Wagner’s Law and Developmental Public Finances

Abstract

In the context of long-term fiscal sustainability challenges and a persistently uncertain geopolitical environment, the developmental role of public finances is gaining increasing prominence. A reallocation in the structure of government spending constitutes a key instrument for strengthening the developmental role of public finances and advancing the sustainable development goals. This study introduces a measurable indicator of the developmental role of public finances and examines the validity of Wagner’s law. Based on a sample of euro area countries, Granger causality panel vector error correction model estimates indicate there is a bidirectional causal relationship between general government expenditure and GDP per capita in the short-run, while the validity of Wagner’s law cannot be confirmed in the long-run. Productive government expenditure results suggest a weak indication of causality from productive government expenditure to GDP per capita. Effective fiscal policy should consider stabilisation needs, structural dynamics, and an appropriate institutional framework, while carefully managing the composition and sustainability of public spending and addressing developmental roles.

1. Introduction

Countries adopt diverse development policies to achieve sustainable economic growth and broader economic development. However, the effectiveness of public expenditure may remain limited, even when programmes are carefully designed. On the one hand, public spending depends on the availability of tax revenues [1,2]. On the other hand, it is constrained by expenditures related to the state’s traditional functions. Attempts to increase tax revenues through alternative tax structures may negatively affect economic growth and, consequently, hinder economic development. At the same time, excessive public spending on traditional state functions reduces the fiscal space available for development-oriented expenditure [3,4,5].

Reallocating general government expenditure is a key instrument for strengthening the developmental role of public finances and advancing sustainable development goals [6,7,8]. These SDGs include, for example, decent work and economic growth, reduced inequalities, quality education, good health and well-being, and industry, innovation, and infrastructure. Fiscal policy plays a central role in shaping economic development conditions. Through effective tax policy, governments can stimulate entrepreneurship, research and development, and investment in sustainable technologies. By adjusting public expenditure across the economic cycle, fiscal policy also promotes macroeconomic stability and supports long-term sustainable growth. Changes in the structure of public expenditure acquire a developmental dimension when resources are directed towards human capital formation, infrastructure development, and institutional improvement. Such reallocation may also reduce social inequalities, thereby enhancing social inclusion and stability. Moreover, changes in expenditure composition can reduce regional disparities and support sustainable development, including environmental and climate objectives. Green budgeting, environmental taxation, and investment in renewable energy are examples of fiscal instruments that facilitate the transition to a low-carbon circular economy.

The efficiency of government expenditure is closely linked to sustainability, as it determines the extent to which public resources generate long-term economic, social, and environmental outcomes. Efficient spending enables governments to achieve policy objectives, maintain fiscal stability, and reduce the risk of excessive deficits and public debt. It also supports productive investment in education, infrastructure, and innovation, thereby fostering sustainable economic growth. In contrast, inefficient spending leads to resource misallocation and limited long-term benefits. From a social perspective, efficient expenditure improves the accessibility and quality of public services, contributing to lower inequality and higher well-being. Inefficiency, by contrast, weakens service provision and reduces the effectiveness of social policies. Environmental sustainability is similarly affected, as efficient allocation supports environmentally sustainable investment, whereas inefficient spending may fail to address environmental challenges or reinforce unsustainable practices see, for example, refs. [9,10,11]. Overall, efficient government expenditure promotes long-term economic stability, social equity, and environmental protection.

With regard to the relationship between economic activity and public expenditure, two dominant yet opposing theories have emerged in the economic literature: Wagner’s law and the Keynesian hypothesis. The theories differ fundamentally in their definition of the functional relationship between economic growth and public expenditure. According to Wagner’s law, economic growth leads to higher public expenditure, whereas under the Keynesian hypothesis the direction of causality is the reverse see, for example, refs. [12,13].

Within the European context, this debate is particularly pertinent due to the historical evolution of welfare states, fiscal integration, and heterogeneous institutional structures across countries. European economies, especially in the post-war period, experienced sustained increases in public expenditure alongside rising income levels, a pattern broadly consistent with Wagnerian dynamics. As economies developed, demand for social protection, infrastructure, and regulatory frameworks expanded, reinforcing the idea that public spending grows as a share of gross domestic product with economic progress. This aligns with the broader interpretation of Wagner’s law that modernization and socio-economic complexity necessitate larger government involvement. At the same time, the Keynesian hypothesis has played a central role in European macroeconomic policy, particularly during crises. Fiscal interventions, such as countercyclical spending during recessions, reflect the Keynesian view that government expenditure can actively shape economic outcomes. In this framework, public spending has multiplier effects, influencing production, employment, and income. The empirical literature emphasises that public expenditure can act as a growth accelerating apparatus [14], especially when used to stabilise cyclical fluctuations or support aggregate demand. European fiscal responses to events such as the global financial crisis and the COVID-19 pandemic illustrate this logic, where expansionary fiscal policies were deployed to mitigate downturns and sustain economic activity see, for example, refs. [15,16,17].

Empirical tests of Wagner’s law in some studies see, for example, refs. [18,19] focus on the general form of the connection between the growth rate of gross domestic product and general government expenditure, but do not take into account the heterogeneity of individual public expenditure categories. Some studies address this limitation see, for example, refs. [20,21,22] by incorporating the economic and functional classifications of public expenditure, whereby the concept of productive public expenditure appears only indirectly. Our research fills this gap by directly incorporating productive public expenditure and introducing a measurable indicator of developmental role of public finances, through which the developmental role of public finances is explicitly integrated into the analysis. This represents an important extension of traditional models, as it provides insight into the structure of public expenditure and its development orientation.

The aim of this paper is threefold: first, to examine the validity of Wagner’s law in euro area countries; second, to test its validity when productive public expenditure is included; and third, to assess the implications for the long-term sustainability of public finances.

The paper makes the following contributions. First, we test the validity of Wagner’s law and examine the causal relationship between the growth rate of gross domestic product and general government expenditure in a dynamic panel data setting, utilising the panel vector error correction model approach. This methodology enables us to estimate both the short-term and long-term causal relationships between the growth rate of gross domestic product and general government expenditure. The panel data setting increases the reliability of our results. Second, in analysing the validity of Wagner’s law, we introduced an indicator of productive public expenditure to “measure” the developmental role of public finances. As noted above, economic growth is also influenced by the composition of public expenditure. Studies that examine the structure of public expenditure by economic classification see, for example, refs. [23,24] find that growth is positively associated primarily with capital expenditure and investment transfers. However, these effects depend on the allocation of capital expenditure and investment transfers and on whether public financing of investment crowds out private investment. To obtain a more comprehensive picture of the developmental role of public finances, the analysis incorporates an indicator of productive government expenditure, which includes expenditure on transport and communication, health, education, and research and development across all categories of the classification of the functions of government. Third, achieving an appropriate balance between economic, social and environmental development, while maintaining the sustainability of public finances, represents a central challenge in the developmental role of public finances. Addressing this challenge requires effective and transparent development planning to ensure the best value for public money, as the primary objective of adjusting the composition of public expenditure is to align it with strategic development priorities. In this context, this study contributes to more informed decision-making related to developmental role of public finances.

