The Effectiveness of Subsidizing Investments in Polish Agriculture: A Propensity Score Matching Approach

Abstract

Evaluation of the effectiveness of state policy instruments is a permanent element of economic science. This paper addresses the issue of investment support under the Common Agricultural Policy (CAP). Using data on Polish farms from 2015–2023, a Propensity Score Matching–Difference in Differences (PSM-DiD) analysis was conducted to assess changes in the economic results of agricultural producers that invest using this support. The comparison of the economic results achieved by supported investors with both non-investing agricultural producers and unsupported investors is a distinguishing element of this study. The relatively rarely used Competitivness Index (CI), which measures the ratio of earned income to the sum of the alternative use of the owned means of production, was used. The positive change in the CI during the analyzed period was 0.14 higher for supported investors than non-investors. No statistically significant change was found were compared to unsupported investors. A clear increase in income, total fixed assets, liabilities, and the level of production in the population of producers using support in relation to non-investors and investing without CAP support was also observed. However, in relationships with investors using their own funds, these differences were mainly due to the difference in the level of investments and were not statistically significant when introducing a correction regarding the scale of the investment. The obtained results remain in line with the results of research shown by a significant part of economists undertaking a similar issue.

1. Introduction

Investing is a fundamental activity of every entity planning to operate over the long term. In the European agricultural sector, it is necessary not only to ensure the development of an agricultural holding, but to keep it on the market in the coming decades [1]. This continuous modernization of farms through investments in order to increase productivity and remain competitive has been recognized as a necessity [2,3] and it applies in particular to highly mechanized European agriculture, the development of which is closely dependent on the pace of implementation of technical progress [4]. The constant need for investments shapes to a large extent the agricultural sectors of every EU Member State and EU agriculture as a whole [3,5]. Due to the growing requirements imposed on agricultural producers relating to environmental factors, a high level of investment is becoming a need not only for individual farms but for the whole sector [6].

The need to carry out the necessary investments is closely related to a growing problem of access to capital for agricultural producers [7,8]. It is additionally deepened by credit restrictions applied by banks and other credit institutions. Lenders are discouraged by high variability of income and strong dependence on the shape of agricultural policy. It has also been observed that lenders do not take into account part of farmers’ assets as a reliable collateral [7]. These are especially significant problems for smaller agricultural producers [9]. All these problems are particularly important in the case of developing economies. The problem of access to capital enabling investment activity ensuring reaching the level of productivity observed in developed countries is another aspect of the discussed issues [8,10,11,12,13]. Many economists indicate that in the case of the Polish agricultural sector these problems are increased by its technological backwardness [1,12,14]. This is due to, among other factors, the fall of the agricultural sector after political changes at the turn of the 1980s and 1990s, which caused a drastic decrease in the level of investment [15]. An additional problem for Polish farms is the need to find a balance in investment activities between economic and environmental policy goals [16].

The issues highlighted above explain why investments made by agricultural producers are strongly supported financially in the European Union. In the 2014–2022 programming period, investment support was one of the most important tools of the Common Agricultural Policy (CAP). However, its relative importance compared to other tools forming the Rural Development Plans (RDPs) varied across EU Member States. In Poland, in this programming period, which is the period under examination in this work, the funds allocated to investment support accounted for 21.4% of all funds designated to the implementation of the RDP. Nearly EUR 3.9 billion was allocated for all investment support activities during this period, the vast majority of which for “modernization of agricultural holdings”. As part of this operation, approximately EUR 2 billion was paid to nearly 42 thousand agricultural producers by the end of 2023 [17].

Investment support in terms of the funding value allocated to its implementation also remains one of the three main instruments of the CAP in the current programming period 2023–2027 [18]. The importance of investment activities for the agricultural sector and the size of the public support allocated for this purpose make research into the effectiveness of this type of support particularly important [19]. The widespread support for investments carried out by agricultural producers, which has been observed for years [5], creates the need for research. On the other hand, a large number of farmers using this support make evaluation more accessible by significantly broadening the research sample. However, as researchers note [20], despite the relatively high popularity of this type of support among agricultural producers, the number of studies addressing this issue still can be considered insufficient and the number of available databases enabling evaluations is very limited. The need for research on the effectiveness of support for farmers’ investment activities is particularly emphasized in the Polish economic literature, where the size of this aid directed to the agricultural sector is noted as being relatively substantial [1,14,15,21,22]. Another aspect worth emphasizing is the assessment of the appropriateness of the targeted support. It is important to examine whether the investment framework defined in the documents regulating this support allows for improved economic performance. Perhaps agricultural producers who invest without support, but with greater flexibility, achieve better results.

The importance of this type of support for agricultural producers is particularly significant in Poland, due to the observed low level of carried out investments. As can be deduced from an analysis of the data on FADN agricultural accounting holdings in the Member States of the European Union in the analyzed period (years 2015–2023), the average annual level of gross on-farm investment on fixed assets in the 27 EU Member States was EUR 12,034. In countries such as Slovakia or the Netherlands, it exceeded EUR 80,000. In contrast, in Poland, it was only EUR 4235 and was one of the lowest in the EU. A comparison of the average net on-farm investment on fixed assets level over the same period is even less favorable for Polish agriculture. The average annual level for the 27 EU countries was EUR 1401. Only in countries such as Denmark, the Netherlands, and Belgium was it higher than EUR 20,000. In seven EU Member States, the value of net investment was negative, which means that the value of depreciation exceeded the value of investment. The Polish agricultural sector had the third lowest level of net investment, with an average of EUR −1183. Only agriculture in Italy and Greece showed worse results, with an average annual net investment of EUR −2744 and EUR −1930, respectively.

