Unlocking the Nexus: Personal Remittances and Economic Drivers Shaping Housing Prices Across EU Borders

Abstract

This study examines the impact of personal remittances on housing prices in European Union (EU) countries, while also accounting for a broader set of macroeconomic, demographic, and structural variables. Using annual data for 27 EU countries from 2007 to 2022, we employ a comprehensive panel econometric approach, including cross-sectional dependence tests, second-generation unit root tests, pooled mean group–autoregressive distributed lag (PMG-ARDL) estimation, and panel causality tests, to capture both short- and long-term dynamics. Our findings confirm that remittances significantly and positively influence long-term housing price levels, underscoring their relevance as a demand-side driver. Other key variables such as net migration, GDP, travel credit to GDP, economic freedom, and real effective exchange rates also contribute to housing price movements, while supply-side indicators, including production in construction and building permits, exert moderating effects. Moreover, real interest rates are shown to have a significant long-term negative effect on property prices. The analysis reveals key causal links from remittances, FDI, and net migration to housing prices, highlighting their structural and predictive roles. Bidirectional causality between economic freedom, housing output, and prices indicates reinforcing feedback effects. These findings position remittances as both a development tool and a key indicator of real estate dynamics. The study highlights complex interactions between international financial flows, demographic pressures, and domestic economic conditions and the need for policymakers to consider remittances and migrant investments in real estate strategies. These findings offer important implications for policymakers seeking to balance housing affordability, investment, and economic resilience in the EU context and key insights into the complexity of economic factors and real estate prices. Importantly, the analysis identifies several causal relationships, notably from remittances, FDI, and net migration toward housing prices, underscoring their predictive and structural importance. Bidirectional causality between economic freedom and house prices, as well as between housing output and pricing, reflects feedback mechanisms that further reinforce market dynamics. These results position remittances not only as a developmental instrument but also as a key signal for real estate market performance in recipient economies.

1. Introduction

The European real estate market has been subject to considerable fluctuations in recent years, reflecting both global economic trends and regional dynamics. Data from the ref. [1] shows that over the past three decades, real house prices in the OECD have risen by nearly 60 index points on average, while ref. [2] highlights a sustained upward trend in property prices in the EU since 2010, resulting in a remarkable 45% increase in the cost of home ownership. However, this growth path experienced a turning point in 2023, marked by the first global downturn in the real estate market after years of continuous expansion, as reported by ref. [3]. Ref. [4] further emphasizes this turning point by highlighting that real estate is overvalued by 15% to 20% in most European housing markets. When examining recent trends, ref. [2] points out that house prices started to outpace inflation in 2015 and rose by 3.0% to 5.5% above inflation between 2016 and 2021. However, this pattern changed in 2022, when house prices only marginally outpaced inflation (+0.3%) and recorded a significant decline of 6.8% annually in inflation-adjusted terms in 2023 amid high inflation rates. This dynamic underscores the complex interplay between economic factors and housing market trends and warrants a closer examination of the underlying causes and implications for stakeholders. In addition to the real estate market, this study places special emphasis on remittances, which are considered as an independent variable. While the impact of immigrants’ remittances on real estate prices has been extensively researched in Africa ref. [5,6,7] and Asia ref. [8,9,10], to the authors’ knowledge, there is currently no published research addressing EU countries, indicating a significant research gap. Therefore, the results of this study will represent an original scientific contribution. Despite the seriousness of these problems, this area of research has not yet been studied in depth.

The aim of this study was to examine a range of explanatory variables beyond personal remittances in order to explain house price fluctuations in EU countries. These include macroeconomic indicators likely to influence real estate markets, such as foreign direct investment, the ratio of travel credit to GDP, the World Risk Index, net migration, economic freedom, and gross domestic product (GDP). Financial and sectoral variables, namely the real interest rate, building permits, production in construction, and the real effective exchange rate, are also included to capture credit conditions, supply-side dynamics, and external competitiveness. Taking into account intra-EU migration flows, this study aims to identify the short- and long-term effects of remittances and other economic determinants on housing prices. This multifactorial approach builds on previous research and aims to enhance the robustness of the results.

