Sustainable Employment Creation through the Polish Investment Zone in Lagging Regions

Abstract

The article aims to identify the firm-level effects of public support within the Polish Investment Zone (PIZ) on firms’ employment located in lagging regions of Poland, signalling insights into sustainable employment creation. Utilising a difference-in-differences framework and accounting for controls, including the firms’ size, age, leverage, other public aid programmes, subsidies, and firm-level fixed effects, our objective is to ascertain the average treatment effects (ATEs) of the programme, particularly concerning the creation of additional workplaces. To ensure the reliability of our findings, we conducted robustness checks utilising alternative econometric approaches and scrutinise changes in the duration of both the pre- and post-treatment periods. No significant ATEs were observed in the year of the treatment or in one to three years following treatment. This outcome remains robust to variations in the econometric approach, the set of variables considered, and alterations in the length of both pre- and post-treatment periods. This article contributes to an ongoing discussion marked by a lack of consensus regarding the effectiveness of special economic zones in fostering sustainable employment and reducing unemployment rates by presenting the effects of the PIZ and directing policy attention towards more qualitative aspects of created workplaces, fostering sustainable employment.

1. Introduction

Support dedicated to specific regions or provided to a selected group of economic entities in the form of geographically targeted tax exemptions has become a popular method of managing economic development initiatives and an integral part of regional development policy. The operation of privileged areas around the world is part of a broader discussion on the issue of regional differences and concepts of regional development, conducted in the light of many economic theories, including institutional economics, new economic geography (NEG), or evolutionary economic geography. Indeed, these instruments exemplify the so-called territorialisation of public interventions aimed at the development of specific areas, referred to in the literature as place-based policies [1,2,3,4].

Privileged areas were introduced by countries worldwide regardless of their socio-political system or level of economic development. However, each country’s economic level determined the specific goals they intended to serve. In most economies, zones were treated as a tool to stimulate economic growth [2,5], exports [6,7,8], regional development [9], and the inflow of additional investment [10]; knowledge and technology transfer were treated as factors for improving competitiveness [11]. In other economies, these policies mainly focused on economic development and revitalisation of less-developed areas [12,13]. The United Kingdom, the United States, and France, among others, have introduced geographically targeted tax incentives to promote economic development, entrepreneurship, job creation, and environmental improvement in the so-called ‘distressed areas’.

The establishment of special economic zones (SEZs) in Poland in the 1990s was a direct response to the negative effects of economic transformation. The systemic transformations revealed not only a clear differentiation in the level of socio-economic development of individual regions, but also a spatial concentration of the adverse effects of the reforms in the form of economic recession, impoverishment of society, and high levels of structural unemployment. For many years, the zones were assigned a role primarily as an instrument for easing labour market tensions. However, in view of the changing needs of the Polish economy, the deepening process of regional polarisation, and the saturation effect associated with the location of investments in economically well-developed areas, this instrument has been completely transformed.

Since 2018, under the new regulations, support in the form of state aid can be obtained throughout the country, although investment projects implemented in so-called problem regions with high unemployment are still given preference. The Polish Investment Zone (PIZ) accelerates economic growth in Poland, reduces unemployment, and stimulates economic development.

The PIZ addresses several dimensions of sustainability. In particular, it refers to sustainable economic development of the regions, sustainable job creation, and the promotion of economic activities with a low negative impact on the environment. The PIZ has the potential to have a significant impact on sustainable employment by creating new jobs, investing in employee skills, and promoting responsible business practices. New jobs are often created in less-developed regions, helping to reduce unemployment and increase labour force participation in these areas. By attracting investors within selected branches of economic activity, the PIZ also encourages the creation of skilled jobs. This in turn leads to the upskilling of the local workforce and an increase in average wage levels. It thus contributes to the creation of stable, better-paid jobs and long-term job security. In addition, PIZ-supported investments often include sustainability and social responsibility components. Investors are encouraged to adopt environmentally friendly practices and look after the welfare of local communities, which may include providing safe, healthy workplaces with decent working conditions.

The dynamic development and evolutionary nature of the solutions used in the framework of zoning policies around the world has resulted in a search for many years for empirical evidence of the impact of this instrument on various socio-economic levels. A particularly important area of interest is the assessment of the impact of zonal investments on improvements in the labour market (employment and unemployment rates). In various analyses, clear evidence has been sought to demonstrate that zones contribute to solving the problems of this market. However, it has not been found either in the extensive foreign literature on, for example, the operation of enterprise zones in the US [3,14] or in France [15], or on the ground of Polish research on SEZs [9,16,17,18].

