farmland transactions is initially collected and documented by the ANR and eventually retrieved from the AMRON database. AMRON offers their information to companies actively participating in Poland’s real estate market. The data consists out of many different documents divided by region. All samples taken together consist of 6,372 farmland transactions in eight different provinces for the period 2010-2013. The data contains transactional information with respect to the source, price, size, geographical location, date, agreement type, constructions, soil quality and typology of the land. The data is an enumeration of individual observations for many different objects and can be categorised as cross-sectional. Therefore the data can be used to assess variability among subjects regardless to differences in time. Data regarding the population, farms, regions and land use classifications are mainly retrieved from the Central Statistical Office (GUS, 2014). Data related to credit availability and subsidies is not added to the model as this is equal for all farmers. Moreover, data regarding lease payments is not added to the analysis as the exact amount of lease is not known.
Selection and filtering
Page | 23 constructions are removed because this could influence the price per hectare significantly. Not many details are provided with reference to the buildings and therefore it is not be possible to take this into consideration. Fourth, transactions without a specific land quality class are removed in order to assess the influence of differences in soil quality in a later stage. For example, comparing land in a perfect arable condition to land covered with trees and sand would not be in favour of this research. Fifth, transactions comprising multiple parcels are removed from the sample. The size and consolidation of land could have a considerable impact on the hectare price and therefore only individual parcels are selected. Furthermore, it is ensured that all transactions took place between the same seller (ANR) and a willing buyer to overcome price differences between ANR sales and private market sales. Seventh, sales prices with abnormal values are removed from the sample in cooperation with Alexander Muhring (Senior Investor, Rabo Farm). This implies that extreme low or high transaction prices are not included in the final sample. As a result, only transactions with a hectare price from 5,200 PLN to 80,000 PLN are taken into account. Furthermore, only transactions from the most recent year (2013) are included. After the adjustments, the remaining samples consist of 1,136 transactions. General linear modelling should provide more insight in the variability of farmland prices in Poland and the differences between the Northern and Southern regions. The following empirical model is developed in accordance with the variables identified by academic literature:
Log(Pi) = β0+ β1Sizei+ β2Population Densityi+ β3Population Growthi (1)
+ β4 Farm Densityi+ β5D1i+ εi
𝐿𝑜𝑔(𝑃𝑖) serves as the dependent variable in this research and is similar to the dependent variable used
in the studies of Chicoine (1981) and Folland and Hough (1991) and is the natural log of the sales price per hectare. The natural log of the sales price per hectare is calculated as follows:
Log (Sales price per hectare) = Total sales price
Page | 24 transaction. Expanding the equation will create a full set of dummies for the three different classes. As a result of testing this independent variable, the original constant will not be included in the expended equation to avoid a singular matrix problem.
The size of farmland serves as the second independent variable and is taken directly from the AMRON database. Size is given as the total area of the parcel expressed in hectares. The third independent variable is population density which can be calculated by dividing the total population by the surface in square meters.
Population Density = Total Population𝑇
Surface (km2) (3)
Information about the total population, surface and land use is not provided by the AMRON database. This data needs to be collected by hand using the locational details and coordinates. AMRON data provides the location of every transaction on four different levels from large to small (i.e. province, district, municipality and precinct). This study uses district as indicator and relates this to information provided by other data bases such as the Central Statistical Office of Poland (GUS, 2014). The country is divided in 379 districts (NTS-4 level) which are often the most detailed level in terms of data availability in databases. This research links data (e.g. population size) to
the district classification of a certain transaction in the original data file. Data about the total population can directly be derived from the Statistical Office database and added to all 1,136 transactions. Furthermore, data about the surface per district is derived from the Urbistat website and added in the same manner.
The fourth independent variable identified by academic literature is population growth. This growth shows the current urban pressure in a particular region at the time of the transaction. Population growth is calculated according to the following equation for 2013. Information from the Central Statistical Office is used to address the total population per district for the relevant periods (GUS, 2014). More than 160 different growth percentages are identified ranging from -3.9% to 19.5%.