For instance, ref. [9] found that remittances to migrant families in Bangladesh improved living standards, including housing, food, education, and healthcare. However, they also noted a correlation between remittances and property price appreciation. Similarly, ref. [11] suggested in Colombia that remittances directed to housing investments could expand housing supply, lowering property prices. Empirical evidence showed a significant negative correlation between received remittances and both sales and rental housing prices, indicating that remittances for housing could stimulate construction, thus reducing property prices. Regarding the other two independent variables, GDP and FDI, findings according to ref. [12] suggest that foreign direct investments in real estate do not cause property price increases or stimulate economic growth in either the short or long term. Instead, property price appreciation has a positive causal relationship with economic growth. While the investigation of the other independent variables is still insufficiently researched, it is conceivable that various factors can influence residential real estate prices. A higher ratio of travel credit to GDP could be a sign of a thriving tourism sector, which could lead to increased demand for residential real estate as this ratio grows. Conversely, countries with lower scores on the World Risk Index may have greater stability and security, attracting higher demand for residential real estate. Positive net migration, which indicates a net inflow of people into a country or region, can increase demand for residential real estate. In addition, a higher degree of economic freedom, characterized by less government intervention and regulation, can foster an environment conducive to investment and entrepreneurship. Consequently, this can stimulate economic growth, job creation, and demand for residential real estate. Movements in the real interest rate critically shape borrowing costs for households and thus influence housing affordability. Changes in real borrowing costs can affect mortgage affordability and credit conditions, with lower rates generally supporting demand and higher rates suppressing it. As highlighted by ref. [13], even modest increases in real interest rates may prompt substantial house price corrections in low-rate environments, underscoring the RIR’s relevance for policy and housing affordability. A higher number of approved permits may indicate anticipated market demand and developer confidence. Ref. [14] suggest that while permits are useful indicators, their long-run impact on price dynamics may be modest due to regulatory and market lags in actual housing delivery. Production in construction reflects the actual delivery of residential units to the market. Increases in construction output may signal a supply-side response to rising prices or broader economic activity. As ref. [15] notes, construction sector dynamics often align with housing and monetary policy cycles, linking production in construction to both demand and investment trends in real estate. The real effective exchange rate is also considered, reflecting changes in a country’s trade competitiveness. While the REER is a common macroeconomic indicator, its direct role in housing price dynamics appears limited. According to ref. [16,17], housing markets in the EU tend to respond more strongly to internal drivers like income and credit availability than to exchange rate fluctuations.

This study attempts to fill this gap by comprehensively analyzing the impact of economic factors on house prices in EU countries, shedding light on previously unexplored dynamics and providing insights that are crucial for policymakers, investors, and stakeholders in the real estate market.

The remainder of this study is structured as follows: the Introduction is followed by the Literature Review, the second part deals with the data and methodology, the third part presents the research Results and Discussion, and the last part is the Conclusion, which summarizes all the results.

2. Theoretical Background and Literature Review

Domestic and foreign investors have a significant influence in real estate market research ref. [18]. However, it is also important to consider the role of the diaspora in shaping the real estate market, since a significant portion of remittances sent by the diaspora to their home countries are often invested in real estate. Already in the 1980s, ref. [19] showed that households mostly spend remittances on basic necessities such as food and clothing, debt repayment, house construction, education, savings, and land purchase.

Table 1. State of the art.