2. Literature Review

A central issue in the literature concerns the theoretical and econometric validity of Wagner’s law. One of the most influential critiques argues that many empirical confirmations of the law are based on spurious regressions arising from non-stationary time series data (Henrekson [25]). This critique highlights that without proper testing for unit roots and cointegration, observed correlations between income and public expenditure may not reflect meaningful long-run relationships. Earlier empirical contributions (see, for example, Afxentiou and Serletis [26], Courakis et al. [27]) similarly emphasise that results are highly sensitive to functional form and measurement choices, with different versions of Wagner’s Law yielding inconsistent outcomes. These methodological concerns have led to a more cautious interpretation of empirical findings and a shift toward more rigorous econometric techniques in later research.

Despite these concerns, there is evidence of support to Wagner’s law, particularly in developing and transitional economies (Abu-Eideh [28], Balkı and Göksu [29]). Empirical analysis of Egypt (Ghazy et al. [30]) finds a stable long-run relationship between economic growth and government expenditure, with causality running from income to public spending, consistent with Wagner’s hypothesis. Similar findings are reported in other developing country contexts, where cointegration and error correction models reveal that rising income levels lead to expansion in public sector activity (Magableh et al. [31]). Studies focusing on transition economies also provide partial support, suggesting that institutional development and restructuring processes during economic transformation contribute to increased public expenditure (Katrakilidi and Tsaliki [32]). Panel data analyses further reinforce these conclusions, indicating that Wagner’s law tends to hold in the long-run across groups of countries, even though short-run dynamics may vary (Magazzino [33]).

At the same time, a considerable number of studies either reject Wagner’s law or find only weak support for it. A recurring finding is the presence of bidirectional or reverse causality between government expenditure and economic growth. In such cases, the evidence aligns more closely with Keynesian theory, which posits that public spending can drive economic growth rather than merely respond to it. Empirical work employing causality tests (see, for example, Afonso and Alves [34]) demonstrates that the direction of causality is sensitive to model specification and estimation technique. In some developed economies, no stable long-run relationship is found, suggesting that public expenditure may stabilise or even decline relative to income as economies mature. Other studies (see, for example, Loizides and Vamvoukas [35]) report inconsistent results depending on the econometric approach used, further underscoring the lack of consensus.

Recent empirical studies, including those based on advanced econometric techniques such as cointegration and Granger causality tests, suggest that the relationship between public expenditure and growth in Europe often reflects a feedback mechanism rather than a unidirectional causality. The former implies that economic growth and public spending reinforce each other over time, with growth enabling higher revenues and spending, and spending in turn supporting further growth. Evidence from cross-country analyses indicates that both Wagnerian and Keynesian effects may coexist, depending on the time horizon, sectoral composition of expenditure, and institutional context. In the European context, the literature emphasises that the strong governance frameworks, fiscal discipline, and alignment with EU policy priorities emerge as critical factors that shape outcomes across countries. Consequently, the evidence highlights the need to move beyond traditional theoretical models toward a more context-sensitive understanding of fiscal policy in Europe (see, for example, Vigny [36]; Gülşen [37]).

Recent contributions to the literature extend the analysis of Wagner’s law by incorporating additional dimensions such as sustainability, institutional quality, and nonlinear dynamics. Research integrating environmental and governance variables shows that public expenditure growth is influenced not only by income but also by policy priorities related to sustainability and institutional effectiveness (Tao et al. [38]). Advances in econometric modelling, including nonlinear and threshold models (Arestis et al. [39]; Efthalitsidou et al. [40]), reveal that the relationship between income and government spending may not be linear, with Wagner’s law holding only beyond certain levels of economic development. A recent study published by Hossain et al. [14] provides further evidence that the validity of Wagner’s law varies across regions and time periods, with globalisation, fiscal frameworks, and institutional factors playing a significant role in shaping outcomes.

Methodological diversity is an important factor explaining the divergent findings in the literature. Studies differ in their choice of variables, such as total government expenditure, per capita expenditure, or expenditure as a share of GDP, as well as in their use of econometric techniques ranging from simple ordinary least squares to more sophisticated approaches such as vector error correction models, autoregressive distributed lag models, and panel cointegration methods. Research by Lamartina and Zaghini [23] emphasises that model specification and data properties critically influence empirical results. Authors emphasise that economic conditions and the level of development play an important role in testing Wagner’s law. In countries with low GDP per capita, characterised by fiscal imbalances and macroeconomic instability, empirical evidence more often supports Wagner’s law.

Based on the literature review, it can be concluded that existing research primarily focuses on testing the validity of Wagner’s law and examining causality using various econometric approaches, with some studies incorporating greater data granularity through disaggregated public expenditure categories. On the other hand, there is little evidence of taking into account the developmental role of public finances. This study addresses this gap by introducing a measurable indicator of the developmental role of public finances captured in a productive government expenditure, with the potential to reorient public expenditure towards more productive and development-oriented uses.

3. Methodology

As indicated in the literature on Wagner’s law, an increase in general government expenditure stems from per capita real gross domestic product growth. This means that general government expenditure is an increasing function of gross domestic product growth. According to Hossain et al. [14], this has two testable implications: one requires a long-run equilibrium association between the two variables, while the other requires a causality direction running from GDP growth to general government expenditure. The general form $IZD=f\left(BDD\right)$ of the relationship between general government expenditure (denoted as IZD) and GDP growth (denoted as BDD) variables is in the logarithmic transformation and in panel data setting (where i is cross-section/country subscript and t is time subscript) [41,42] with random error term μ presented as ${lnIZD}_{it}=ln\alpha +\beta {lnBDD}_{it}+{ln\mu }_{it}$. This equation serves as the basis for testing the long-term association among the variables and, consequently, for assessing the validity of Wagner’s law. To test the validity of Wagner’s law, we used the ratio of general government expenditure to GDP as the dependent variable and real gross domestic product per capita as the independent variable see, for example, refs. [14,43]. Additionally, to assess the robustness of the results, we replaced the ratio of general government expenditure to GDP with productive public expenditure (PID) as the dependent variable in the specification. In line with the methodology of Bańkowski et al. [44], productive government expenditure (PID) includes expenditure on transport and communication, health, education, and research and development across all categories of the classification of the functions of government (e.g., environmental protection, defence, public order and safety). These types of spending are linked to higher long-term growth and stronger productive capacity. Theoretical alignment is based on endogenous growth theory, where government expenditure is productive if it raises the productivity of labour and capital or increases total factor productivity. In this framework, education and health are especially central because they build human capital, which is a core driver of sustained economic growth. Education enhances skills, knowledge, and adaptability of the workforce, directly increasing labour productivity and the ability to adopt new technologies. Health expenditure improves workers’ physical and cognitive capacity, reduces absenteeism, and lengthens effective working lives, all of which raise output per worker. Healthier and better-educated individuals contribute to higher innovation, faster diffusion of technology, and greater efficiency across the economy. This makes such spending qualitatively different from non-productive expenditure, as it shifts the production frontier outward. Infrastructure and research and development also support productivity by improving efficiency and fostering innovation. These types of spending expand the economy’s productive capacity rather than only supporting consumption or redistribution. The classification of the functions of government is in use in all EU countries. Consistent with the literature see, for example, refs. [45,46], public debt, unemployment rate and trade openness were included as control variables to account for prevailing macroeconomic conditions, as well as population ages 65 and above to control for demographic structure. To capture structural breaks, such as global financial crisis (2008–2009), Eurozone debt crisis (2011–2012) and COVID-19 pandemic (2021–2023) a dummy crisis variable (DUMC) was included. All data expressed in nominal terms were price-deflated using the harmonised consumer price index (annual average index). The list of variables used, along with their descriptions and data sources, is presented in

Table 1. Description of variables and data sources.