The research question posed in this paper is whether support for investments carried out by agricultural producers allows for a competitive advantage over the group of farms where investment activity is relatively low and the group where investments are carried out without the use of pro-investment support. To answer this question, individual farms’ data and PSM-DiD methodology were used. It was decided that the point of reference would not be the entire subpopulation of all other agricultural holdings but two different subgroups. The first one consists of agricultural holdings that did not carry out any investments of an appropriate scale at all. The second control group are farms that, for various reasons, took on the burden of investment without the support offered by the CAP. Establishing two control groups as well as employing a relatively rarely applied competitiveness index can be considered as an element of this study’s innovative approach.

2. Literature Review

Initially, co-financing investments from public funds may seem to have only positive consequences for the agricultural sector and, above all, the beneficiaries of such support. However, a careful study of all the theoretical consequences of such support forces us to seriously consider the possible negative effects. It is therefore worth looking at both the positive and negative possible effects of pro-investment support. The former certainly include the possibility of improving the profitability of production, proper correction of the production structure in relation to changing market conditions, increased competitiveness of both individual producers on a given market and the entire sector in international terms, and the possibilities of obtaining benefits from achieving economies of scale [23,24,25,26,27]. The most common negative consequences of providing pro-investment support include the following effects: transfer of funds to producers who would also invest without the offered support (deadweight loss), the increased risk of unsuccessful investments due to the reduced cost of capital, time-consuming and cost-intensive procedures, or the obligation to meet limits specified in support programs. These limits can also reduce benefits of scale. The reduced cost of capital also encourages some producers to choose capital-consuming variants that may lead to choices that are far from optimal at a given time [23,24,28,29]. Economists also emphasize the phenomenon of the reduced concern of agricultural producers regarding the future effects of investments when pro-investment support is treated as an element of income [26]. Others note that this type of support can be the cause of the artificial maintenance of farms when their economic existence is ensured only by such support [30].

Above, several of the most frequently indicated theoretical premises indicating a positive and negative impact of investment support on the results obtained by beneficiaries of such programs have been cited. These phenomena have been the subject of a number of works by agricultural economists. The research results obtained by them will be presented below. It is also worth adding that similar assessments of the balance of negative and positive impact were studied in relation to other forms of support offered to farmers in CAP. A broad review of this type of research, covering 41 economic articles referring to the impact of CAP payments on the productivity of farms, does not indicate that the answer to the question about the net effect is simple [30]. The research results in fifteen cases indicate a negative effect, in four cases a positive effect, and in the remaining cases an ambiguous effect. Below, the results relating exclusively to pro-investment support are presented.

As in the case of other support instruments offered to agricultural producers under the CAP, also in the case of investment support, one can find studies in the literature with completely different conclusions. In order to structure the literature review undertaken, it was decided to present the findings of economists addressing this issue within three groups. The first of these presents studies whose authors have demonstrated the positive impact of the type of support analyzed on the performance of agricultural producers. The positive effects of investment support were indicated by, among others, Ratinger et al., who proved a positive impact on labor productivity for medium-sized farms in Czechia [5]. The favorable impact of this kind of support on gross value added, income, and employment in Czech agriculture also has been proven [19]. There is evidence for positive impact on firm productivity in small agricultural firms in Sweden [26]. Supported investments also have a beneficial impact on total factor productivity in German dairy farms [31]. The improvement of the economic situation due to this kind of support was proven on the example of all Polish farms [14] as well as on the subpopulation of dairy farms [32]. The beneficial effect of the discussed type of support has been also proven in case of farm performance in Slovakia [33], as well as Austria [34], Czechia [35], and Bulgaria [36]. There are also studies that indicate that the positive impact of investment support is limited to specific periods [37] or production types [35].

The second group of economic studies cited in this article included those where, on the basis of calculations, it was decided to regard this type of support as having more losses than benefits. There are numerous studies that indicate a negative impact of investment support under the CAP. This applies, for example, to the impact on efficiency and productivity of European farms [38] or technical efficiency of German organic dairy farms [39]. The negative impact of subsidies on total factor productivity of southern Italian farms has also been detected [40]. Despite the positive impact on other variables, in the case of research on the analyzed form of support in Slovak conditions, a negative impact on labor productivity was proven [19]. The overinvestment observed among beneficiaries of pro-investment activities in Poland should also be considered as an unfavorable effect of support [41]. The deadweight effect was also discovered in research examining the issue of investment support in the German, Slovak, and Polish food sectors [19,42,43]. It was also noted during research on investment support in the aquaculture sector in Sweden [44].

The third group of highlighted studies contains those scientific papers in which the authors, on the basis of the information collected and the calculation results obtained, were unable to make a clear assessment of pro-investment support policies in agriculture. In Mary’s research on farms in France, no statistically significant effect of pro-investment support was detected [45]. The same results applied to employment effect in the German agricultural sector [46], labor productivity on Lithuanian farms [47], or efficiency and productivity in Europe as a whole [48,49]. In some of the studies, the results obtained indicated a positive effect of subsidies in one period and a negative effect in another [50]. Taking into account the significant number of studies in which the effects of pro-investment support were found to be statistically insignificant, the results of studies in which a significant time lags between investment and accumulated observable effects was observed should be considered interesting [23,51]. An insufficiently long period between obtaining support and conducting the study may be one of the reasons for the lack of statistically significant impact of investment subsidies on the economic performance of farms.