The lack of consensus in the debate on the effectiveness of zones in stimulating employment or reducing unemployment is explained, among other things, by the variety of incentives used under the programme to create an appropriate investment climate, the socio-economic context in which the zones operate, or, finally, the adopted method of analysis. However, from the point of view of any economy, an unambiguous assessment of the effectiveness of this type of public intervention is critical. This is because it allows us to rationalise and thus increase the effectiveness of the designed programmes and to decide on their legitimacy and possible continuation.

Therefore, this article seeks an answer to the research question of the effects of support in the Polish Investment Zone on firm-level employment. The aim of this article strictly follows the above-formulated research question. It aims to identify the firm-level effects of public support within the Polish Investment Zone (PIZ) on firms’ employment, signalling insights into sustainable development and sustainable employment. Using a difference-in-differences framework and controlling for factors such as the firms’ size, capital efficiency, age, leverage, other public aid programmes, subsidies, and other firm-level fixed effects, we seek to discern the programme’s average treatment effects (ATEs), specifically in terms of additional workplaces. Alternative econometric approaches are employed for robustness checks, along with examinations of changes in the length of the pre- and post-treatment periods.

This article forms part of an extended discourse with no consensus concerning the efficacy of specialised economic zones in generating job opportunities and curtailing unemployment rates [19]. This article advances by showing novel aspects of firm-level effects of SEZ support within a recently introduced PIZ programme, which has not been evaluated to date. The programme encompasses a series of changes compared to its predecessor and other SEZ programmes around the world (e.g., quantitative and qualitative criteria, territory of operation, etc.). By controlling for a broader set of firm-level variables than previous research, including other sources of government aid and subsidies, we can better isolate the true effect of SEZ privileges on firm-level job creation. Most previous studies in Poland used aggregated data and therefore cannot capture the heterogeneity of individual firms. This article also shows the efficacy of public spending, i.e., creating employment within the programme and signalling potential strands of subsequent empirical trials, including the need to capture the heterogeneity of programme outcomes. The implications stemming from the obtained results form intriguing insights into sustainable employment in less-developed regions, particularly signalling the importance of quantitative aspects of created workplaces.

The remainder of this article is as follows: The next section depicts empirical contributions to the effects of zonal policies on reducing unemployment and employment creation at regional- and firm-levels. The following part of the article depicts the data and research methodology utilised in the counterfactual approach. In the stylised facts section, we present a series of characteristics on the research sample of SEZs. In the results and discussion section, we present the effects of the econometric inquiry, conduct robustness checks, and compare the findings with other empirical works. In the concluding section, we provide implications for economic policy and potential future strands of the research of zonal policies.

2. Literature Review

The scientific literature provides comprehensive analyses assessing zones’ effects and their significance for individual economies. In particular, SEZs have had a good chance of achieving their goals in countries such as India, China, and the United States. This is because these countries are pioneers in conducting zone policies and are examples of success stories of using this instrument to drive exports, attract foreign investment, or generate employment. On the other hand, the experience of zones in Africa suggests that these policies can also be costly and socially damaging [20,21,22,23,24].

Considering the array of research approaches addressing the impact of public interventions on labour market improvement in diverse economies and the varied effects observed in the analyses conducted, comparing them becomes nearly impossible. This fact underscores the challenge of generalising empirical evidence within this research area. In more-or-less methodologically sophisticated analyses of the operation of zones in India, China, the US, France, or on the grounds of Polish studies, one does not find conclusive evidence that zones contribute to stimulating jobs and reducing unemployment. This fact is explained by the variety of incentives used under the programme, the different socio-economic contexts in which the zones operate, and the research approach used to evaluate the impact of the interventions undertaken.

In India, the zones’ impact on employment at the national level is assessed as limited. However, they have contributed to employment generation at the regional level [25,26], as well as to changes in employment patterns in urban and rural areas [27]. Given that SEZs are scattered throughout the country, they primarily represent an employment opportunity for the unskilled and low-skilled [28]. In China, on the other hand, through the strengthening of economic agglomeration processes and the consequent migration of labour from the countryside to the cities [16,29], estimating the actual volume of employment and assessing the effects of China’s SEZ policies in this regard is extremely difficult. Nevertheless, Lu et al. [2] have made efforts to estimate that the programme’s net benefits in the employment domain are approximately 35%.

Much of the controversy over the effectiveness of such instruments has centred on enterprise zones (EPZs). These programmes were supposed to support economic development and encourage employment generation in underdeveloped regions of the US and France. However, the results of the studies conducted are mixed. On the one hand, enterprise zones are found to contribute to reducing the unemployment rate [14,30,31,32,33,34] or increasing employment [15,33,35,36,37,38,39]. On the other hand, many researchers do not find a significantly significant impact in this regard [3,40,41,42,43,44,45]. Moreover, analyses of the same US programme (EPZ) measured by the same method (PSM) lead to contradictory results [14,43]. In addition, several publications also highlight the issue of heterogeneity of effects across sectors/industries [38,39,46,47], depending on the size or status of the company (new, existing, disappearing) [38,46,48] and the location of the zone [49].