Variable	Description	Data Source
IZD	Annual total general government expenditure, percentage of gross domestic product	Eurostat [gov_10a_exp] https://doi.org/10.2908/GOV_10A_EXP
BDD	Real gross domestic product per capita in EUR	Eurostat [tipsna40] https://doi.org/10.2908/TIPSNA40
PID	Annual productive government expenditure, percentage of gross domestic product	Eurostat [gov_10a_exp] https://doi.org/10.2908/GOV_10A_EXP
JD	Government consolidated gross debt, percentage of gross domestic product	Eurostat [gov_10dd_edpt1] https://doi.org/10.2908/GOV_10DD_EDPT1
BRE	Unemployment rate, percentage of active population	Ameco https://dashboard.tech.ec.europa.eu/qs_digit_dashboard_mt/public/sense/app/667e9fba-eea7-4d17-abf0-ef20f6994336/sheet/2f9f3ab7-09e9-4665-92d1-de9ead91fac7/state/analysis (accessed on 16 November 2025)
POP	Population ages 65 and above, percentage of total population	World Bank World Development Indicators https://databank.worldbank.org/reports.aspx?source=2&series=SP.POP.0014.TO.ZS&country=# (accessed on 20 April 2026)
TRADE	Trade openness, sum of exports and imports of goods and services, percentage of gross domestic product	World Bank World Development Indicators https://databank.worldbank.org/reports.aspx?source=2&series=SP.POP.0014.TO.ZS&country=# (accessed on 20 April 2026)

.

In all the models, the variables are in their natural logarithms. The parameter α denotes the intercept, while β is the slope coefficient, which measures the elasticity in a double log model (estimation was performed in Stata 13.0). For the validation of Wagner’s law, the estimated regression coefficient associated with real GDP per capita must be greater than or equal to zero. In addition, evidence of causality running from BDD to IZD, together with confirmation that the variables are stationary, further supports the validity of Wagner’s law [14].

The empirical analysis covers a panel of 20 euro area countries: Austria, Belgium, Croatia, Cyprus, Estonia, Finland, France, Germany, Greece, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, the Netherlands, Portugal, Slovakia, Slovenia, and Spain. Data for the period 1997–2023 were obtained from the Eurostat, AMECO and World Bank databases. Descriptive statistics for the panel 1997–2023 are presented in

Table 2. Descriptive statistics.

Variable	Obs	Mean	Std. Dev.	Min	Max
lnIZD	540	3.801473	0.1614558	3.025291	4.172848
lnBDD	540	10.07222	0.6098733	8.474286	11.52456
lnPID	540	2.638213	0.1804708	2.120761	3.020483
lnJD	540	3.951783	0.7598146	1.360977	5.344246
lnBR	540	2.083194	0.4623219	0.8329091	3.325036
lnTRADE	540	4.675547	0.4899664	3.674744	6.021453
lnPOP	540	2.813029	0.195992	2.278742	3.187128
DUMC	540	0.2592593	0.4386345	0	1

.

To assess the long-run (cointegrating) relationship between general government expenditure and the explanatory variables, the empirical analysis uses the panel vector error correction model (PVECM), which is valid when the variables are integrated of order one, I(1), and share one or more cointegrating relationships [47]. In the PVECM long-run equilibrium relationships and short-run dynamics among variables are simultaneously analysed. The model equations in the case of general government expenditure (similar in the case of productive government expenditure) are:

$$\begin{array}{cc}\hfill ∆{lnIZD}_{it}=c_{1i}+& {\sum }_{j=1}^q{\beta }_{11ij}{∆lnIZD}_{it-j}\hfill \\ \hfill & +{\sum }_{j=1}^q{\beta }_{12ij}{∆lnBDD}_{it-j}\hfill \\ \hfill & +{\sum }_{j=1}^q{\beta }_{13ij}{∆lnJD}_{it-j}\hfill \\ \hfill & +{\sum }_{j=1}^q{\beta }_{14ij}{∆lnBR}_{it-j}\hfill \\ \hfill & +{\sum }_{j=1}^q{\beta }_{15ij}{∆lnTRADE}_{it-j}+{\sum }_{j=1}^q{\beta }_{16ij}{∆lnPOP}_{it-j}\hfill \\ \hfill & +{\sum }_{j=1}^q{\beta }_{17ij}{∆DUMC}_{it-j}+{\beta }_{18i}{\epsilon }_{it-1}+{\mu }_{1it}\hfill \end{array}$$

$$\begin{array}{cc}\hfill ∆{lnBDD}_{it}=c_{2i}+& {\sum }_{j=1}^q{\beta }_{21ij}{∆lnIZD}_{it-j}\hfill \\ \hfill & +{\sum }_{j=1}^q{\beta }_{22ij}{∆lnBDD}_{it-j}\hfill \\ \hfill & +{\sum }_{j=1}^q{\beta }_{23ij}{∆lnJD}_{it-j}\hfill \\ \hfill & +{\sum }_{j=1}^q{\beta }_{24ij}{∆lnBR}_{it-j}\hfill \\ \hfill & +{\sum }_{j=1}^q{\beta }_{25ij}{∆lnTRADE}_{it-j}+{\sum }_{j=1}^q{\beta }_{26ij}{∆lnPOP}_{it-j}\hfill \\ \hfill & +{\sum }_{j=1}^q{\beta }_{27ij}{∆DUMC}_{it-j}+{\beta }_{28i}{\epsilon }_{it-1}+{\mu }_{2it}\hfill \end{array}$$

$$\begin{array}{cc}\hfill ∆{lnJD}_{it}=c_{3i}+& {\sum }_{j=1}^q{\beta }_{31ij}{∆lnIZD}_{it-j}\hfill \\ \hfill & +{\sum }_{j=1}^q{\beta }_{32ij}{∆lnBDD}_{it-j}\hfill \\ \hfill & +{\sum }_{j=1}^q{\beta }_{33ij}{∆lnJD}_{it-j}\hfill \\ \hfill & +{\sum }_{j=1}^q{\beta }_{34ij}{∆lnBR}_{it-j}\hfill \\ \hfill & +{\sum }_{j=1}^q{\beta }_{35ij}{∆lnTRADE}_{it-j}+{\sum }_{j=1}^q{\beta }_{36ij}{∆lnPOP}_{it-j}\hfill \\ \hfill & +{\sum }_{j=1}^q{\beta }_{37ij}{∆DUMC}_{it-j}+{\beta }_{38i}{\epsilon }_{it-1}+{\mu }_{3it}\hfill \end{array}$$

$$\begin{array}{cc}\hfill ∆{lnBR}_{it}=c_{4i}+& {\sum }_{j=1}^q{\beta }_{41ij}{∆lnIZD}_{it-j}\hfill \\ \hfill & +{\sum }_{j=1}^q{\beta }_{42ij}{∆lnBDD}_{it-j}\hfill \\ \hfill & +{\sum }_{j=1}^q{\beta }_{43ij}{∆lnJD}_{it-j}\hfill \\ \hfill & +{\sum }_{j=1}^q{\beta }_{44ij}{∆lnBR}_{it-j}\hfill \\ \hfill & +{\sum }_{j=1}^q{\beta }_{45ij}{∆lnTRADE}_{it-j}+{\sum }_{j=1}^q{\beta }_{46ij}{∆lnPOP}_{it-j}\hfill \\ \hfill & +{\sum }_{j=1}^q{\beta }_{47ij}{∆DUMC}_{it-j}+{\beta }_{48i}{\epsilon }_{it-1}+{\mu }_{4it}\hfill \end{array}$$