The heterogeneous results should not be surprising. It should be remembered that each of the cited studies concerned a different country, a population of agricultural producers with dissimilar characteristics, and finally various time horizons. It should be clarified that the focus of most agricultural economists on data from a single country is not solely due to the availability of data—although this may play a role—but is also the result of differences between EU Member States in the detailed rules governing the conditions for obtaining subsidies for investments in agricultural holdings [19]. There is one more substantial difference between the studies cited—the studies were conducted using different methodologies. For instance, in the case of research on the impact of different types of CAP support on farm productivity, more than ten different methodologies were used. The most common is stochastic frontier analysis and data envelopment analysis [30].

When it comes to the assessing the effectiveness of investment support, the fundamental problem is constructing a control group with appropriate characteristics. Agricultural holdings are highly heterogeneous and participation in investment support programs is voluntary. As a result, there are systemic differences between farms benefiting from pro-investment support and other agricultural holdings [3]. In a very limited number of studies, the results of which were cited above, the random nature of support use was assumed [45,50]. In most studies, researchers saw the need to use a counterfactual approach to overcome this evaluation problem. One of the most commonly used solutions is the use of the Propensity Score Matching (PSM) method [3,5,19,47,52,53]. A significant number of such studies additionally use Difference-in-Difference (DiD) estimators [3,19,40,46,53]. It is also worth noting that, most commonly, the Nearest Neighbor Method (NNM) is applied [3,5,52]. In some studies, the Coarsened Exact Matching technique is also used [23,26]. Studying the impact of pro-investment support from the micro perspective using data collected in the FADN databases is the most common approach in the literature on this issue. Only exceptionally in the literature on the impact assessment of subsidies encouraging agricultural producers to invest are less sophisticated techniques used, such as correlation analysis and various regression methods [21] or survey research [54].

An analysis of the literature addressing the issue of the effectiveness of investment support directed at agricultural producers also made it possible to observe considerable diversity in terms of the methodology used. This concerns not only the type of econometric methods applied but also the set of indicators used. The most commonly used indicators of changes occurring on farms as a result of using pro-investment subsidies include gross value added, farm profits, farm employment, and labor productivity [14,19,23,26,32,35,37,46,47]. In a large number of publications addressing the issue of the impact of investments on the functioning of farms, changes in the value of total factor productivity (TFP) are analyzed [8,23,37,45,53,55,56,57].

It should also be borne in mind that the issue of the benefits of investment support and the effects of the investment itself is in itself a very broad issue. The level of benefits resulting from investment activities is affected by existing infrastructure in agricultural areas, which is mentioned by agricultural economists analyzing Bulgarian conditions [58]. It should be also remembered that this work focused only on short-term effects related to investment support policy. Meanwhile, the phenomenon of changes in investment outlays can be analyzed in a broader context, taking into account even problems related to sustainable development [59].

3. Materials and Methods

3.1. Data Source

The main source of the data used in the calculations comes from the Polish FADN (Farm Accountancy Data Network) database. The FADN database is commonly used in economists’ research to assess various issues related to agriculture in its broad sense. This applies to both the analysis of agricultural management efficiency [60,61] and the possibility of using FADN data to assess various agricultural policy instruments [62]. This is mostly due to numerous advantages that the FADN database has. First of all, it offers a wide set of indicators regarding the production and financial situation of agricultural producers. In addition, the methodology for obtaining, collecting, and sharing data is largely unified for most European Union (EU) member states, which allows for the geographical expansion of the scope of the analyzed phenomena in the future. An extremely important advantage of using this data source is the possibility of creating panels of farm data continuously reporting on economic activity in agriculture. This, in turn, allows for the assessment of the long-term impact of the analyzed phenomena. The Polish FADN database contains information on a relatively large number of farms. Ultimately, it was decided to analyze the period between 2015 and 2023. During this period, the number of farms in individual years ranged from 12,116 in 2020 to 11,029 in 2023. This allowed for the isolation of a data panel of 4495 agricultural producers continuously submitting their data to FADN. It is worth mentioning that the large number of agricultural producers in Poland in general, and those included in the FADN database in particular, helped to avoid the problem noted in the literature [63] that consists of difficulties in finding farmers who do not benefit from pro-investment support.

3.2. Variables Selections

The assessment of changes in the efficiency of farms carrying out investments using the PSM methodology requires a number of stages of processing the available data. In the first step—after generating the data panel—two basic concepts used in the work were defined: investment and subsidized investment. In the FADN database, agricultural producers are characterized by two variables relating to investment. These are gross investments (variable SE516) and net investments (SE521). It was determined that the analysis would cover the values of net investments, which is common in the literature [64,65], mostly due to the need to take into account the differences in investment effort between small and large agricultural producers. The value of net investment includes the value of depreciation. Therefore, a larger producer with a larger initial stock of physical capital and a higher level of depreciation must make a correspondingly larger investment than a smaller producer to achieve the same value of net investment. Then, the threshold value of purchases made in the farm was determined, from which one can talk about an investment. Based on the data analysis, it was decided to set the threshold at the level of PLZ (Polish zloty) 100,000 (approximately EUR 25,000). Since it should be taken into account that the implementation period of some investments exceeds one calendar year, an additional threshold was the total value of the investment in two consecutive years exceeding PLZ 120,000.

An agricultural producer who carried out investment activities with a value exceeding the established thresholds is treated as an investor in a given year. If in the same year he received a subsidy for investment purposes with a value exceeding PLZ 50,000, this investment is perceived as a subsidized investment. Other agricultural producers who did not carry out investment activities in the analyzed period were included in the control group.