Authors	Study Place and Period	Methodology	Findings
[6]	Ghana, Accra	Survey	Rising remittances or improved access to finance can strongly affect housing demand and prices, potentially slowing migration.
[7]	Accra, Ghana, 1990–2003	Variant of the traditional mono-centric city model to calculate the city’s housing supply elasticity	Remittances boost housing supply but also raise costs due to its inflexible housing market.
[20]	Nairobi city	Stepwise regression modeling	House prices correlate with economic fundamentals like money supply, population growth, and inflation, but are particularly influenced by population growth.
[5]	Kenya, 1970–2008	Autoregressive distributed lag (ARDL)	The short- and long-run elasticities show that remittances influence housing construction demand in Kenya.
[21]	Nepal, 2005–2010	Analysis of behavioral basis	Remittance recipients invested in real estate, driving property price increases.
[22]	Mexico, monthly CPI data for 272 consumption items for the period 1996:01–2007:06	Vector autoregressive model	Remittances primarily benefit non-tradable services like rental housing, electricity, cars, and restaurants, with their impact increasing over time.
[23]	Kenya, 2003–2012	Multiple linear regressions method	Investing remittances in government securities provides external financing to enhance real estate, infrastructure, and tourism industries in Kenya.
[8]	Nepal, 2000/01–2010/11	Descriptive analysis	Remittance-funded imports crucially generate revenue, contributing to a notable rise in real estate and housing prices.
[24]	Nairobi, Kenya, 2000–2013.	Descriptive technique, regression analysis, ANOVA	Diaspora remittances, exchange rates, inflation, currency circulation, and real GDP growth individually do not affect real estate investment in Nairobi, but their combined changes do influence it.
[9]	Bangladesh, data collected in 2000, 2005, and 2010	Ordinary least squares (OLS) and quantile regressions	Both domestic and foreign remittances significantly impact housing prices, though the pricing of housing attributes may vary across different price ranges.
[25]	Kenya, 2006 quarter one to 2015 quarter four	Panel data regression model	The study found a strong positive correlation between real estate prices and diaspora remittances.
[26]	Ghana, 1986–2017	Autoregressive distributed lag model (ARDL)	Real estate prices are cointegrated with exchange rates, remittances, and inflation, with remittances having a positive effect on real estate prices.
[10]	Large cities in Pakistan, 2015–2016	Heckman’s sample selection model and Household Integrated Economic Survey (HIES)	Remittances significantly boost housing demand by increasing households’ purchasing power, driving up the demand for new homes.
[27]	Jordan, 1990–2015	Descriptive and analytical approaches	Remittances positively impact Jordan’s real estate market: increased funds from expatriates correspond to market growth.
[11]	Colombia, 2000–2016	VAR mode	Remittances briefly affect the housing market, but their influence diminishes over time. They boost housing supply, resulting in lower house prices.
[28]	Pakistan	Johansen test, Chow breakpoint test	Remittances have a positive impact on house prices due to the linkages between the development of urban areas and employment.
[29]	Developing countries from the period 1950 to 2007	Conceptual and theoretical literature review	Remittances are frequently invested in housing and land, leading to rising real estate prices in migrant-sending areas.
[30]	Kenya, 2004–2020	Autoregressive distributed lag model (ARDL)	Remittances are strongly linked to house prices. Adjusting for GDP per capita, lending rates, and construction costs, a 1% increase in remittances equals a 0.10% rise in house prices.
[31]	Kenya, quarterly time series data from the period 2004 to 2020	Autoregressive distributed lag model (ARDL)	Real remittances have a strong positive relationship with real house prices in the long run.
[32]	Kenya, 2004–2013	Descriptive research design	Diaspora remittances, GDP, unemployment, lending rates, and inflation are not statistically significant in influencing real estate growth in Kenya at a 5% level of significance.
[33]	Nairobi, Kenya	Granger causality tests	Housing prices correlate with lending rates, GDP, real estate loans, and construction costs, but not with diaspora remittances according to Granger causality tests.
[34]	Nepal, 2008–2010	Descriptive analysis	The 2009/10 slowdown in remittance growth led to declining real estate and housing prices, liquidity issues, and balance sheet deterioration for BFIs.
[35]	Rural highland Ecuador, village Xarbán	Questionnaire, in-depth interviews, participant observation	Remittances intended for rural Ecuador’s housing sector remained unutilized, resulting in a surplus supply and driving house prices down.

Source: own research.

provides a comprehensive examination of prevailing research studies into the relationship between remittances and fluctuations in real estate prices.

From the previous research results, it can be concluded that the relationship between remittances and real estate prices is very complex and varies from country to country. Additionally, according to the geographical criterion, the research is primarily focused on African countries (Kenya and Ghana), in Latin America (Ecuador, Colombia, and Mexico), and in Asia (Pakistan, Bangladesh, Nepal, and Jordan), indicating that the European region remains unexplored, thus closing the research gap in this study.