Population Growth = ( Total PopulationT+1−Total PopulationT
Total PopulationT )*100% (4)
Page | 25 The fifth independent variable is farm density. This factor implies the number of farms per square kilometre for a particular district. GUS (2014) data is used because the AMRON data does not provide information about the number of neighbouring farms on a specific location. Again the identified districts serve as indicator and the most recent data is based on the agricultural census 2010. Furthermore, the data is filtered for private farms and the total number of agricultural holdings. Only significant holdings with at least five hectares are taken into account. It is interesting to note that the surface taken in the equation is the total surface of arable land and does not include rural area to provide more specific results. Farm density percentages are identified ranging from 0.7% to 8.2%.
Farm Density= ( Total number of farm residents
Surface (km) ) * 100% (5)
Average farmland prices reported by the ANR experience a large annual appreciation as presented in the table below. Therefore, an individual annual analysis will provide more reliable knowledge about the influence of the independent variables on farmland price variability. For this reason, only the transactions from the most recent year (2013) are included in the samples. Comparison with other years will lead to misleading results.
A dummy variable is included to estimate the model for both the Northern and Southern provinces. This study is mainly based on AMRON data collected in eight different provinces (i.e. Dolnosląskie, Opolskie and Lubuskie; Kujawsko-Pomorskie and Wielkopolskie; Pomorskie, Warmińsko-Mazurskie and Zachodniopomorskie). As mentioned before, all circumstances that could influence the price and are impossible to assess need to be controlled. Therefore a distinction is made between the Northern and Southern provinces as both regions possess a unique set of characteristics that could influence farmland prices. The descriptive statistics of all estimated variables are provided in the following table:
Year 2010 2011 2012 2013
Average price per ha.
% increase PLN 13.639,12 PLN 14.697,82 +8% PLN 16.434,26 +12% PLN 22.230,37 +35%
Table 6. Average farmland price increase 2010-2013 in PLN.
Variable Definition No. Obs.
(N=) Mean SD Min. Max.
Page | 26
3.3 Estimation procedure
Three different samples are used to estimate the model. The first analysis is based on the different value drivers within the whole country. The sample size (N) of the transactional data (2013) is respectively 1136. The second estimation provides more insight in the differences within the country. This analysis is therefore based on the data distinguished by the Northern (Warmińsko-Mazurskie, Pomorskie and Zachodniopomorskie) and Southern (Dolnosląskie, Opolskie and Lubuskie) region again focusing on the most recent transactions (2013). The sample size (N) of the second analysis is respectively 790 (Northern provinces) and 242 (Southern provinces).
Assumptions
Four different assumptions apply regarding the use of a multiple regression method with cross-sectional data (i.e. linearity and additivity, statistical independence, homoscedasticity and normality). General multiple regression analysis will only provide accurate estimates about the relation between the dependent and independent variables if this relationship is linear in nature. Plots of the residuals against the predicted values show a symmetrical distribution around the horizontal line with an approximately stable variance. This confirms the existence of a linear relationship between the dependent and independent variables for both analyses (i.e. the estimations based on year and based on region and year). Another issue regarding this research concerns the possibility of multicollinearity. An issue with multicolinearity will affect calculations with respect to the individual predictors. The variance inflation factor (VIF) measures how much the variance of an estimated regression coefficient is increased by collinearity. VIFs of less than 5 for all variables indicate that no problematic multicollinearity is present (VIF around or more than 5 could indicate the presence of multicolinearity). The correlation between the variables is also assessed because the VIF method is not binding. The
6 The number of observations (N=) that report a poor, average or good classification are respectively 478, 270
and 388.
7 The explanatory variable land quality has no mean and standard deviation as it is only categorized in three
different classes.