In the Polish literature, a longstanding and substantial debate exists concerning the efficacy of zone policies in stimulating job growth and alleviating unemployment. Similar to findings in other countries, the empirical results of these studies frequently yield conflicting conclusions. Analyses of the impact of SEZs on the economic performance of counties by Ciżkowicz et al. [50,51] indicate a positive impact of zones on employment. Nazarczuk and Cicha-Nazarczuk [52] confirmed the positive impact of the instrument on both employment and unemployment rate reduction, and this is true regardless of the modelling approach or the size of the group of counties studied. Ambroziak [53], as well as Ambroziak and Hartwell [9], conversely, concluded that the zoning policy reduced the unemployment rate only in the poorest and least-developed counties.

Analyses conducted at the municipal level are not so conclusive. A study by Jensen and Winiarczyk [29] and then Jensen [16], using advanced spatial econometric models, found a small but statistically significant and positive effect on municipal employment. In contrast, an evaluation by Trzciński et al. [54] found that SEZs had no effect on reducing unemployment levels in municipalities.

Indeed, to date, the employment effects of zones in Poland have been analysed almost exclusively in a regional dimension that considers counties or municipalities. The low availability of micro-data and the high cost of obtaining them have led to few attempts to identify these effects at the firm level [17,18].

Analyses conducted at the firm level constitute the fastest-growing area of current research. In the proposed topic, the utilisation of firm-level data has primarily focused on examining the effects of enterprise zones in the US [36,38,45,46,47,48] or France [15,39]. However, these analyses, most of which neglected to take into account the various dimensions of heterogeneity among firms, led to contradictory conclusions. Positive employment effects of such programmes have been estimated by O’Keefe [36], Billings [38], and Givord et al. [39]. In contrast, the lack of an employment effect was indicated by Bondonio and Greenbaum [48], Neumark and Kolko [45], and Lynch and Zax [46].

Cicha-Nazarczuk [17] underscores the imperative of acknowledging the heterogeneous nature of areas subject to intervention and the diversity among economic entities targeted by zone policies. This diversity plays a pivotal role in shaping the effects of special economic zones (SEZs) on the local labour market and employment growth within firms. The study further delineates the potentially varied impacts of SEZs at two distinct levels of data aggregation: counties and firms. Notably, positive and statistically significant effects on the labour market are identified. The investigation reveals that the magnitude of employment effects induced by SEZs exhibits variations contingent upon the contextual conditions under scrutiny. In turn, the analyses conducted by Dugiel et al. [18] lead to the conclusion that regional state aid to companies within SEZs has no effect on generating additional employment effects.

So far, analyses of employment effects carried out in Poland have dealt with the instrument of public intervention, which has been the special economic zones in operation since the 1990s. In recent years, however, this instrument has undergone a complete transformation. The enactment of a new law on 10 May 2018, aimed at supporting new investments, introduced the concept of assisting investors through regional investment aid across the entire country. However, due to the fact that investment projects implemented in so-called problem regions with high unemployment continue to be rewarded, it is necessary to determine to what extent the support provided under the Polish Investment Zone can contribute to improving the situation in local markets and whether the employment created in companies is the result of the real impact of the support provided.

3. Materials and Methods

The data utilised in this paper were obtained from the following: (i) SEZ-managing companies, supervising the issuance of decisions of support within the PIZ; (ii) InfoCredit, being the firms’ financial data provider; and (iii) EU regional aid intensity maps. Following Gal [55] and Cevik and Miryugin [56], the data prior to the counterfactual analysis were stripped of negative, missing, or zero observations in the following variables: total assets and employment. Next, the data were winsorised between the 1st and 99th percentile of their distribution to eliminate the impact of spurious observations (significant outliers) on the results following the approach of Öztekin [57].

The final dataset covers years 2015–2022 and encompasses ca. 320 companies, out of which 29 are the beneficiaries of the decision of support within the Polish Investment Zone. The control group firms were drawn from a large InfoCredit database as a random selection of companies operating within the same sectors as the treatment group also residing in the territory of Warmińsko-Mazurska SEZ and Suwalska SEZ but are not (and were not in the past) beneficiaries of public aid support within the zones. Since the data availability for micro and small companies is very limited, in the cases of such firms, the available data are scarce. Therefore, the research sample is limited and offers better comparability in medium- and bigger-sized companies. Knowing the restrictions above, the results should not be generalised over the whole sample of firms participating in the programme. Similarly, a company’s financial data delay further restricted the sample until 2022.