$$\begin{array}{cc}\hfill ∆{lnTRADE}_{it}=c_{5i}+& {\sum }_{j=1}^q{\beta }_{51ij}{∆lnIZD}_{it-j}\hfill \\ \hfill & +{\sum }_{j=1}^q{\beta }_{52ij}{∆lnBDD}_{it-j}\hfill \\ \hfill & +{\sum }_{j=1}^q{\beta }_{53ij}{∆lnJD}_{it-j}\hfill \\ \hfill & +{\sum }_{j=1}^q{\beta }_{54ij}{∆lnBR}_{it-j}\hfill \\ \hfill & +{\sum }_{j=1}^q{\beta }_{55ij}{∆lnTRADE}_{it-j}+{\sum }_{j=1}^q{\beta }_{56ij}{∆lnPOP}_{it-j}\hfill \\ \hfill & +{\sum }_{j=1}^q{\beta }_{57ij}{∆DUMC}_{it-j}+{\beta }_{58i}{\epsilon }_{it-1}+{\mu }_{5it}\hfill \end{array}$$

$$\begin{array}{cc}\hfill ∆{lnPOP}_{it}=c_{6i}+& {\sum }_{j=1}^q{\beta }_{61ij}{∆lnIZD}_{it-j}\hfill \\ \hfill & +{\sum }_{j=1}^q{\beta }_{62ij}{∆lnBDD}_{it-j}\hfill \\ \hfill & +{\sum }_{j=1}^q{\beta }_{63ij}{∆lnJD}_{it-j}\hfill \\ \hfill & +{\sum }_{j=1}^q{\beta }_{64ij}{∆lnBR}_{it-j}\hfill \\ \hfill & +{\sum }_{j=1}^q{\beta }_{65ij}{∆lnTRADE}_{it-j}+{\sum }_{j=1}^q{\beta }_{66ij}{∆lnPOP}_{it-j}\hfill \\ \hfill & +{\sum }_{j=1}^q{\beta }_{67ij}{∆DUMC}_{it-j}+{\beta }_{68i}{\epsilon }_{it-1}+{\mu }_{6it}\hfill \end{array}$$

$$\begin{array}{cc}\hfill ∆{DUMC}_{it}=c_{7i}+& {\sum }_{j=1}^q{\beta }_{71ij}{∆lnIZD}_{it-j}\hfill \\ \hfill & +{\sum }_{j=1}^q{\beta }_{72ij}{∆lnBDD}_{it-j}\hfill \\ \hfill & +{\sum }_{j=1}^q{\beta }_{73ij}{∆lnJD}_{it-j}\hfill \\ \hfill & +{\sum }_{j=1}^q{\beta }_{74ij}{∆lnBR}_{it-j}\hfill \\ \hfill & +{\sum }_{j=1}^q{\beta }_{75ij}{∆lnTRADE}_{it-j}+{\sum }_{j=1}^q{\beta }_{76ij}{∆lnPOP}_{it-j}\hfill \\ \hfill & +{\sum }_{j=1}^q{\beta }_{77ij}{∆DUMC}_{it-j}+{\beta }_{78i}{\epsilon }_{it-1}+{\mu }_{7it}\hfill \end{array}$$

∆ is the first difference operator, μ is the random error term, q is the lag value, and ε is the error correction term. Prior to applying PVECM methodology, panel unit root and cointegration tests were performed.

4. Results

Using the Im et al. [48] panel unit root test and the Fisher-type augmented Dickey–Fuller (ADF) test proposed by Choi [49], the condition for the variables being either I(1) or I(0) was checked. The null hypothesis for these tests is that all the panels contain a unit root.

Table 3. Panel unit root tests.

Variable		IPS Test	Fisher-Type ADF Test
lnIZD	Level 1st difference	−3.2769 (0.0005) −12.0939 (0.0000)	−3.4104 (0.0003) −14.0020 (0.0000)
lnBDD	Level 1st difference	−0.8304 (0.2032) −10.2872 (0.0000)	−0.3958 (0.3461) −11.1826 (0.0000)
lnPID	Level 1st difference	−2.9769 (0.0015) −11.6748 (0.0000)	−4.6105 (0.0000) −12.7753 (0.0000)
lnJD	Level 1st difference	1.8845 (0.9703) −8.7431 (0.0000)	0.1894 (0.5751) −8.1593 (0.0000)
lnBR	Level 1st difference	1.8750 (0.9696) −7.4204 (0.0000)	−2.5448 (0.1555) −10.2716 (0.0000)
lnTRADE	Level 1st difference	0.4674 (0.6799) −12.0884 (0.0000)	0.8626 (0.8058) −18.1728 (0.0000)
lnPOP	Level 1st difference	−0.5042 (0.3070) −4.2822 (0.0000)	5.4077 (1.0000) −3.4760 (0.0003)
DUMC	Level 1st difference	−5.1423 (0.0000) −13.6295 (0.0000)	−4.1746 (0.0000) −22.7793 (0.0000)

Notes: p-value in parenthesis.

points out unit root test results. According to both tests, variables lnIZD, lnPID and DUMC are stationary at level, the others are non-stationary at level. However, they become stationary after first differences, since the null hypothesis of a unit root is rejected at 1% level. The variables are integrated of order one, I(1) and are appropriate for further cointegration analysis.

To tackle the endogeneity concern, we applied instrumental variables two-stage least squares for panel data models (IV-2SLS), as in our case GDP can influence government expenditure and vice versa. According to the literature see, for a discussion, refs. [50,51] we included lagged GDP as an instrument, which proved to be suitable enough, since it is a statistically significant predictor in the first-stage model (in the case of general government expenditure as well as in the case of productive government expenditure with p-value = 0.000). Furthermore, the F-statistics is in both cases greater than ten (in the case of general government expenditure F = 53.55; in the case of productive government expenditure F = 10.59). Estimation results in

Table 4. IV-2SLS coefficients estimates.

	General Government Expenditure	Productive Government Expenditure
Variable	Coefficient	Coefficient
lnBDD	−0.1014435 (0.003)	0.0700417 (0.160)
lnJD	0.1313602 (0.000)	−0.0013732 (0.946)
lnBR	0.0420298 (0.002)	0.0392121 (0.056)
lnTRADE	−0.2665221 (0.000)	−0.119718 (0.019)
lnPOP	0.1488021 (0.003)	0.1789748 (0.017)
DUMC	0.0527004 (0.000)	0.0639695 (0.000)

Notes: p-value in parenthesis.

indicate similar results to those obtained using PVECM, confirming the robustness of results.

We proceeded with a panel cointegration test to confirm a long-run relationship between the dependent and explanatory variables, and applied the second-generation Westerlund’s cointegration test, Pedroni test and Kao test [52,53,54,55,56], to verify cointegration between the natural logarithm of ratio of general government expenditure to GDP (the ratio of productive government expenditure to GDP) and a set of explanatory variables. The results in

Table 5. Cointegration test results.