In order to avoid distortion of the comparison results by investment activities of farms in the years 2021–2023, farms were removed from the analyzed data panel if investments exceeded the established threshold values in that period. This restriction was introduced so that the economic results of the surveyed agricultural producers in 2022 and 2023 were not dependent on investment activities carried out in recent years. The control group therefore includes only farms that did not carry out sufficiently large investments in the years 2016–2023. On the other hand, both surveyed groups of investors refer to farms that carried out investments only in the years 2016–2020. This additional condition caused the number of surveyed farms to decrease by 992 farms (22.1% of the original data panel) to a final 3501 agricultural holdings.

This significant reduction in the size of the study population is purely due to concern for the representativeness of the results. Agricultural producers that are in the middle of the investment process may show results far from the long- or medium-term average since shortly after an investment is made, farms tend to show the weaknesses associated with incurring its cost without showing the benefits that usually become apparent later.

The undertaking of investment activity by agricultural producers referred to one five-year period. It covers the years 2016–2020 and refers to the period of validity of the Rural Development Programme 2014–2022. This choice of cut-off dates results from the desire to maintain at least a two-year period between the end of the investment period and the year from which the data needed to conduct the efficiency analysis will come.

In this article, it was decided to use the competitiveness index (CI), the formula of which was developed based on the competitiveness index developed by Werner Kleinhanss [66]. Its essence is to compare the value of income obtained from agricultural activity to the hypothetical value of the alternative use of the owned means of production. The value of this coefficient was estimated for each farm from the analyzed panel. This is an indicator often used in the works of Polish agricultural economists [66,67,68]. It has a number of beneficial properties and is ideally suited to benchmarking when assessing the effects of investment processes carried out in agriculture.

The alternative value of own labor was determined as the ratio of the annual work unit of used own labor and the average rate for hired labor in FADN farms in a given year, separately for each of the four macroregions of Poland. The alternative value of land was calculated as the ratio of own land resources in the farm and the average land rental rate in FADN farms in a given year. In this case, regional differentiation of this rate was also assumed. On the other hand, the alternative cost of equity capital was estimated on the basis of its value and the level of interest on total deposits for the household sector.

Summing up, the formula for calculating the competitiveness index is as follows:

CI = Income/(ACw + ACl + ACc)

(1)

where

• ACw—the alternative cost of labor;
• ACl—the alternative cost of land;
• ACc—the alternative cost of capital.

A value of the competitiveness index exceeding 1 indicates full coverage of the costs of production factors, while a value less than one indicates their incomplete coverage. To calculate the value of the competitiveness index, the time series of various variables relating to the state of capital and land ownership by individual agricultural producers and the level of own labor involvement in the farm were downloaded from the FADN database.

In addition to estimating the change in the average level of the competitiveness index for the studied groups in the analyzed years, changes in the level of other indicators characterizing the economic situation of farms were also analyzed. Finally, this work presents changes in variables such as the level of income obtained by farmers, the value of assets held, the level of debt, and the value of production.

A number of various variables were also used in the initial phase of the analysis, when an analysis of their correlation with the level of value of investments made in farms was carried out. The aim of this stage of the research was to abstract from a wide range of variables those whose average level is the most strongly differentiated between groups of farms carrying out and not carrying out investments.

Table 1. FADN variables used in study.

Name of Variable	Symbol of Variable	Purpose of Use
Economic size of farm	SE005	M 1
Total labor input (AWU)	SE010	M
Own labor input (AWU)	SE015	M, I 2
Hired labor inputs (AWU)	SE020	M
Area of agricultural land (ha)	SE025	M, I
Area of rented agri. land (ha)	SE030	I
Total output (PLZ)	SE131	R 3
Family farm income (PLZ)	SE420	R
Total assets (PLZ)	SE436	I, R
Fixed assets (PLZ)	SE441	R
Farm buildings (PLZ)	SE450	M
Total liabilities (PLZ)	SE485	M, R
Average farm capital (PLZ)	SE510	M
Land Value		I

¹ M—matching procedure; ² I—competitiveness index calculation; ³ R—Results.

presents a full list of variables from the FADN database used in this study. The third column presents information on the nature of their use. Only those variables that, based on the analysis of correlations with the aggregated variable defining the level of investment, did not show significant differences at the level of average values between the analyzed subgroups of farms (non-investing, investing independently, and investing using subsidies) were omitted here.

The variable marked with the name Land Value does not have its own symbol in the FADN methodology. It was additionally calculated by the Polish FADN employees specifically for the reported demand from the author. The abbreviation AWU refers to a unit measurement and means Annual Work Unit.

3.3. Methods Used

In this paper the PSM-DiD method was used. The creators of the PSM concept are considered to be P.R. Rosenbaum and D.B. Rubin, who in 1983 published an article entitled “The central role of the propensity score and observational studies for causal effects” [69]. This method was created to overcome problems related to comparing the results obtained by units subject to the influence of a given event with those that were free from this influence. It is a counterfactual method that allows for minimizing selection bias when estimating the average effect of the impact on units subjected to the intervention. The need to use this method results from the fact that in economic research, it is most often impossible to conduct experiments based on randomization. In observational studies, selection bias exists, consisting in the imbalance of variables between the treated/study group and the control group. Adding the DiD method helps through eliminating the impact of other policies or macroeconomic factors affecting the sector. This technique is seen as a highly applicable estimator of any policy impact, especially when the outcome data is available before and after the policy’s implementation. In the first step, policy beneficiaries are matched with comparable non-beneficiaries using the PSM technique. In the next step, the net effect of support is calculated on the basis of the value of the differences of the analyzed variables before and after the implementation of the policy for both groups.

The purpose of impact evaluation is therefore to measure the effect of the impact on selected variables. These variables constitute the basis for calculating the effect of the intervention. Most often, in this type of research, the aim is to estimate the average effect of the impact on units subjected to the intervention, which is called the ATT (Average Treatment in the Treated).