To grasp the preliminary effects of the Polish Investment Zone, this article adopts a counterfactual approach to obtain the average treatment effects (ATE) of the programme. The treatment variable, $D_{it}$, depicts whether in a certain year, t, company i had an active decision of support (Equation (1)):

$$D_{it}=\left\{\begin{array}{c}1ifcompanyiistreatedinyeart\\ 0intheothercase\end{array}\right\}$$

(1)

The treatment varies over time, as companies received public support in different years and entered the PIZ in different years of the study. Since the treatment variable is binary and time-varying, this article follows the approach by Cerulli and Ventura [58], being a generalisation of the difference-in-differences estimator, which is capable of estimating ATEs with a binary time-varying treatment and many pre- and post-treatment periods. The method, compared to other difference-in-differences frameworks, enables verification of the lack of the anticipatory effect and identifies potential delays in the programme’s effects.

The following equation was estimated for the employment variable $Y_{it}$:

$$Y_{it}={\mu }_{it}+{\beta }_{-2}{D}_{it-2}+{\beta }_{-1}{D}_{it-1}+{\beta }_0{D}_{it}+{\beta }_1{D}_{it+1}+{\beta }_2{D}_{it+2}+{\beta }_3{D}_{it+3}+{\gamma x}_{it}+u_{it}$$

(2)

where $x_{it}$ is a vector of control covariates with a corresponding conformable coefficient vector $\gamma$, ${\mu }_{it}$ depicts a fixed effect (in an FE specification), ${\beta }_0$–${\beta }_3$ measure the impact of treatment in the year of granting the decision of support and in one to three lags, whereas ${\beta }_{-2}-{\beta }_{-1}$ measure the impact of treatment in two and one leads, respectively. The number of leads and lags results from the number of years in the dataset for treated and non-treated firms. This paper uses the longest possible number of pre- and post-treatment periods, which does not significantly reduce the total number of firms in the estimates. However, other possible scenarios in this respect have been examined, ultimately leading to similar results.

To obtain more robust estimates, this paper utilises two consecutive approaches to obtain ATEs, through FE and OLS models, in each of the cases with robust to heteroskedasticity standard errors. The FE specification enables overcoming the selection-on-observables problem. The procedure yields causal estimates if the parallel trend assumption is met together with the lack of a so-called anticipatory effect. The latter indicates the effect of the current treatment on the pre-treatment levels [58], which is not the case if the pre-treatment estimates are not significant. The former tests whether the outcome variables of the treatment and control groups followed the same trend in the pre-treatment period. This paper uses two parallel tests to test this assumption: the lead and the time trend approach. Their insignificant values indicate that the parallel test is fulfilled.

The dependent variable represents companies’ employment (in levels, vacancies), as according to McConnell [59], the estimation of differences-in-difference in logs does not provide true difference-in-difference estimates but rather a proxy for relative growth rates among groups. Control variables depict the companies’ size, age, financial standings, capital efficiency, regional aid intensity, and control for other forms of public aid these enterprises may receive. The latter include subsidies and public aid support programmes. The complete list of variables utilised in this study is presented in

Table 1. Descriptive statistics of the variables utilised in this study.

Variable	Description	Source	Obs	Mean	SD	Min	Max
emp	Firms’ employment	Infocredit	1157	203.8	391.3	1.0	2970.0
D	Treatment variable (dummy)	SEZs	2014	0.1	0.3	0.0	1.0
year	Year	Infocredit	2014	2018.7	2.2	2015.0	2022.0
ROE	Return on equity [ratio]	Infocredit	1934	0.2	6.6	−191.0	129.7
subs	Subsidies [m PLN]	OCCP	2014	0.2	0.9	0.0	18.7
pai	Received public aid [m PLN]	OCCP	2014	1.7	9.2	0.0	231.7
rai_i	Regional public aid intensity [%]	EU RAM	2014	45.6	6.7	35.0	50.0
age	Companies’ age [years]	Infocredit	2013	16.0	15.6	0.0	147.0
lev	Total debt to total assets [ratio]	Infocredit	1916	1.1	10.8	0.0	463.5
size	Total assets [m PLN]	Infocredit	1945	73.4	215.1	0.0	2711.8
FOE	Foreign-owned entity (dummy)	Infocredit	2014	0.2	0.4	0.0	1.0

Source: own calculations. Information: SEZs—companies supervising special economic zones in Poland; EU RAM—the EU regional aid maps; OCCP—the Office of Competition and Consumer Protection; FOE—minimum threshold of 25% foreign capital is applied.

. The variables are not log-transformed, as this would exclude many observations due to frequent zero values for certain variables (e.g., subsidies, public aid, liabilities). We do not detect a multicollinearity problem, as the mean VIF is very low (1.19).