	General Government Expenditure	Productive Government Expenditure
Pedroni test	Stat.	Stat.
Panel v-Statistic	0.336 (0.3684)	−0.3298 (0.3708)
Panel rho-Statistic	0.5094 (0.3052)	1.023 (0.1531)
Panel PP-Statistic	−4.031 (0.0000)	−3.423 (0.0003)
Panel ADF-Statistic	−1.712 (0.0434)	−0.4742 (0.3177)
Group rho-Statistic	2.375 (0.0088)	2.539 (0.0056)
Group PP-Statistic	−3.482 (0.0002)	−3.982 (0.0000)
Group ADF-Statistic	−1.519 (0.0643)	0.08968 (0.4642)
Westerlund test	Stat.	Stat.
Gt	−0.083 (0.020)	1.314 (0.210)
Ga	5.730 (0.980)	5.516 (0.960)
Pt	−0.112 (0.050)	0.823 (0.180)
Pa	3.435 (0.860)	3.251 (0.490)
Kao test	Stat.	Stat.
Modified Dickey–Fuller t	−3.2213 (0.0006)	−4.3135 (0.0000)
Dickey–Fuller t	−2.4460 (0.0072)	−4.2307 (0.0000)
Augmented Dickey–Fuller t	−1.7966 (0.0362)	−5.4984 (0.0000)
Unadjusted modified Dickey–Fuller t	−4.1371 (0.0000)	−4.4040 (0.0000)
Unadjusted Dickey–Fuller t	−2.8322 (0.0023)	−4.2624 (0.0000)

Notes: p-value in parenthesis.

present the test statistics and the significance levels in parenthesis for the rejection of the null hypothesis of no cointegration. In the case of the Westerlund test, the panel as a whole were (Pt and Pa) and the mean-group statistics were (Gt and Ga). The rejection of the null hypothesis for the Pt and Pa test statistics should be taken as evidence of cointegration for the panel as a whole, whereas the rejection of the null hypothesis for the Gt and Ga test statistics should be taken as evidence of cointegration in at least one of the cross-sectional units [52,53]. We performed tests for cointegration relationship, taking into account all the explanatory variables. The results of the Westerlund test show, that for the general government expenditure there is evidence of cointegration relationship, whereas for the productive government expenditure the null of no cointegration cannot be rejected. Furthermore, the results of the Pedroni test show evidence of cointegration relationship among the variables in the case of general government expenditure as well as in the case of productive government expenditure. To reinforce these findings, the Kao test was performed, where the null of no cointegration is rejected at the 5% level. Although in the case of productive government expenditure the evidence is not uniformly strong across all statistics, the overall balance of results indicates the existence of a stable long-run equilibrium relationship among the variables. Failure to reject the null hypothesis of no cointegration in the case of the Westerlund test does not necessarily invalidate the existence of a long-run relationship. When combined with the significant Pedroni PP statistics and the strongly significant Kao results, the overall evidence still supports the presence of panel cointegration. These results suggest a stable and long-run equilibrium relationship among the dependent and explanatory variables.

The Bai-Perron test for multiple structural breaks (

Table 6. Structural break results.

Test	Statistic	Prob.	Interpretation
Bai-Perron UDmax	13.42	Exceeds 1%, 5%, and 10% critical values	Multiple breaks detected
Sequential test (F(1|0))	13.42		First break supported
Sequential test (F(2|1))	11.33		Second break supported
Sequential test (F(3|2))	10.00		Third break supported
Breakpoint 1	2002		95% CI: 2001–2003
Breakpoint 2	2008		95% CI: 2007–2009
Breakpoint 3	2014		95% CI: 2013–2015

) rejects the null hypothesis of no breaks (UDmax = 13.42, p < 0.01), indicating the presence of structural breaks in the panel relationship. The sequential breakpoint tests further suggest three structural breaks, in 2002, 2008 and 2014, with confidence intervals of 2001–2003, 2007–2009 and 2013–2015, respectively.

Additionally, we checked pre- and post-break effects by estimating separate regimes using dummy interactions (

Table 7. Pre- and post-break effects.

Regime	Effect of GDP per Capita	Prob.	Interpretation
Regime 1 (before first break)	−0.3045618	0.000	Baseline
Regime 2 (break year 2002)	−0.2607501	0.004	Significant weaker negative effect
Regime 3 (break year 2008)	−0.3085016	0.812	Insignificant
Regime 4 (break year 2014)	−0.3416962	0.023	Significant stronger negative effect

). The baseline regime estimates indicate a statistically significant negative relationship between GDP per capita and general government expenditures. Structural-break interactions reveal that this relationship changes across regimes identified by the Bai-Perron procedure. In regime 2, the negative effect weakens significantly, while in regime 4 the negative relationship becomes significantly stronger relative to the baseline period. No statistically significant slope change is observed in regime 3. The breakpoint period around 2002 coincides with economic downturn in the euro area and dot-com bubble; 2008 corresponds to global financial crisis, while 2014 corresponds to Eurozone debt crisis.

The presence of cross-sectional independence was tested using the Pesaran test [57]. The null hypothesis of cross-sectional independence was rejected at 1% level in the case of general government expenditure and in the case of productive government expenditure (p-value = 0.0000), indicating significant heteroskedasticity across panel units. Therefore, clustered standard errors were included in PVECM estimation.

A long-term cointegration coefficients assessment was carried out by a PVECM fixed effects estimator for general government expenditure and productive government expenditure (

Table 8. Long-run estimates (general government expenditure).

Variable	Coefficient	Std. Error	Prob.
lnBDD	−0.1765603	0.0298724	0.000
lnJD	0.1320885	0.0134522	0.000
lnBR	0.0297136	0.0133834	0.027
lnTRADE	−0.2279974	0.0328363	0.000
lnPOP	0.1772425	0.0490939	0.000
DUMC	0.0546281	0.0074382	0.000

and

Table 9. Long-run estimates (productive government expenditure).

Variable	Coefficient	Std. Error	Prob.
lnBDD	0.0166003	0.0451241	0.713
lnJD	−0.0046361	0.0203204	0.820
lnBR	0.0276224	0.0202164	0.172
lnTRADE	−0.0768993	0.0496012	0.122
lnPOP	0.2437761	0.0741594	0.001
DUMC	0.065151	0.0112358	0.000

). Based on the results from Table 8 in the long-run there is a cointegration relation among the general government expenditure, GDP per capita, public debt, unemployment, trade openness, population ages 65 and above and dummy crisis variable. All the estimated coefficients are statistically significant at a 5% level. As for the GDP per capita, this negatively influences the general government expenditure, as well as trade openness.

The results from Table 9 point out that in the long-run there is a cointegration relation among the productive government expenditure, population ages 65 and above and the dummy crisis variable. All the other estimated coefficients are not statistically significant.

The presence of a cointegration relationship among variables necessitates that there is causality in at least one direction. Causality connections between variables were evaluated by the Granger causality test [58] based on PVECM framework. The Granger causality test investigates the short-run, long-run and joint causal relationship among the variables. For the long-run causal relationship the error-correction term ECT (in

Table 10. Granger causality (general government expenditure).