The PSM method itself consists in matching the group subjected to the intervention with such a control group, selected from the population of units not subjected to the impact, that the distributions of the characteristics of the selected explanatory variables in both groups will be balanced. The entire selection process must be based only on the observed characteristics of explanatory variables and all variables influencing participation in the program and the result of the output variable must be observed by the researcher. One of the model assumptions is that each unit of the population must have a positive probability of reaching the group of beneficiaries.

Most often, the values of the propensity score function are determined as theoretical values of the model estimated for this purpose. This model is only a means to achieve the goal, which is to balance the variables. When estimating it, it is necessary to take into account variables that affect both the selection and the final result of the output variable. A key step in the Propensity Score Matching method is to assess the match between the control group and the intervention group, because the quality of the estimated effects of the interactions depends on the quality of this match.

In the research described in this paper, matching was perform using two methods. First, the Nearest Neighbor Matching (NNM) method was applied. This method involves searching through the list of treated units and selecting the closest eligible control unit to be paired with each treated unit. It is worth mentioning that each pairing occurs without analysis of how it affects the overall matching within the groups being analyzed. In this sense, one could state that there are no optimizations for any criterion. Due to its simplicity, NNM is considered the most common form of matching [70]. However, since results obtained using this method are quite often unsatisfactory, the Optimal Full Matching (OFM) method was used as well. In this technique, every treated and control unit is assigned in the sample to one subclass each [67]. Each subclass contains one treated unit and one or more control units or one control units and one or more treated units. It is optimal in the sense that the chosen number of subclasses and the assignment of units to subclasses minimize the sum of the absolute within-subclass distances in the matched sample. Weights are computed based on subclass membership, and these weights then function like propensity score weights and can be used to estimate a weighted treatment effect.

After matching each method’s procedures and estimating its effects, the computation to estimate the average treatment in the treated (ATT) was applied. Cluster-robust standard errors were analyzed, with pair membership as the clustering variable. The comparison of ATT results for different subpopulations was the final goal of the calculations, which indicated differences in the efficiency of the analyzed ways of carrying out investment activities on farms in Poland.

4. Results

4.1. Size of the Studied Populations

As already mentioned, the number of farms in the analyzed panel that continuously kept agricultural accounting records in the years 2015–2023 amounted to 3501. Based on the data analysis, it was found that in the specified five-year period (2016–2020), 824 farm owners, which represents almost one quarter of the entire sample, carried out at least one investment whose value exceeded the established threshold values. Among those farm owners, 277 benefited from support under one of the measures supporting investment projects. Their number corresponded to 7.9% of all surveyed farmers. Almost twice as many, i.e., 547, farm owners carried out investments without support from public funds.

Summing up, three subgroups of farms were studied. In 277 farms, at least one investment was made with the support of the Common Agricultural Policy. In 547 cases, at least one investment was made and none of them was related to support from public funds. However, for both of these subgroups of investors, the control group was the one consisting of 2677 farms in which no investment of a sufficiently large scale was made in the five-year period.

It should also be noted that it was intentional to present the results without a regional or agricultural production type disaggregation. The PSM-DID methodology used in this paper makes divisions of this kind unnecessary. Moreover, the results obtained with a greater decomposition of the surveyed population did not differ significantly from the results obtained for the entire group. In the author’s opinion, excessive divisions could adversely affect the readability of the results or the assessment of their significance.

4.2. Characteristics of Analyzed Subpopulations

The identified populations of agricultural producers making investments in the analyzed period differ significantly from the farms in the control group, which confirms the correctness of the choice of the PSM method for studying the analyzed phenomena. Moreover, the differentiation also applies to the two analyzed subgroups of farm owners who carried out investments on a sufficiently large scale. The diversity of the studied subpopulations is presented below in

Table 2. Differences between the averages of selected variables for the years 2015–2016 between the group of farms carrying out subsidized investments and the group of farms not investing.

Variables	Non-Investing Producers	Supported Investors	Unsupported Investors
Total labor input (AWU)	1.80	2.03	2.10
Area of agricultural land (ha)	27.96	45.47	50.87
Total production (PLZ)	167,361	290,492	395,528
Total assets (PLZ)	1,104,844	1,741,526	2,043,141
Farm income (PLZ)	56,419	101,993	130,975
Total liabilities (PLZ)	72,104	158,171	206,971

Source: Own calculations.

, using a few of the most important characteristics of the analyzed groups. The six key—from the point of view of the calculations conducted—presented variables are the total labor input, total agricultural area, total assets, total production, family farm income, and total liabilities. It should be recalled that the initial state was determined for each farm for the average of 2015–2016. Similarly, the final state did not refer to a single year, but to the state in 2022–2023. Two-year variables were used to minimize the impact of short-term changes in price relations on the analyzed variables, as was done in other works on similar topics [12,71].

It should be added that the set of variables used during the matching procedure was changing. However, it was noted that extending this set did not significantly affect the results obtained. Hence, it was decided to use a relatively short list of variables used in the calculation of the PSM in order to increase the clarity of the calculations carried out.

The average values of selected variables in the three subgroups studied, presented in Table 2, clearly indicate that there is a significant differentiation in the scale of production activity conducted between the groups studied. This is primarily due to the difference in the levels of use of production factors.

Non-investing farms are characterized by the lowest average level of labor, land, and capital involved in production. On average, the largest are those farms that carry out investments without using support offered under the common agricultural policy. The level of labor involved in production is 17% higher than the average for non-investing farms. In the case of the land factor, this difference is as much as 82%, and for the capital factor almost 85%.