The most restrictive variable regarding the number of observations is employment due to the difficulty of obtaining such data at the firm level in Poland. Other variables mostly have a comparable number of observations.

4. Results

4.1. Research Sample

The Polish Investment Zone area covers the entire territory of Poland. This study presents the case of two of its sub-zones located in the northeastern part of the country. The region is a fairly homogeneous area with similar growth prospects, including two regional capitals: Bialystok (293k inhabitants) and Olsztyn (168k inhabitants). The Warminsko-Mazurska and Suwalska special economic zones are similar in many aspects, as they operate in a structurally similar economic environment.

As a consequence, the number of issued decisions of support are very similar between these zones. As the programme started in September of 2018, the number of decisions issued then is the lowest, reaching its peak in 2021, the year of extraordinary boost after the negative shock caused by the COVID-19 pandemic. Until 2022, within their scope of operation, firms received 196 support decisions, with a few benefiting from multiple support decisions (2–3). According to the SEZs’ managing data, 177 unique firms benefited from the programme until the end of 2022. Together, the area under the supervision of these SEZs represents approximately 10.3% of the total number of decisions issued between 2018 and 2022 in the PIZ.

4.2. Main Results

The results of difference-in-differences estimations provide rather vague effects of the PIZ programme on firms operating within it. The significant difference between the number of employed in treated firms and a control group of not-treated ones in the time of the study remains significant only in estimation 1 in

Table 2. The effects of Polish Investment Zone on companies’ employment.

Variables	(1)	(2)	(3)	(4)
D(t − 2)	25.47	14.52	6.45	6.40
	(16.25)	(17.01)	(12.87)	(13.19)
D(t − 1)	14.79	7.56	0.16	0.14
	(9.05)	(10.12)	(10.11)	(10.12)
D	23.89 **	16.06	8.13	8.10
	(12.05)	(12.62)	(9.65)	(9.71)
D(t + 1)	−2.31	−9.80	0.88	0.89
	(10.16)	(11.27)	(10.95)	(10.95)
D(t + 2)	14.04	5.95	−0.78	−0.80
	(12.29)	(12.69)	(10.03)	(10.05)
D(t + 3)	31.79	23.85	21.30	21.35
	(20.63)	(20.65)	(16.06)	(16.07)
ROE	1.46 ***	1.53 ***	1.73 ***	1.73 ***
	(0.55)	(0.49)	(0.35)	(0.36)
pai	0.84	0.71	−0.32	-0.32
	(1.34)	(1.29)	(1.11)	(1.11)
subs	8.05 **	7.68 **	9.18 *	9.18 *
	(3.78)	(3.77)	(5.05)	(5.05)
age		7.92 **	2.46	2.48
		(3.56)	(2.40)	(2.45)
size			0.74 ***	0.74 ***
			(0.21)	(0.21)
lev				−0.87
				(8.62)
Constant	215.12 ***	63.76	95.91 *	96.23 *
	(6.61)	(69.76)	(50.26)	(50.07)
Observations	579	579	579	578
No. of firms	168	168	168	168
F	5.049	6.142	11.95	11.09
R2 within	0.139	0.155	0.383	0.383
R2 between	0.104	0.0450	0.564	0.565
R2 overall	0.090	0.067	0.593	0.593
sigma_u	382.3	385.4	270.3	270.2
sigma_e	52.12	51.69	44.22	44.33
rho	0.982	0.982	0.974	0.974
P_test_l	0.325	0.511	0.882	0.889
P_test_t	0.424	0.378	0.242	0.241

Source: own estimates. The table presents estimates using the FE framework within the difference-in-differences approach. Robust standard errors are presented in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1. P_test_l—parallel trend test using the leads approach (p-value); P_test_t—parallel trend test using the time trend approach (p-value).

, where a limited scope of variables is utilised to control firm-level differences. The result indicates increased employment creation in PIZ enterprises by ca. 23.89 vacancies over not-treated entities, controlling for the overall efficiency of the firms’ equity (ROE) and other forms of received aid, like subsidies (subs) or public aid (pai). No delayed effects in the years following treatment can be grasped. It must be mentioned that utilisation of the FE framework crosses out some time invariant variables, like ownership status (FOE) and regional aid intensity (mostly stable over time with the exception of 2022), which are included in the OLS approach.

The introduction of new variables, like the firms’ age (estimation 2, Table 2) and especially their total assets (which are a proxy for firm size, estimation 3), greatly enhances the fitness of the models and takes out the significance from the ATEs observed in the first estimation. Generally, older and bigger firms employ more, similar to those with higher efficiency of equity and those receiving higher volumes of subsidies. The ratio of debt-to-assets, here, a proxy of the level of leverage, was not significant in explaining the differences in the firms’ employment.