Examining the research results according to Table 1, it can be concluded that the majority of studies have demonstrated a positive impact of remittances on real estate price growth. Authors that support this finding are ref. [5,6,7,8,9,10,20,21,22,23,24,25,26,27,28,29,30].

In contrast to previous studies, some studies found no evidence of a relationship between remittances and real estate prices ref. [32,33], while some studies even found a negative relationship ref. [11,34,35]. Considering the lack of research on this topic in Europe, and recognizing that Europe is currently experiencing high migration flows and a significant rise in house prices, this topic holds substantial relevance for all stakeholders within EU countries.

3. Data Description

The study is based on data from 27 member states of the European Union, covering the period from 2007 to 2022. This timeframe was selected as it represents the most extensive dataset available for the variables under analysis. Spanning over more than a decade, it offers a comprehensive lens through which the profound economic changes and structural transformations across the EU can be systematically observed and analyzed. The variables used in the research are represented in

Table 2. Variables’ description.

Variable	Definition	Source
HP	House price index (index, 2015 = 100)	Eurostat database
RGDP	Personal remittances, received (% of GDP)	World Bank database
FGDP	Foreign direct investment, net inflows (% of GDP)	World Development Indicators
TGDP	Travel credit to GDP in %	Eurostat database
R	World Risk Index	The data of the World Risk Index by Bündnis Entwicklung Hilft and Institute of International Law of Peace and Armed Conflict
NM	Net migration (number of immigrants minus the number of emigrants)	World Bank database
EF	Index of economic freedom	Heritage foundation
GDP	GDP, current prices (billions of USD)	World Bank database
REER	Real effective exchange rate (index, 2015 = 100)	Eurostat database
RIR	Real interest rate (proxy: deflated EMU convergence criterion)	Eurostat database
PCON	Production in construction (index, 2021 = 100)	Eurostat database
PER	Building permits (residential buildings, except residences for communities)	Eurostat database

.

To detect the variability and distribution characteristics of the variables, the basic statistical descriptions of house prices (HP), remittances to GDP (RGDP), foreign direct investment to GDP (FGDP), gross domestic product (GDP), travel credit to GDP (TGDP), real interest rate (RIR), building permits (PER), production in construction (PCON), real effective exchange rate (REER), risk index (R), net migration (NM), and economic freedom (EF) are presented in

Table 3. Descriptive statistics.

	Mean	Median	Max.	Min.	Std. Dev.	Obs.
HP	12.30771	12.00000	113.8000	43.60000	20.94517	428
RGDP	1.567392	1.035035	7.531268	0.035033	1.446421	428
FGDP	10.41578	3.035411	449.0828	394.4716	44.71416	428
GDP	571.3626	236.6590	4281.348	7.989000	894.2807	428
TGDP	0.043230	0.027471	0.191559	0.003638	0.037450	428
RIR	3.094741	2.608814	21.75595	−0.482042	2.802699	428
PER	62.49346	27.45000	633.4000	2.100000	99.16485	428
PCON	99.50794	96.10000	593.1000	42.50000	48.03532	428
REER	100.2984	100.0000	124.7335	88.98679	3.670696	428
R	3.351776	2.580000	10.40000	0.390000	2.557551	428
NM	44213.51	9781.000	3366387.	−254292.0	194234.7	428
EF	69.30607	69.50000	82.60000	53.20000	5.799759	428

Source: Authors’ construct. Note: On average, HP was 12.31, which indicates growth; RGDP accounted for 1.57%; FGDP represented 10.42%; GDP stood at USD 571.36 billion; TGDP comprised 0.043%; and RIR averaged 3.09%. PER and PCON had mean values of 62.49 and 99.51, respectively. The average REER was 100.30. The R averaged 3.35, NM was 44,213.5 individuals, and the EF averaged 69.31. Country-level scatter plots of HP against the RGDP are presented in Appendix A.

.

4. Research Methodology

This study adopts a quantitative research design, employing econometric analysis to empirically assess the causal relationships between RGDP and HP, along with other variables, across EU member states. The quantitative approach is appropriate given the study’s objective to analyze dynamic interactions using large-scale panel data, where patterns and effects are best identified through statistical modeling ref. [36]. Panel data methods enable the exploration of both cross-sectional variation and temporal dynamics, which is particularly valuable when evaluating economic relationships in a heterogeneous setting such as the EU ref. [37].