Land Quality6 Average arable condition of the land classified as poor,
average or good 1136 n/a7 n/a Poor Good Size Total size of the land in hectares 1136 18,60 12,56 5 112 Population growth Current local population growth in percentage 1136 0,021 0,024 -0,037 0,195 Population density Total number of inhabitants per km2 in region 1136 61,99 33,90 20 200 Farm density Number of farms divided by total arable land in region 1136 0,025 0,090 0,007 0,070
Page | 27 correlation matrix confirms the findings and shows no indication of possible multicolinearity and therefore no further transformation of variables in necessary (i.e. highest correlations coefficient is 0.28 between farm density and population growth for the total sample of 2013). The same applies for the second analysis based on year and region (i.e. the highest value found is a 0.27 correlation between farm density and population density in the Southern region). Furthermore, plotting the errors indicate the possible existence of heteroskedasticity. Therefore the samples are tested for conditional homoscedasticity by using the White test in order to establish the optimality of the least squares estimator. The test rejects the hypothesis of homoscedasticity at the 5% level for most estimates. To solve this problem the regressions will be estimated by fixing the standard errors. The standard formula for the errors is wrong and the test can be deceptive although the method is unbiased. Heteroskedasticity robust standard errors are used to establish consistent estimates of the standard errors for all relevant samples at a sufficient 5% significance level. Another question is whether there exists autocorrelation between the error terms. The cross-sectional data does not include a time component or clusters and therefore no autocorrelation is present. The normality of the residual distribution is assessed based on graphical representation and the Jarque-Bera test statistic. The sample sizes of the first analyses are probably large enough to be less concerned about non-normality compared to the small samples used in the next analysis based on year and region. The histograms provide no evidence against a normal distribution (i.e. all histograms are bell-shaped for both analyses). Furthermore, the Jarque-Bera statistic is not significant at the 5% level given the p-value of more than 0.05.
4. Estimation results
The estimation output of the three tested samples show comparable results which are presented in Table 8. The tested variables show to have a significant impact on farmland value variability (P<0.05). The findings show the significant influence of three agricultural factors (i.e. Land Quality, Size and Farm
Density) and two non-agricultural factors (i.e. Population Growth and Population Density).
Table 8. Model Estimation Results
Dependent Variable: Log(total price/hectare)Variables Total (2013) North (2013) South (2013)
Land quality (poor) 4,0473*** 4,0430*** 4,0059***
(272.88) (250.04) (95.04)
Land quality (average) 4,0936*** 4,0878*** 4,0686***
(234.98) (232.58) (76,32)
Land quality (good) 4,2039*** 4,1522*** 4,2433***
(247.11) (236.70) (74.46)
Size 0,0008** 0,0008** 0,0011
Page | 28 The coefficients are an estimate of the parameter and generally indicate to be significant at a sufficient 1% level. Taken into account that Size shows to be significant at a 5% level for the sample Total (2013) and North (2013) but not for the sample South (2013). Moreover, Farm Density shows to be significant at a 5% level for the sample South (2013). The t-statistic is found by dividing the coefficient by the standard error that represents an estimate of the coefficients standard deviation. It can be argued that this is a measure of precision since a large coefficient relative to the standard error generally predicts significance. Furthermore, some dissimilarity between the sample sizes is present. This could explain some of the differences in significance regarding Size and Farm Density. Furthermore, a difference can be noticed between the coefficient estimates of Size and Population Density and Land Quality,
Population Growth and Farm density. The impact of Size and Population Density appears to be marginal
but this is explained by the difference in measurement unit of the variables (i.e. respectively percentages and absolute numbers).
The goodness of fit statistic answers the question whether the model and the independent variables actually explain variations in the dependent variable (Brooks, 2002, p. 107). This research uses the adjusted R² since this measure takes into account the number of explanatory variables added to the model. The adjusted R² provides very comparable results differing from 0.454 for Total (2013) to 0.520 for South (2013). A small increase can be noticed for the samples focusing on a particular region and the best fit is reported for the sample South (2013). The explanatory power of the model is not completely satisfactory although it might be more than expected beforehand. These expectations were based on the reliability and availability of the data and the use of different sources. Also the difficulties relating to the valuation of real estate with respect to the numerous unique value drivers could make this type of valuation a major challenge. It can be stated that it is not possible to determine an accurate farmland price purely on the method and variables used in this research. This is contrast
Population growth 1,3112*** 1,7424*** 2,0187*** (6.67) (8.24) (5.69) Population density 0,0021*** 0,0022*** 0,0016*** (11.57) (12.43) (3.63) Farm density 2,4712*** 2,7113*** 3,7333** (4.60) (4.83) (2.25) Adjusted R2 0,454 0,488 0,520 Included observations 1.136 790 242 Mean 4,347 4,325 4,366