The re-run estimations (available on request) showed some degree of heterogeneity in the results (in comparison to Table 2). As the dataset is not large, we were only able to incorporate a limited number of strata. Regarding age, the obtained results showed no significant ATEs (for all periods) for firms aged 10 and below. Conversely, more established firms (aged above 10) showed significant ATEs in the year of the treatment (34.4). Both estimations were run without controlling for the size of the company, as the effect disappeared when size was also controlled for.

In the second trial, we also found some heterogeneity in the significance of the ATEs when stratifying variable size (total assets). The obtained results lead to finding particular ATEs in the year of treatment for the following strata (in m PLN): <10 (no effect), 10–50 (significant positive ATE), 50–250 (no effect), 250+ (weakly significant positive effect). The ATEs above were only significant without additional controls for age. Once included, the significant ATEs vanished. Due to a limited number of firms in the dataset, we could not explore the issue in greater detail.

The robustness checks include the use of different frameworks to obtain difference-in-difference estimates, as well as playing with the length of the pre- and post-treatment periods.

Table 3. The effects of Polish Investment Zone on firms’ employment in different periods.

Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
D(t − 3)	11.28	11.69	12.11
	(16.19)	(15.22)	(16.60)
D(t − 2)	9.00	8.27	9.60	6.40	8.53	10.57
	(10.81)	(10.46)	(9.88)	(13.19)	(15.45)	(14.82)
D(t − 1)	5.99	8.00	5.90	0.14	0.52	−2.07	9.13	8.73	6.07
	(11.82)	(10.95)	(9.72)	(10.12)	(9.39)	(8.24)	(8.66)	(9.54)	(9.24)
D	2.90	−4.55	−7.51	8.10	9.76	10.03	5.03	3.75	3.78
	(10.56)	(10.79)	(11.49)	(9.71)	(9.74)	(10.25)	(8.07)	(7.64)	(8.83)
D(t + 1)	−7.12	−11.54	−12.04	0.89	−8.01	−11.27	7.98	8.00	5.76
	(11.91)	(12.81)	(20.06)	(10.95)	(11.53)	(16.07)	(10.51)	(11.16)	(11.81)
D(t + 2)	7.98	19.25		−0.80	5.38		4.12	−0.05
	(26.72)	(38.71)		(10.05)	(15.82)		(7.69)	(11.31)
D(t + 3)				21.35			3.15
				(16.07)			(9.82)
/some of the output omitted/
Obs	418	509	567	578	683	750	743	858	932
No. of firms	150	164	179	168	184	196	185	198	208
Pre	3	3	3	2	2	2	1	1	1
Post	3	2	1	3	2	1	3	2	1
F	19.79	3.517	3.257	11.09	4.149	3.497	11.25	10.36	10.13
R2	0.597	0.408	0.253	0.593	0.476	0.329	0.596	0.539	0.447
P_test_l	0.738	0.650	0.638	0.889	0.858	0.747	0.293	0.361	0.512

Source: own estimates. The table presents estimates using the FE approach within the difference-in-differences approach. Robust standard errors are presented in parentheses. P_test_l—parallel trend test using the leads approach (p-value). Pre and Post refer to the number of pre- and post-treatment periods used in the estimations.

shows results with different numbers of pre- and post-treatment periods, using the FE approach and the set of variables used in estimation 5 of Table 2. Some of the results are omitted for the sake of clarity. None of the scenarios (Table 3) led to significant ATEs being found in the year of treatment or in the subsequent periods.

In the OLS framework, one can observe overinflated, though still insignificant, ATEs in the treatment period, as well as in the pre- and post-treatment periods (

Table 4. The effects of Polish Investment Zone on firms’ employment—OLS estimates.