To achieve this, the study applies a suite of panel econometric techniques designed to account for the structure and characteristics of macroeconomic panel data. Firstly, we conduct residual cross-section dependence tests to evaluate potential contemporaneous correlations among countries, which, if present, may bias standard estimators. To determine the integration properties of the variables, we apply second-generation panel unit root tests, which account for cross-sectional dependence. Given that the variables are integrated of mixed order and cointegrated, we employ the pooled mean group–autoregressive distributed lag (PMG-ARDL) estimator ref. [38], which allows for heterogeneous short-run dynamics while assuming homogenous long-run coefficients across EU countries. This approach is particularly suited to datasets with moderate time dimensions and mixed stationarity properties. Finally, to assess the direction of causality, we use the ref. [39] Granger non-causality test, providing robust evidence of short-run causal relationships among variables. This multi-step approach provides methodological rigor in detecting dynamic and causal links, directly supporting the aim of identifying how RGDP and other relevant variables influence HP developments in EU countries.

Before assessing the stationarity of the data, conducting a cross-section dependence test on the panel data is necessary. Ref. [40] proposed a straightforward test to assess the cross-sectional dependence (CD) of errors that can be applied to various panel models where N > T, in comparison with the Breusch–Pagan LM test, which performs badly when there are larger number of units and smaller number of time periods. According to ref. [40], this test relies on computing pairwise correlation coefficients of OLS residuals obtained from individual regressions within the panel.

$$CD=\sqrt{{\displaystyle \frac{2T}{N\left(N-1\right)}}\left(\sum _{i=1}^{N-1}\sum _{j=i+1}^N\widehat{{\rho }_{ij}}\right)}$$

(1)

where pairwise correlation coefficients can be written as $\widehat{{\rho }_{ij}}=\sum _{t=1}^T{e}_{it}{e}_{jt}/{\left(\sum _{t=1}^T{e}_{it}^2\right)}^{{\displaystyle \frac{1}{2}}}{\left(\sum _{t=1}^T{e}_{jt}^2\right)}^{{\displaystyle \frac{1}{2}}}$ where eit represents the OLS residuals based on T observations ei = (ei1, ei2,….,eiT)’. With cross-sectional dependence between the variables, a second-generation unit root test must be applied. According to refs. [41,42], they proposed a technique utilizing an ADF regression incorporating a lagged cross-sectional mean and its first difference to account for cross-sectional dependence. This approach is referred to as the Cross-Sectional Dickey Fuller test (CADF), which is given by following equation:

$$∆y_{it}={\alpha }_i+{\rho }_i{y}_{i,t-1}+d_0{\overline{y}}_{t-1}+\sum _{j=0}^p{d}_{j+1}∆{\overline{y}}_{t-j}+\sum _{k=1}^p{c}_k∆y_{i,t-k}+{\epsilon }_{it}$$

(2)

where ${\overline{y}}_t$ and $y_t$ represent the averages for the lagged and first differences in each cross-section series. Ref. [42] introduced a modified IPS statistic derived from the mean of individual CADF values. He averaged the t-statistics on the lagged CADF values to compute the CIPS statistics, denoted as the Cross-Sectionally Augmented IPS. This is expressed as follows:

$$CIPS= N^{-1}\sum _{i=1}^N{CADF}_i$$

(3)

After checking the stationarity of the data, the next step was to examinate the panel heterogeneity. Within a panel data framework, it is pivotal to determine whether the data generating process exhibits homogeneity or heterogeneity, especially concerning the presence or absence of individual effects. Ensuring homogeneity of the slope coefficients holds considerable importance in various analytical contexts, such as selecting the appropriate unit root, evaluating cointegration, and investigating causality ref. [43]. The homogeneity test holds significance in determining whether the unknown slope coefficients (β) remain consistent across sections, although this consistency may not always hold true over time ref. [44]. In instances of panel heterogeneity, adopting the assumption of uniform slopes can lead to inaccurate estimates ref. [45]. The examination of country heterogeneity is facilitated through the application of the Hsiao test ref. [46].