Dependent Variable: d(lnIZD)	Chi-sq. (p-Value)
d(lnBDD)	11.79 (0.0028)
d(lnJD)	4.20 (0.0546)
d(lnBR)	0.95 (0.3431)
d(lnTRADE)	0.01 (0.9382)
d(lnPOP)	5.14 (0.0353)
d(DUMC)	19.96 (0.0003)
All	11.49 (0.0000)
ECT	−0.1476071 (0.001)
Dependent variable: d(lnBDD)
d(lnIZD)	21.60 (0.0002)
d(lnJD)	0.05 (0.8282)
d(lnBR)	0.07 (0.7891)
d(lnTRADE)	0.01 (0.9278)
d(lnPOP)	8.82 (0.0079)
d(DUMC)	24.49 (0.0000)
All	22.19 (0.0000)
ECT	−0.0753314 (0.019)
Dependent variable: d(lnJD)
d(lnBDD)	2.03 (0.1708)
d(lnIZD)	1.42 (0.2483)
d(lnBR)	0.35 (0.5621)
d(lnTRADE)	1.77 (0.1990)
d(lnPOP)	4.07 (0.0579)
d(DUMC)	17.78 (0.0005)
All	10.77 (0.0000)
ECT	0.3940529 (0.000)
Dependent variable: d(lnBR)
d(lnBDD)	8.06 (0.0105)
d(lnJD)	0.02 (0.8980)
d(lnIZD)	0.01 (0.9248)
d(lnTRADE)	2.22 (0.1525)
d(lnPOP)	7.84 (0.0114)
d(DUMC)	28.73 (0.0000)
All	14.72 (0.0000)
ECT	0.2675828 (0.009)
Dependent variable: d(lnTRADE)
d(lnBDD)	4.66 (0.0439)
d(lnJD)	4.97 (0.0381)
d(lnBR)	25.77 (0.0001)
d(lnIZD)	0.60 (0.4467)
d(lnPOP)	2.85 (0.1077)
d(DUMC)	31.19 (0.0000)
All	23.00 (0.0000)
ECT	−0.0796836 (0.101)
Dependent variable: d(lnPOP)
d(lnBDD)	5.69 (0.0276)
d(lnJD)	3.95 (0.0614)
d(lnBR)	2.80 (0.1105)
d(lnTRADE)	5.29 (0.0330)
d(lnIZD)	0.61 (0.4445)
d(DUMC)	1.57 (0.2260)
All	10.75 (0.0000)
ECT	0.0009933 (0.812)
Dependent variable: d(DUMC)
d(lnBDD)	5.96 (0.0246)
d(lnJD)	3.07 (0.0961)
d(lnBR)	7.25 (0.0144)
d(lnTRADE)	27.91 (0.0000)
d(lnPOP)	20.68 (0.0002)
d(lnIZD)	4.65 (0.0440)
All	35.06 (0.0000)
ECT	1.54846 (0.013)

Notes: p-value in parenthesis.

and

Table 11. Granger causality (productive government expenditure).

Dependent Variable: d(lnPID)	Chi-sq. (p-Value)
d(lnBDD)	2.19 (0.1550)
d(lnJD)	2.77 (0.1125)
d(lnBR)	0.70 (0.4123)
d(lnTRADE)	0.01 (0.9813)
d(lnPOP)	3.11 (0.0937)
d(DUMC)	0.22 (0.6417)
All	2.39 (0.0681)
ECT	−0.2496287 (0.000)
Dependent variable: d(lnBDD)
d(lnPID)	3.19 (0.0900)
d(lnJD)	0.03 (0.8549)
d(lnBR)	0.01 (0.9170)
d(lnTRADE)	1.70 (0.2076)
d(lnPOP)	6.64 (0.0185)
d(DUMC)	33.38 (0.0000)
All	13.46 (0.0000)
ECT	0.0068591 (0.688)
Dependent variable: d(lnJD)
d(lnBDD)	2.10 (0.1635)
d(lnPID)	0.02 (0.8798)
d(lnBR)	0.70 (0.4139)
d(lnTRADE)	0.34 (0.5672)
d(lnPOP)	3.14 (0.0924)
d(DUMC)	15.82 (0.0008)
All	11.43 (0.0000)
ECT	0.0042985 (0.939)
Dependent variable: d(lnBR)
d(lnBDD)	6.79 (0.0173)
d(lnJD)	0.01 (0.9809)
d(lnPID)	0.49 (0.4932)
d(lnTRADE)	1.15 (0.2977)
d(lnPOP)	8.85 (0.0078)
d(DUMC)	29.21 (0.0000)
All	10.39 (0.0000)
ECT	0.0192349 (0.712)
Dependent variable: d(lnTRADE)
d(lnBDD)	3.88 (0.0637)
d(lnJD)	5.49 (0.0301)
d(lnBR)	22.06 (0.0002)
d(lnPID)	2.31 (0.1449)
d(lnPOP)	1.89 (0.1847)
d(DUMC)	20.30 (0.0002)
All	17.00 (0.0000)
ECT	0.0230368 (0.457)
Dependent variable: d(lnPOP)
d(lnBDD)	4.19 (0.0547)
d(lnJD)	3.26 (0.0868)
d(lnBR)	2.43 (0.1353)
d(lnTRADE)	6.83 (0.0171)
d(lnPID)	0.01 (0.9715)
d(DUMC)	1.83 (0.1915)
All	6.03 (0.0011)
ECT	0.0012767 (0.592)
Dependent variable: d(DUMC)
d(lnBDD)	3.72 (0.0689)
d(lnJD)	4.50 (0.0472)
d(lnBR)	6.71 (0.0180)
d(lnTRADE)	23.79 (0.0001)
d(lnPOP)	21.24 (0.0002)
d(lnPID)	6.93 (0.0164)
All	27.28 (0.0000)
ECT	0.7564395 (0.002)

Notes: p-value in parenthesis.

the coefficient for ECT is reported with p-values) has to be negative and significant. Based on the results from Table 10 this can be confirmed in the case of general government expenditure as a dependent variable and GDP per capita as a dependent variable. For the dependent variable general government expenditure, GDP per capita, population ages 65 and above, dummy crisis variable and public debt show a statistically significant Granger causal effect, suggesting short-run causality from these variables to general government expenditure. In the case of GDP per capita as a dependent variable, general government expenditure, population ages 65 and above and dummy crisis variable exhibit statistically significant Granger causal effect at 5% level, with p-values of 0.0002, 0.0079 and 0.0000, respectively. Based on these results it can be concluded, that there is a bidirectional causal relationship between general government expenditure and GDP per capita in the short-run, while the estimated long-run equilibrium relationship between GDP per capita and general government expenditure is negative; therefore, we cannot confirm the validity of Wager’s law in the long-run. For the dependent variable trade openness, GDP per capita, public debt, unemployment and the dummy crisis variable demonstrate statistically significant Granger causality at the 5% level, implying that changes in the variables mentioned contribute to the short-run variations in trade openness. Structural shocks, captured by dummy crisis variable, are Granger caused in the short-run by all the other variables included, indicating that GDP per capita, unemployment, trade openness, population ages 65 and above, general government expenditure at the 5% level and public debt at the 1% level contribute to the short-run variations in structural shocks, captured by the dummy crisis variable.