Variables	(1)	(2)	(3)	(4)	(5)
D(t − 2)	71.61	72.99	73.06	20.18	18.59
	(82.20)	(82.21)	(81.37)	(58.41)	(58.64)
D(t − 1)	−57.18	−56.79	−55.09	−29.27	−29.26
	(102.80)	(102.94)	(101.18)	(74.77)	(74.86)
D	62.51	62.75	50.40	56.09	56.00
	(108.29)	(108.44)	(106.97)	(67.37)	(67.47)
D(t + 1)	−80.72	−80.93	−87.67	−59.40	−59.31
	(119.63)	(119.80)	(117.75)	(57.90)	(58.00)
D(t + 2)	−19.32	−18.85	−18.14	24.69	24.63
	(110.30)	(110.38)	(107.87)	(58.13)	(58.18)
D(t + 3)	151.03	151.50	148.14	80.89	81.50
	(114.43)	(114.45)	(111.62)	(67.51)	(67.52)
ROE	−4.52	−4.38	−1.67	0.63	0.66
	(4.66)	(4.62)	(3.91)	(2.04)	(2.05)
rai_i	6.65 ***	6.84 ***	6.83 ***	3.74 ***	3.84 ***
	(1.84)	(1.77)	(1.84)	(1.27)	(1.30)
pai	7.08 *	7.11*	7.40 *	2.90	2.88
	(3.75)	(3.76)	(3.79)	(2.58)	(2.58)
subs	116.24 ***	115.83 ***	110.30 ***	60.76 ***	60.76 ***
	(29.94)	(30.19)	(29.87)	(21.73)	(21.76)
FOE		14.69
		(28.49)
age			4.85 ***	2.59 ***	2.58 ***
			(1.30)	(0.97)	(0.97)
size				1.02 ***	1.02 ***
				(0.15)	(0.15)
lev					−4.77
					(4.56)
Constant	−146.93 *	−159.43 **	−251.15 ***	−142.71 ***	−143.65 ***
	(74.87)	(71.88)	(76.11)	(51.26)	(51.39)
Observations	579	579	579	579	578
F	5.944	5.843	6.482	25.95	23.97
R2	0.200	0.201	0.226	0.625	0.625
P_test_l	0.682	0.671	0.663	0.923	0.926

Source: own estimates. The table presents estimates using OLS within the difference-in-differences approach with 2 pre-treatment and 3 post-treatment periods. Robust standard errors are presented in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1. P_test_l—parallel trend test using the leads approach (p-value).

). This suggests that factors other than the PIZ programme influence employment changes in firms. Some changes in the significance of the covariates and their coefficient values are noticeable when compared to the FE estimates. In particular, ROE has lost its significance in favour of regional aid intensity (rai_i) and the value of received public aid from other programmes (pai). Both have a positive effect on job creation, meaning that more jobs were created in firms that received higher volume of public aid and were located in less economically developed areas. The ownership status (FOE) has no statistically significant effect on employment changes in this case, while other variables retain their significance. The OLS estimates are less robust than the FE estimates due to the problem of selection-on-observables. Similarly to the FE estimates, changes in the length of the pre- and post-treatment periods do not change the significance of the treatment variable. Their results are available on request.

Both approaches suggest that there are no significant effects of the PIZ on job creation by firms relative to the control group. This observation remains robust to changes in the estimation framework and to variations in the length of the pre- and post-treatment periods.

5. Discussion

To the authors’ knowledge, this study is the first to evaluate the Polish Investment Zone, so direct comparisons between studies evaluating the same programme are not possible. The closest studies are firm-level evaluations of Poland’s former Special Economic Zone programme. Using an econometric approach similar to ours, Dugiel et al. [18] also found insignificant ATEs on firm-level effects of the former SEZ programme within a different period (2007–2017). In comparison to our study, they utilised a larger dataset, but controlled for a slightly different set of variables and did not test whether the results obtained were sensitive to different lengths of the pre- and post-treatment periods. They also did not try to find heterogenous effects of the programme.

Using kernel-based difference-in-differences combined with a propensity score matching technique, Cicha-Nazarczuk [17] found positive employment effects for a different time frame (2007–2012). However, the magnitude of these effects varies according to the contextual conditions analysed, namely, the structural characteristics of the intervention areas (counties) and the sector and subsection of economic activity. It has not been possible to establish conclusively whether firms with domestic and foreign capital generate different employment effects. As the author points out, the problem with access to firm-level data translated into a small number of firms in the counterfactual survey. This affected both the scope of the study and the verification methods adopted. Furthermore, the challenging time period of the analysis (2007–2012), which coincided with the global financial crisis and the sovereign debt crisis in the Eurozone, may have influenced the scale of employment effects achieved and resulted in a slight overestimation.

Our results fit into the long-lasting discussion on the effects of public aid programmes within zones worldwide. Similarly to other firm-level studies, mainly conducted in the US [3,40,41,42,43,44,45], we did not find statistically significant effects of the programme on job creation in firms.

Our results are also contrary to the ones established by O’Keefe [36], Billings [38], Givord et al. [39], and Givord et al. [15] for the US and France, who found positive employment effects, which can be assigned to zonal policies. The disparity may stem from different programmes, contextual factors, initial zone conditions, the scope of incentives offered, geographical context, zonal firm density, region type in which zones were established, the applied method, etc.

The discrepancy in the assessment of the impact of US or French enterprise zones may be attributed to a multitude of factors. Key sources of these differences, frequently highlighted in the literature, include the nature of the zone policies and the level of tax breaks offered in each state (in the US) or municipality (in France). The distinctive socio-economic context in which these zones operate is also a significant factor. Finally, the estimated magnitude of the effect is sensitive to the method or data analysis technique employed.