The Hsiao test [36] is utilized to examine country heterogeneity. According to ref. [46], the homogeneity test is based on three hypotheses. H1 suggests that the slope coefficients are uniform, contrasting with the alternative hypothesis that allows for heterogeneity. Conversely, H2 assumes a homogeneity corresponding to that of H1, but its alternative form allows for the possibility of heterogeneity. Finally, H3 proposes complete uniformity of slope coefficients, while its alternative hypothesis acknowledges the potential for partial uniformity. Acceptance of the null hypothesis indicates homogeneity among sectional individuals.

This study employs an advanced analytical technique, the pooled mean group–autoregressive distributed lag (PMG-ARDL) model, to explore the immediate and long-lasting effects of remittance inflows, GDP fluctuations, economic freedom levels, risk factors, and tourism dynamics on housing prices. This model, pioneered by ref. [47], offers a robust framework for assessing the complex interplay among these variables over both short- and long-term horizons. Ref. [38] proposed the pooled mean group (PMG) assessment procedure as a solution for heterogeneity bias caused by heterogeneous slopes in dynamic panels. PMGs allow for a higher degree of heterogeneity of parameters in growth regressions ref. [38].

The primary advantage of PMG estimators lies in their efficacy when dealing with a limited number of observations, i.e., the number of N is approximately 20–30 countries; secondly, they offer the simultaneous correction of autocorrelation issues while demonstrating minimal susceptibility to outliers, and thirdly, their attribute enhances the robustness of the analysis, ensuring that the model’s performance remains stable and reliable even when confronted with anomalous data points ref. [38]. The equation of the panel PMG-ARDL is specified as follows

$$∆ln{HP}_{it }={\vartheta }_i(ln{HP}_{i,t-1}+{\sigma }_i\prime X_{i,t})+{\sum }_{j=1}^{p-1}{\beta }_{ij}∆ln{HP}_{i,t-j}+{\sum }_{j=0}^{q-1}{\pi }_{ij}^{\prime }∆X_{i,t-j}+{\omega }_i+{\epsilon }_{it}$$

(4)

In Equation (4), ${\vartheta }_i$ represents the specific coefficient indicating the speed of adjustment for each group, while ${\sigma }_i$ represents the vector of variables under analysis, namely GDP, NM, EF, FGDP, TGDP, R, RIR, PER, PCON, and REER, measuring their long-term impact on hp. The parameter $\left(ln{HP}_{i,t-1}+{\sigma }_i^{\prime }{X}_{i,t}\right)$ denotes the error correction speed of adjustment term (ECT), which detects any deviations from the equilibrium relationship in the long run. A negative value of the error correction term is anticipated, signaling a reversion to the long-term equilibrium. The coefficients ${\beta }_{ij}$ reflect the short-term dynamic relationships between the explanatory variables and the target indicator. The ${\pi }_{ij}^{\prime }$ capture the immediate and lagged marginal effects of changes in the exogenous variables X on the short-term dynamics of ΔlnHP for each group. ${\epsilon }_{it}$ stands for the error term, while p and q denote the optimal lag orders, and ${\omega }_i$ represents the constant term.

The ARDL method accommodates diverse stationarity properties, as it can be applied when factors exhibit different levels of stationarity: either stationary at the first difference (I(1)), at levels (I(0)), or a combination of both. This adaptability underscores the method’s efficacy across various empirical contexts, ensuring its applicability in scenarios characterized by disparate temporal characteristics of the variables under consideration. This approach proves beneficial in situations where there is a belief that the long-term balance between variables is comparable across countries, or at least within a certain group of countries. As ref. [48] stated, it permits country-specific short-term adjustments, considering the diverse effects of factors like vulnerability to financial crises, external shocks, stabilization policy, and monetary policy.