The results, presented in Table 11, indicate that for the dependent variable productive government expenditure, none of the independent variables (GDP per capita, public debt, unemployment, trade openness, population ages 65 and above, dummy crisis variable) demonstrate statistically significant Granger causal effect, suggesting no short-run causality from these variables to productive government expenditure. In the case of GDP per capita as the dependent variable, only population ages 65 and above and the dummy crisis variable demonstrate statistically significant Granger causality at the 5% level, with p-values of 0.0185 and 0.0000, respectively, whereas productive government expenditure shows an effect with a p-value of 0.0900, suggesting a weak indication of short-run causality from productive government expenditure to GDP per capita. For the dependent variable unemployment, the independent variables GDP per capita, population ages 65 and above and the dummy crisis variable, show a statistically significant Granger causal effect at the 5% level, implying that changes in these variables contribute to short-run variations in unemployment. For population ages 65 and above, trade openness Granger causes population ages 65 and above at the 5% level, whereas the opposite cannot be confirmed. For the dummy crisis variable, all the independent variables included (GDP per capita, public debt, unemployment, population ages 65 and above, trade openness, productive government expenditure) show a reasonably strong indication of short-run causality.

5. Discussion

Based on the results of our study, there is a bidirectional causal relationship between general government expenditure and GDP per capita for the sample of euro area countries for the period 1997–2023 in the short-run, while the validity of Wagner’s law cannot be confirmed in the long-run. For productive government expenditure, results suggest a weak indication of short-run causality from productive government expenditure to GDP per capita.

As noted in the literature [43,59,60], public finance performs three basic functions: allocation, stabilisation, and redistribution. The allocation function concerns the correction of market failures in the functioning of individual markets. Its fundamental objective is to increase the volume and efficiency of the use of production factors and to improve their allocation by ensuring that production factors are employed in those economic sectors or activities where their contribution is greatest. The stabilisation function refers to the regulation or reduction in fluctuations in overall economic activity and is therefore also referred to as countercyclical policy (expansionary policy stimulates economic growth during a recession, while restrictive policy slows it down during periods of economic overheating). The redistributive function refers to the correction of the market-based distribution of income.

From a dynamic perspective, the validity of Wagner’s law is linked to the evolution of the fundamental functions of public finances, which adjust over time to structural changes in the economy and society. In line with stabilisation-oriented fiscal policy, the government should increase its expenditure during a recession. This generates additional demand, which can contribute to a revival of economic activity. The effect is stronger if the higher expenditure is financed through borrowing rather than through new taxes. Investment expenditure is expected to have the greatest impact on economic recovery, although in practice governments often increase other types of spending instead of investment (for example, job retention subsidies). In addition to raising general government expenditure, the government can also stimulate demand by reducing certain types of taxes, thereby encouraging an increase in other forms of expenditure. During an expansionary phase, high demand relative to available capacity generates rising inflation. The government restrains excessive demand by reducing general government expenditure or increasing taxes. Firms operate at virtually full capacity utilisation. To meet strong demand, they hire additional labour. Given tight labour market conditions, they can recruit additional workers only by offering higher wages. Higher wages, however, raise firms’ costs, which, in a context of strong demand, can be passed on into higher final prices, thereby triggering inflation. The stabilisation function of public finances gains importance particularly in the context of macroeconomic fluctuations, financial crises and asymmetric shocks. Through fiscal policy, the government influences aggregate demand, employment and price stability. In modern economies, which are increasingly exposed to global risks, the scope of stabilisation measures generally expands, leading to higher public expenditure and public debt.

The allocative function of public finances corrects market failures. It is therefore important that the method of financing used does not create new market distortions. From the allocative function perspective, a neutral tax system would be optimal—that is, one that interferes as little as possible with the economic decisions of taxed entities and thereby generates the smallest possible deadweight welfare loss. In addition to the structure and level of taxation, the allocative function of public finances is realised through the level of general government expenditure and its allocation across different purposes. The allocative function thus relates to the provision of public goods and the correction of market failures, such as externalities and monopolistic structures. With economic development and technological progress, the complexity of economic relations increases, leading to a greater need for public services in areas such as infrastructure, education, research and development, and environmental protection.

The redistributive function of public finances, on the other hand, corrects the distribution of income that would arise solely from the free market operation. The redistributive function is associated with reducing income and social inequalities and with ensuring social security. With rising incomes, urbanisation and demographic changes, such as population ageing, expectations regarding social protection, pension systems and healthcare services increase. Consequently, government’s role in income redistribution strengthens, which over the long term contributes to higher general government expenditure.

The aforementioned functions of public finances are dynamically interrelated and influence both the level and the structure of general government expenditure. Their evolution over time suggests the validity of Wagner’s law. The empirical results reveal a statistically significant negative long-run relationship between GDP per capita and general government expenditure, indicating that Wagner’s law is not supported in the long-run. However, the Granger causality results show that GDP per capita positively and significantly affects general government expenditure, while general government expenditure also positively and significantly affects GDP per capita in the short-run. Productive government expenditure results suggest a weak indication of causality from productive government expenditure to GDP per capita and not vice versa see, for example, refs. [61,62].

The latter may also indicate the operation of performance-based budgeting and a tendency towards evidence-based policy making. Development policies comprise a set of development programmes and measures that are interconnected to ensure an effective attainment of the defined development objectives, while countries must also take into account internationally agreed commitments. Development planning and budget planning, within the broader public finance framework, are subject to continuous refinement and upgrading. The successful implementation of development strategies requires effective and transparent development planning, ensuring the best value for public money, as the fundamental purpose is to align general government expenditure with the country’s development priorities [63].

The role of government spending composition emerges as an important factor as not all public expenditures have equal growth effects. Productive government spending, particularly in education, health, green transition and innovation, tends to support Keynesian growth effects, whereas inefficient or consumption-oriented expenditure undermines growth performance [64]. Furthermore, sustainability and fiscal constraints play an important role. Some studies [65] argue that while Keynesian expansionary spending may generate short-term growth, it can lead to long-term inefficiencies or debt burdens if not matched by revenue growth or productivity gains. This introduces a critical limitation to the Keynesian framework in contemporary contexts characterised by high public debt.

A recurring finding across multiple studies is the presence of bidirectional causality. This suggests a feedback loop: economic growth increases fiscal capacity and public demand (Wagner), while government spending simultaneously stimulates growth (Keynes). However, this bidirectional causality is not universal. Some country-specific analyses, especially those focusing on fiscally constrained or institutionally weak economies, find unidirectional causality from spending to growth, supporting Keynesian view [29]. Others identify the reverse, particularly in more mature economies, where growth precedes expenditure expansion, consistent with Wagner’s law [66].

The Hausman test supported dynamic fixed effects estimator suggesting that the long-run coefficients are homogeneous across the euro area countries included in the sample. In other words, the long-run relationship between the variables is statistically similar across member states, implying that economic variables adjust toward a common equilibrium path in the long term. This finding is consistent with the high degree of economic and monetary integration within the euro area, where common monetary policy, integrated financial markets, and coordinated macroeconomic frameworks may contribute to similar long-run dynamics across countries. Additionally, fiscal coordination mechanisms associated with the Maastricht Treaty, the Stability and Growth Pact, the European Semester, and post-crisis reforms have likely contributed to more similar fiscal behaviour across member states imposing common fiscal discipline and strengthened budgetary surveillance.

Furthermore, the results indicate limited parameter heterogeneity across euro area countries, at least with respect to the estimated long-run coefficients. Cross-country heterogeneity in the long-run effects is relatively weak, supporting the assumption that euro area economies respond in a broadly similar manner to changes in the explanatory variables over the long-run.