Agricultural holdings where farmers carried out investments using support in the analyzed period are located between the two groups mentioned above in the discussed context. However, it can be noticed that the differences between farms using pro-investment support and farms investing without this aid are smaller than the differences between farms investing and those not carrying out investment projects at all.

Differences in the distribution of four selected variables between control group and farmers investing using support are presented in Figure 1. It can be seen that among non-investing farms there are particularly many units with a very low level of economic size or with a small area of land used for agriculture. As for the level of debt, in both groups discussed there is a large percentage of farms with no financial liabilities; however, among investing farms there is a significant group of farms with a significant level of debt. It is also clearly visible that in the control group there is a much more frequent possession of a low level of capital.

4.3. Matching Results

In the first step, the matching procedure was performed using the NNM. Although this method significantly improved the matching of the groups, the final data resulting from the reconstruction of the control group should be considered not fully satisfactory. Balance had improved significantly; however, overall, it still was not good enough. This statement applies primarily to the comparison of means of variables such as total labor input, the value of buildings or capital, and the level of liabilities. Failure to correct these differences indicated the possibility of improving the matching of the control group to the group of investing farms.

In the next step, OFM was used. As was mentioned, in this technique units are assigned in the sample to one subclass each in a way that makes it possible to minimize the sum of the absolute within-subclass distances in the matched sample. Overall differences in the results obtained using these two methods are shown in

Table 3. Differences between the averages of selected variables for the years 2008–2009 between the group of farms carrying out subsidized investments and the group of farms not investing for two selected matching procedures.

Variables	Nearest Neighbor Procedure			Optimal Full Matching Procedure
	Means Treated	Means Control	eCDF Mean	Means Treated	Means Control	eCDF Mean
SE005	56,194.6	56,841.9	0.03	56,194.6	55,714.5	0.02
SE010	2.03	2.08	0.03	2.03	2.04	0.02
SE020	0.23	0.26	0.02	0.23	0.21	0.03
SE025	45.47	45.23	0.05	45.47	45.00	0.03
SE450	274,207.0	265,052.9	0.02	274,207.0	275,516.0	0.02
SE485	158,170.6	150,177.8	0.07	158,170.6	156,831.0	0.07
SE510	872,577.4	855,354.1	0.01	872,577.4	874,606.7	0.01

Source: Own calculations.

, where the means of seven variables from 2015–2016 are presented on the example of the matching of the group of investors with support and the control group. In addition to the differences in these means, the values of the mean eCDF function are also presented. As can be seen from the comparison in Table 3, the matching results for the OFM method are characterized by more desirable features. Similar differences were also observed when assessing the matching effects in the case of the other three analyzed groups.

It is interesting to compare the selection of units for the control group after matching using these two different methods. Figure 2 presents two graphs showing the distribution of individual observations (agricultural producers) in the control group and in the treated group in relation to the value of the propensity score function. The upper graph shows the matches for the NNM, which can be recognized by the significant number of unmatched observations. These are mainly observations for which the computed value of the PSM function is relatively low, as in this range there is a significant quantitative predominance of observations from the control group. The information relating to these observations is permanently lost. The lower graph, on the other hand, illustrates the assignment of different weights to individual observations from the control group in the OFM method. Differences in the size of the circles assigned to each observation correspond to differences in the value of the assigned weights. All the information from the control group is fully used here, which improves the quality of the comparisons made between the treated and control groups. Given that the distribution of the population of agricultural producers carrying out investments in terms of the PSM function computed value differs significantly from this distribution in case of control groups, the information loss associated with the use of the NNM is significant. This is an additional factor confirming the validity of the choice of the OFM for assessing differences between two examined subpopulations of agricultural producers.

4.4. The Average Treatment Effect in the Treated (ATT) Results

The final stage of the calculations was to compute the ATT effect for every analyzed pair of subgroup of agricultural producers. Initially, it was planned to analyze only the results of changes in the level of the competitiveness index, but due to the lack of unambiguous results, an analysis of the ATT effect was also carried out for changes in the level of such variables as the agricultural income obtained (SE420), the value of fixed assets (SE441), the level of total liabilities (SE485), and the value of production (SE131).

It is worth clarifying that, due to the use of the PSM-DID methodology, all changes are given in nominal values. Inflationary changes affected each of the subpopulations analyzed to the same extent. This also applies to some extent to changes in agricultural commodity prices.

In the first approach, the ATT effect was estimated for a pair of producers not investing and investing with the support. The results are presented in the

Table 4. Selected variables ATT estimations for group of supported investors compared to non-investing producers.

Variables	Estimate	Std. Error	z	Pr(>|z|)
Competitiveness index	0.14	0.08	1.73	0.0828
SE420	99,171	13,229	7.5	<0.001
SE441	696,397	39,393	17.7	<0.001
SE485	142,857	18,301	7.81	<0.001
SE131	213,152	25,021	8.52	<0.001

Source: Own calculations.

. When comparing these groups, the ATT effect related to the competitiveness indicator is relatively insignificant. Although an increase in the level of this indicator (0.14) is observed as a result of carrying out investments using public support, the value of the p statistic means that statistical significance can only be considered when using the relatively unrigorous threshold level of p = 0.1. The statistical power is much greater when estimating the ATT effect for the remaining four variables studied. It should be recalled that the averages for 2015–2016 are compared with the average for 2022–2023. In the case of income (SE420), it can therefore be said that carrying out an investment with support causes an increase in income by almost PLZ 100,000 higher than in the control group. This increase is certainly the result of the rise in variables such as fixed capital (SE441—an increase of almost PLZ 700,000) or production value (SE131—an increase of over PLZ 200,000). The significant increase in the fixed capital value is due to the investments made and to the substantial level of capital depreciation in farms without sufficiently large investments. The increase in capital has a positive effect on income indirectly by increasing production possibilities. However, taking into account the methodology used by FADN, its direct impact is negative, because increased capital resources cause a growth in the value of depreciation, which reduces the amount of income obtained.