Although PMG-ARDL will reveal key assumptions in this research paper, these findings need further clarification through the Dumitrescu and Hurlin ref. [39] causality test in order to identify causal relationships. Essentially, this test is considered an advancement over the Granger causality test ref. [49]. Unlike conventional Granger causality methodology, the Dumitrescu–Hurlin test assumes the variability in all coefficients across different cross-sectional units. Through Monte Carlo experiments, the authors have demonstrated the robust performance of this test even when applied to datasets with a short time span, particularly in the presence of cross-sectional dependence. The test can be used for balanced and heterogeneous panels. It is applicable in scenarios where the number of observations (N) is increasing over time or when the number of observations exceeds the number of time periods (N > T), and vice versa. The authors considered the following panel data model:

$$y_{1,t}={\alpha }_i+{\sum }_{i=1}^K{\gamma }_i^{\left(k\right)}{y}_{i,t-k}+{\sum }_{i=1}^K{\beta }_i^{\left(k\right)}{\chi }_{i,t-k}+{\epsilon }_{i,t}$$

(5)

where intercept ${\alpha }_i$ and coefficients ${\beta }_i{=\beta }_i^{\left(1\right)},\dots ,{\beta }_i^{\left(k\right)}$ are assumed to be time-invariant, while the autoregressive parameters ${\gamma }_i^{\left(k\right)}$ and regression coefficient ${\beta }_i^{\left(k\right)}$ are deemed to vary across cross-sections. Here, K denotes the lag length. In line with ref. [44], the Dumitrescu–Hurlin test assumes the possibility that causality exists between certain individuals within a dataset, albeit not necessarily across all units.

$$\begin{array}{cc}{H}_1:\hfill & {\beta }_{i1}=\dots ={\beta }_{iK}=0 \forall i=1,\dots ,N\hfill \\ & {\beta }_{i1}\ne 0 or\dots or {\beta }_{iK}\ne 0 \forall i=N_1+1, \dots ,N\hfill \end{array}$$

(6)

where $N_1$ is unknown but satisfies the condition 0 ≤ $N_1$/N < 1. If $N_1$ equals 0, causality is present across all individuals. Conversely, if the condition $N_1$/N < 1 is not met, indicating $N_1$ is less than N, causality is absent for all individuals, thereby reducing H1 to H0.

5. Results and Discussion

Since this research investigates the association between the HP as the dependent variable and RGDP, EF, NM, GDP, FGDP, TGDP, REER, RIR, PCON, and PER as independent variables, the first step was to check the stationarity of the panel data. First-generation unit root tests, as introduced by refs. [45,50,51,52], assume the independence of residuals in the cross-section. Therefore, it is important to check for the presence of cross-sectional dependencies between the residuals to ensure the validity and robustness of these tests.

Based on the findings from

Table 4. Residual cross-section dependence test.

Test	Statistic	D.F.	Prob.
Breusch–Pagan LM	1466.070	351	0.0000
Pesaran scaled LM	42.08559		0.0000
Pesaran CD	22.53363		0.0000

the null hypothesis of no cross-section dependence in the residuals can be rejected at 1% of the significance level. Therefore, it can be concluded that the second-generation unit root test should be applied.

Table 5. CIPS test.

Variables	Log	First Difference	Level of Integration
LHP	−3.149 ***	−3.344 ***	I(0)
LGDP	−0.013	−1.741 **	I(1)
LFGDP	−2.304 **	−4.413 ***	I(0)
LTGDP	−18.597 ***	−3.170 ***	I(0)
LR	−2.222 ***	−3.455 ***	I(0)
LEF	−1.700 **	−2.399 ***	I(0)
LGDP	−0.918	−1.654 **	I(1)
LNM	−0.393	−12.682 ***	I(1)
LREER	−0.49933	−1.78584 **	I(1)
LPCON	−1.49497	−2.53514 ***	I(1)
LPER	−1.66043 **	−2.89411 ***	I(0)
LRIR	−2.12628	−3.13201 ***	I(1)

Note: *** and ** refer to the significance level at 1% and 5%, respectively. Lag length selection is based on Akaike info criterion.

presents the results of the CIPS unit root test. It can be concluded that most variables are stationary after logarithmic transformation. However, RGDP, GDP, NM, REER, RIR, and PCON became stationary after taking the first difference, implying they are integrated of order I(1). The Hsiao test assessing the homogeneity of slopes within the panel data is displayed in Table 5.