It is evident that the findings on the net impact of zone policies on national or regional labour markets are contradictory. This is despite the use of both counterfactual methods (difference-in-difference, propensity score matching) and complex econometric models. A further area of concern is the process of identifying geographical areas that accurately reflect the boundaries of the zones (tax jurisdictions, municipalities, censuses, or various geolocation techniques). This is particularly relevant in the context of counterfactual methods, where the way in which the control sample was selected for the treatment group and the identification of factors that would affect the accuracy of the matches is of great importance. Furthermore, discrepancies may arise from the geographical context, the density of firms in the zone, the types of regions in which the zones were established, etc.

This study has a few limitations, mostly originating from the scope of the research. Since the collected data encompass only a part of the Polish Investment Zone, the results cannot be easily generalised over the entire territory of the PIZ. In this respect, some regional or sectoral heterogeneity may affect the results obtained in less-developed areas of Poland in particular. Secondly, the difficulty in obtaining data on firms’ employment is a serious concern for researchers analysing different aspects of corporate performance in Poland, which limits the number of firm-level observations. Thirdly, the timespan of the conducted research was very stressful for entrepreneurs, while the world became more volatile, uncertain, complex, and ambiguous—a VUCA environment [60]. The effects of numerous shocks (e.g., the COVID-19 pandemic, Russian invasion of Ukraine, energy prices surging) could impact decisions postponing or aborting investment projects [61], including employment creation in firms, their internationalisation [62], etc.

With larger datasets, one could inspect additional layers of potentially uneven effects of the programme on firms receiving public aid, located in various region types, or operating within different economic sectors. These aspects should direct future research agendas, with a particular emphasis on the qualitative aspects of employment creation. This should include an in-depth analysis of the stability of employment, the type of workplaces, remuneration levels, the required competences, and other relevant factors.

6. Conclusions

This paper evaluates one of the aspects of the Polish Investment Zone operation, which is the employment creation of firms, operating within the two of its subzones and receiving public aid. The results provide no statistically significant effect combined with the average treatment effect in this regard on employment. Therefore, based on the effects of the conducted estimates, we cannot accredit additional employment generated in firms purely to the operation of the programme, but rather to differences among the firms. However, one cannot unarguably state, due to the limited research sample size, that the programme results are not heterogeneous due to the firms’ characteristics, sections of economic activity, the firms’ location, etc.

The lack of a significant quantitative impact of the programme on business employment (one of its objectives) may be seen by some as a failure of the programme in terms of increasing sustainable employment. However, it is important to consider other aspects that contribute to sustainable development and employment. These include the changes in the quality of jobs created in the zones, which may involve greater demand for highly skilled, specialised, and well-paid workers. These improvements contribute to sustainable employment by fostering a skilled workforce that supports long-term economic growth. By promoting the creation of higher quality jobs, the programme can stimulate positive structural change in the regions. This change helps regions to move towards the production of higher value-added products and services, thereby overcoming some of the structural barriers faced by economically depressed areas. These changes are crucial for sustainable development, as they increase the competitiveness of these regions.

The PIZ programme encourages firms to align their activities with regional smart specialisation strategies, thereby promoting greater local and regional embeddedness. This alignment not only supports regional economic goals, but also ensures that business activities are more sustainable and closely integrated with local development plans. Such integration is essential for sustainable development, as it promotes long-term regional growth and stability.

While the direct quantitative impact on employment numbers may be limited, the potential of the PIZ programme to improve the quality of jobs and regional economic structures suggests a potentially broader and more sustainable impact. By focusing on these aspects, the programme can support sustainable employment and contribute to the long-term sustainable development of lagging regions. As the situation in the Polish labour market improved, the quantitative aspects of job creation appear to have become less important (in comparison to the previous SEZ programme) in favour of other aspects, such as the creation of skilled, highly specialised, and well-paid workplaces, employment stability, supporting employees in acquiring education and professional qualifications, increasing firms’ B&R activity, local and regional embeddedness, the exploration of national and regional competitive advantages, cooperation with the local education sector, promoting economic activities with low negative impact on environment, etc. It is recommended that actions be taken to further increase regional firms’ embeddedness, stability of workplaces, and firms’ resilience, particularly in the context of the VUCA economy, in order to promote more sustainable regional economies.

Future research should continue to explore these qualitative benefits and the wider implications of the programme for sustainable economic growth. This could include identifying the quality of jobs created in PIZs, their level of technical sophistication, the skills and qualifications required, the level of salaries, and job stability. However, this type of research would require a questionnaire survey among firms operating in the PIZ. Another potential strand of the research could inspect the sustainable development of regions through the PIZ.