In addition to these changes, which should definitely be considered positive, there is also an increase in the SE485 variable, which indicates the level of indebtedness of agricultural holdings. Despite using public support, farms carrying out investments had to use external financing, and to a much greater extent than was the case in the control group.

To sum up, there is a significant positive impact of investments on the functioning of agricultural holdings. In addition to the relatively small increase in the competitiveness index, there is a strong increase in both the level of production and income obtained. A significant rise in the level of fixed assets held is also observed, which is also associated with an increase in the level of indebtedness.

The next compared pair of selected groups of agricultural producers is the subpopulation of producers carrying out investments without support offered under the CAP and the control group created by using the OFM technique. The results of the ATT effect estimation for the five variables studied are presented in

Table 5. Selected variables ATT estimations for group of producers carrying investment without public support compared to non-investing producers.

Variables	Estimate	Std. Error	z	Pr(>|z|)
Competitiveness index	−0.01	0.08	−0.07	0.942
SE420	56,276	12,539	4.49	<0.001
SE441	399,022	37,883	10.5	<0.001
SE485	100,837	9917	10.2	<0.001
SE131	136,025	20,953	6.49	<0.001

Source: Own calculations.

.

First of all, it should be noted that there are no statistically significant changes in the competitiveness index. In the case of the remaining variables, however, statistically significant increases are observed. Substantial growth of the fixed assets level is observed. It can also be stated that as a result of carrying out investment activities in the period 2016–2020, there is significant growth in the value of production by over PLZ 136,000 and the value of income by over PLZ 55,000. Similarly to the case of comparisons relating to investors using support under the CAP, in this case a significant increase (over PLZ 100,000) in incurred financial liabilities is also observed.

The results are therefore largely consistent with those observed for the group of investors using support. However, it should be noted that there was no increase in the competitiveness index and significantly lower increases in the remaining variables presented. Both the increase in income and the value of fixed capital was lower by approx. 43%. In the case of the rise in the value of production, the increase for investors not using support was lower by over 35%. The relatively smallest differences refer to the growth in the value of incurred liabilities, which were lower in the analyzed case by less than 30% compared to those investing with support.

Comparison of ATT effects estimated based on the matching of two separate groups does not allow for relative assessment of the superiority of some results over others. This is because ATT values refer to comparisons with the control group, while these may differ significantly. The shape of the control group depends on the characteristics of the treated groups. Meanwhile, as shown in Table 2, farms carrying out investments without support are characterized by a significantly higher level of production factors involved in production. This makes the control group in their case also economically stronger, which may influence the relative weakening of the observed ATT effects. However, this is not the only difference between the two groups of agricultural holdings carrying out investments.

Table 6. Mean value of investments and pro-investment support in 2016–2020 in analyzed subgroups.

Variables	Non-Investing Producers	Supported Investors	Unsupported Investors
Mean value of investments	16,872	542,823	357,408
Mean value of pro-investment support	424	161,671	1622

Source: Own calculations.

presents the differentiation of the three analyzed groups in terms of the average level of investments made in 2016–2020 and the average level of investment support received. It is worth noting that the non-zero support values in the group of farms that are considered not to benefit from support at all result from the adopted thresholds of the pro-investment support value. The situation is similar with the positive value of investments in the control group. Farms from this group made small investments, but none of them exceeded the threshold set in the study.

The most important information contained in Table 6 is the significant difference in the mean value of investments carried out in the years 2016–2020. Agricultural producers who received pro-investment support could afford significantly higher expenses. The average level of investment in their case is 48% higher than that observed in the group of farms not using support.

The last stage of the research involved creating a control group and comparing the results achieved by two groups of agricultural producers carrying out investment projects. Due to the significant difference in the size of the groups, it was decided that the subgroup for which the control group would be created would be investors using support under the CAP. The same procedure was carried out as in the earlier estimation of the ATT effect and the results of these calculations are presented in

Table 7. Selected variables ATT estimations for group of supported investors compared to producers carrying investment without public support.

Variables	Estimate	Std. Error	z	Pr(>|z|)
Competitiveness index	0.06	0.92	0.61	0.537
SE420	38,472	13,686	2.81	0.005
SE441	283,893	44,181	6.43	<0.001
SE485	48,666	21,237	2.29	0.022
SE131	73,675	26,704	2.76	0.006

Source: Own calculations.

.

Based on the calculations performed, it can be stated that the economic situation of agricultural producers using pro-investment support has changed more favorably. A higher increase in income or production is observed here than in the case of farmers not using support. A significantly higher increase in fixed assets is also observed. The only thing that may be surprising is the higher increase in liabilities. It should also be noted that no statistically significant differences are observed in the change of the competitiveness index between the studied subgroups.

Apart from the ATT effect for the value of liabilities, these results should not be surprising. They result from a significant difference in the value of the investments themselves. Taking this into account, it was decided to create an additional matching process, which resulted in the creation of a new control group. A new element here is the fact that the value of investments was taken into account when performing the matching, so that this difference does not affect the values of the estimated ATT effect. The results of these additional calculations are presented in

Table 8. Selected variables ATT estimations for group of supported investors compared to producers carrying investment without public support when the value of the investments made was taken into account.