Nevertheless, the homogeneity assumption should be interpreted cautiously as some degree of heterogeneity may still exist in the short-run due to country-specific institutional structures, fiscal conditions, labour market rigidities, or asymmetric shocks. Therefore, while the long-run equilibrium relationships appear homogeneous, short-run adjustment dynamics may continue to differ across euro area member states.

Economic conditions and increasing demands from taxpayers for improved efficiency, quality and performance of the public sector are compelling many European countries to modernise their public systems, with budgetary developments representing a key pillar of policy development and accountability. The introduction of performance-based budgeting facilitates a clearer understanding of how funds are used. Performance-based budgeting involves the systematic use of information on effects and results to support the implementation of government programmes and decision-making on resource allocation. Performance-based budgeting is a strategy aimed at enhancing transparency and accountability, as well as ensuring alignment between taxpayers’ expectations and government provision. General government expenditure should support society in achieving its development objectives and contribute to social protection. Public order and safety are necessary for a stable institutional environment. General government expenditure on social protection, healthcare and recreation supports labour market participation and social development; housing and community amenities (for example, water supply) are a prerequisite for adequate living standards, while spending on environmental protection underpins sustainable development. Even where the objective is to increase social welfare, general government expenditure may not necessarily be allocated in the most efficient manner. Nordhaus [67] highlighted the problem of the political business cycle, in which policymakers treat policy decisions primarily as a means of pursuing their private objectives.

In addition, the paper has the following implications. The sustainability and developmental role of public finances are questionable under current uncertain geopolitical conditions, due to increasing demands and pressures to raise public expenditure on defence. Uncertain geopolitical conditions have posed the greatest test to the European security architecture since the Second World War [68]. Restoring European defence capabilities and increasing the production capacity of the European defence industry will require substantial public and private investment over an extended period [69]. In its ReArm Europe plan, the European Commission identified the coordinated activation of the national escape clause of the Stability and Growth Pact as one of five pillars for increasing European defence spending. It proposes the coordinated activation of the national escape clause by all Member States to unlock additional flexibility for higher defence expenditure. The delimiting of flexibility in the case of activating the national escape clause relates to defining the scope of the additional expenditure, that can benefit from flexibility, as well as its overall amount and duration. The flexibility would cover an increase in total defence expenditure, including both investment and current expenditure. Expenditure financed by loans under the SAFE instrument would automatically benefit from flexibility upon activation of the escape clause. The flexibility will allow for a deviation from the agreed expenditure path equivalent to the increase in defence expenditure. The national escape clause will be available for a period of four years, starting from 2025, with the total amount of additional eligible expenditure capped at 1.5% of GDP, using 2021 as the reference year (the year immediately prior to the exceptional circumstances triggering the activation of the national escape clause).

One of the conditions for activating the national escape clause is to ensure that such a deviation does not endanger fiscal sustainability over the medium term. A simulation of fiscal sustainability assessment for EU Member States that requested activation of the national escape clause, conducted by [70], shows that an increase in defence expenditure during the period covered by the escape clause would require greater adjustment measures in the subsequent period in order to preserve fiscal sustainability. Compared to the levels projected under the approved net expenditure paths, the average deficit-to-GDP and debt-to-GDP ratios in 2028 could be higher by 1.3 percentage points and 2.6 percentage points respectively, if the maximum allowed increase in defence expenditure (by 1.5% of GDP) were to be implemented gradually over the 2025–2028 period. Therefore, higher expenditure in the 2025–2028 period would, in the second round of medium-term fiscal-structural plans starting in 2029, on average require additional fiscal efforts of 0.4 percentage points, or 0.25 percentage points in the case of an extended seven-year adjustment period, in order to meet the conditions linked to the reference values for public debt and the budget deficit. Following the activation period of the national escape clause, new medium-term fiscal-structural plans will have to account, from the outset, for higher levels of public debt and budget deficit. This could potentially lead to greater adjustment requirements to achieve fiscal sustainability and may affect the developmental role of public finances due to pressure to reduce productive public expenditure as a result of rising defence spending.

Furthermore, while economic growth contributes to an expansion of the tax base, this does not necessarily lead to more efficient mobilisation of general government revenue or to its targeted allocation to key areas of development. The quality of fiscal institutions, the efficiency of the tax system, and the degree of budgetary responsibility and strategic planning play an important role. In the absence of an appropriate institutional framework, additional general government revenue generated by economic growth may remain underutilised or may be directed towards unproductive expenditure that does not yield long-term developmental effects. Moreover, maintaining the developmental role of public finances requires active and deliberate political action. Increases in general government expenditure in areas such as education, research and development, public infrastructure and institutional strengthening are not the result of market processes, but of strategic decisions by policymakers. Without clear development priorities and appropriate fiscal governance, economic growth, even when generating additional general government revenue, will not necessarily lead to productive expenditure and a strengthening of the developmental role of public finances see, for example, ref. [71]. In this context, there is a risk that policymakers’ decisions are influenced by the prevailing political cycle.

In addition, in preventing the lack of fiscal discipline and preserving the developmental role of public finances, independent fiscal oversight institutions, i.e., fiscal councils, play an important role. They influence the conduct of fiscal policy. These institutions are independent of the government (i.e., not subject to detrimental political incentives) and play a significant role in shaping fiscal policy. In practice, fiscal councils monitor the preparation of macroeconomic forecasts, produce analyses for budget preparation and oversee budgetary developments in relation to established objectives. They also assess the budgetary impact of specific measures, which may include recommendations regarding compliance of fiscal policy with fiscal rules laid down at both the national and EU levels in the context of economic governance in the Member States. High-quality and realistic economic growth forecasts are expected to limit excessive growth in general government expenditure, which is often the result of overly optimistic macroeconomic projections. The objective is to preserve the sustainability of public finances and their developmental orientation see, for example, refs. [72,73].

6. Conclusions

This study introduces a measurable indicator of the developmental role of public finances and examines the validity of Wagner’s law, taking into account productive government expenditure in addition to general government expenditure. Our panel vector error correction model estimates indicate bidirectional causality running from general government expenditure to GDP per capita and vice versa in the short-run, while in the long-run the validity of Wagner’s law cannot be confirmed. For productive government expenditure, results suggest a weak indication of short-run causality from productive government expenditure to GDP per capita and not vice versa. Although economic growth contributes to higher general government revenue, its use often follows the inertia of existing budgetary structures and short-term redistributive and political priorities, rather than being directed towards productive, development-oriented expenditure with high long-term multiplier effects.

Wagner’s law and the Keynesian hypothesis should not be viewed as competing explanations but rather as complementary frameworks. The methodological advancements in the recent literature have been instrumental in reconciling previously conflicting results, revealing a more nuanced reality in which causality is often bidirectional, conditional, and shaped by the composition and efficiency of public expenditure.

Future challenges may include maintaining the efficiency of public expenditure under changing fiscal and economic conditions supporting sustainable development goals. Additional challenges could arise from balancing economic, social, and environmental objectives within fiscal policy frameworks. The effective allocation of public resources toward sustainable and inclusive development is also likely to remain an important policy issue.

While this study provides important insights, certain limitations remain. The analysis focuses solely on euro area countries, and future research could extend the framework to non-euro area countries and countries with high and low income per capita, generate alternative proxies for productive government expenditures or employ a composite index approach. Such extensions would further enhance our understanding of the validity of Wagner’s law.