Variables	Estimate	Std. Error	z	Pr(>|z|)
Competitiveness index	0.06	0.12	0.47	0.638
SE420	17,784	16,235	1.10	0.273
SE441	94,248	48,956	1.93	0.054
SE485	7776	21,676	0.36	0.720
SE131	11,891	33,954	0.35	0.726

Source: Own calculations.

.

If the matching procedure neutralizes the influence of the higher average investment value, the differences in the economic situation of agricultural producers from the two investor groups become statistically insignificant. Only in the case of one variable was statistical significance proven. This refers to the change in the value of fixed capital, which in the group carrying out investments with support from public funds increased significantly compared to the control group. In the case of the remaining variables, no statistically significant difference was found between the groups studied.

In summary, the results of the calculations indicate that over the years, farms carrying out investments have been characterized by significantly better economic results. It was also observed that a comparison of the results of farms of investors using support offered under the CAP with farms not carrying out investments brings more favorable results, especially in terms of the increase in the value of production, fixed capital and income than in the case of farms of investors not using support. A direct comparison of the two subgroups of agricultural producers carrying out investments, when the differences resulting from the larger scale of investments among farms whose owners used support are eliminated, does not bring a conclusion as to the better economic form of either of the two groups. In each case, however, an increase in the value of financial liabilities is observed among agricultural producers carrying out investments.

5. Discussion

The results presented in this paper constitute a voice in the discussion on the effectiveness of this part of the CAP, which is responsible for supporting investments made by agricultural producers. The results obtained by farmers who took advantage of the possibility of co-financing their own investment costs indicate a largely beneficial use of public funds in this area. Beneficiaries of pro-investment support not only significantly increased the scope of their total fixed assets, but also obtained significantly higher income and production value than the control group consisting of farms not making sufficiently large investments. It is worth remembering, however, that this happened with a simultaneous significant increase in liabilities. Only in the case of the competitiveness index growth may the relationship be seen as less clear, although still visible and statistically significant at the level of p = 0.1.

What constitutes a unique element of this research in the perspective of the discussion on the effectiveness of pro-investment support within the CAP is the extension of the analyzed subpopulations of agricultural producers to include those agricultural holdings in which investments are made using own funds. Comparison of agricultural producers carrying out investments without using CAP instruments with non-investors did not show the difference in changes of CI. However, there is an increase in the value of other variables like total fixed assets, income, and production value. It should be noted, however, that these increases were significantly lower than those shown by CAP support beneficiaries and amounted to between 57% and 64% of those increases.

As for the comparison of investors using CAP support with those who carried out the investment only using their own funds, there was no statistically significant increase in the level of CI. However, there was a significantly higher increase in income, total fixed assets, and the level of production. However, this growth resulted mainly from the difference in the scale of investments. The average value of investments supported by public funds was 52% higher than investments carried out without this support. When the differences in the scale of investments were reduced as part of the PSM procedure, it turned out that the statistically significant increase concerned only the higher growth of total fixed assets.

These results should be considered quite interesting. In their context, it is much more difficult to formulate conclusions about the beneficial impact of investment support under the CAP. This impact obviously exists, but it mainly consists of expanding the population of agricultural producers who can afford to make investments. However, another conclusion resulting from the calculations can be considered as defending pro-investment activities against criticism. The effectiveness of investments made using support is in no way lower than the effectiveness of investments made by producers that, for various reasons, did not use this support. This may mean that the framework defining the permissible shape of subsidized investments did not negatively affect their effectiveness. The implications for CAP politics seem clear. The investment support policy clearly improves the situation of agricultural producers and allows them to carry out investments on a sufficiently higher scale, which is particularly important in the case of agricultural producers from countries such as Poland, where financial restrictions are still a serious barrier to overcome.

6. Conclusions

The results obtained through this research enrich the knowledge on investment processes in European agriculture and the role of state support in this process. It can be also a promising starting point for further research on this topic. For example, the analysis can be expanded to examine different production types or regions. The latter expand of scope investigation can apply to both, narrowing to regions as well as extending it to other EU Member States. It is also worth considering the use of alternative indicators of the efficiency of individual groups of farms. For example, as part of the verification of the correctness of the results, the studied groups of agricultural producers were divided into subgroups conducting similar production directions. Preliminary analysis results that will be developed in subsequent articles shows that farmers specializing in animal production are characterized by a stronger increase in the value of the competitiveness index than it is in the case of farmers specializing in plant production. This happened despite the fact that they obtained lower production increases, and the smaller increase in the value of liabilities. Nevertheless, when directly comparing both groups of investors, it is noticeable that the differences in favor of those using support are much lower in the case of farmers conducting animal or mixed production than in the case of those focusing exclusively on plant production. These results, after more detailed processing and control, will be published in subsequent articles addressing the issue of changes in the efficiency of the conducted activity in the group of actively investing agricultural producers.

Finally, the restrictions of this analysis should be mentioned. First of all, the main emphasis was placed on indicating the effects of support for the widest possible group of agricultural producers. As a result of this approach, limited attention was paid to the diversity of results between smaller subpopulations of agricultural producers. Attention should also be paid to the geographical restriction. Agricultural producers from only one country were analyzed. Meanwhile, although there are a common CAP framework, some differences are noticed in detailed solutions relating to individual support instruments between European Union Member States. It should also be remembered that the entire investment was analyzed without their detailed division. It can be presumed that investments of different types brought benefits with a significantly different scale. For example, construction investments could be associated with benefits different from those obtained when buying harvesters or tractors. However, it may be a significant difficulty in conducting such an analysis to find an effective way to distinguish between data on various types of investments from existing databases.