The results presented in

Table 6. Hsiao tests.

Hypothesis	F-Stat	p-Value
H1	4.628546	3.67 × 10−17
H2	3.0644841	1.64 × 10−10
H3	8.241880	7.06 × 10−24

suggest that the estimated statistical tests and associated values permit rejection of the null hypothesis positing homogeneous slopes. This underscores the significance of accounting for both slope and intercept heterogeneity. Dumitrescu and Hurlin ref. [39] modified the Granger ref. [49] non-causality test for heterogeneous panels.

The subsequent crucial phase involved estimating both the long-run and short-run effects, for which panel PMG-ARDL models were employed. To systematically explore the heterogeneity of theoretically relevant determinants of HP, we estimate a series of progressively extended models. The specifications in Models 1–3 are guided by theoretical relevance, empirical precedent, and incremental model-building strategies recommended in the panel time series literature refs. [38,53]. All models include RGDP, the core variable of interest, reflecting the aim of this study. Furthermore, the models incrementally incorporate additional variables based on their theoretical relevance to housing prices. This layered approach facilitates a clearer interpretation of marginal contributions and potential collinearity among regressors ref. [54]. The extension to Models 4–7 responds to further robustness needs, incorporating key financial and supply-side variables such as the REER, PCON, PER, and RIR. These additions ensure a more comprehensive coverage of both demand and supply dynamics affecting EU housing markets. Thus, the model design reflects a structured, theory-informed, and methodologically consistent progression aligned with best practices in applied panel econometrics.

Specifically, the selected PMG-ARDL models adhere to an ARDL (1,1,1,1,1) specification based on the Akaike information criterion for lag determination. The outcomes of these PMG-ARDL models are methodically detailed in

Table 7. PMG-ARDL empirical results.

lHP (Dependent Variable)
Variables	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6	Model 7
Long-run coefficients
dlRGDP	2.175321 (0.2391)	3.603530 (0.0466) **	5.006294 (0.0028) ***	10.68209 (0.0000) ***	9.962634 (0.0000) ***	2.409121 (0.0000) ***	3.268322 (0.0000) ***
lFGDP							−0.011647 (0.0000) ***
lTGDP			−1.058647 (0.0000) ***			−1.581399 (0.0000) ***
lR		0.027437 (0.8631)
dlGDP	40.76021 (0.0000) ***	43.15926 (0.0000) ***	26.51823 (0.0000) ***	15.36940 (0.0000) ***	47.51800 (0.0000) ***		0.734377 (0.0000) ***
dlNM	0.859937 (0.0000) ***	1.122729 (0.0000) ***
lEF			−0.663049 (0.0000) ***			−0.853115 (0.0000) ***
dlREER				50.22280 (0.0002) ***
dlPCON					−7.049899 (0.0000) ***
lPER						−0.302570 (0.0000) ***
dlRIR							−0.577096 (0.0000) ***
Short-run coefficients
ECM	−0.201062 (0.0000) ***	−0.194577 (0.0000) ***	−0.261180 (0.0000) ***	−0.180429 (0.0000) ***	−0.209562 (0.0000) ***	−0.338483 (0.0000) ***	−0.562253 (0.0000) ***
d(dlRGDP)	−0.564203 (0.2000)	−0.905408 (0.0512) *	−1.139116 (0.0049) ***	−2.788085 (0.0002) ***	−3.048129 (0.0000) ***	−0.652742 (0.5787)	−4.077597 (0.0666) *
d(lFGDP)							0.652075 (0.0543) *
d(lTGDP)			−0.123730 (0.4946)			−0.370664 (0.3654)
d(LR)		−1.702792 (0.1421)
d(dlGDP)	−4.301475 (0.0000) ***	−4.204911 (0.0000) ***	−3.585476 (0.0000) ***	−1.747789 (0.0942) *	−6.594748 (0.0001) ***		−2.179229 (0.2254)
d(dlNM)	−0.914532 (0.1228)	−0.951713 (0.1017)
d(lEF)			8.771247 (0.0846) *			−3.954954 (0.5375)
d(dlREER)				−22.02609 (0.0034) ***
d(dPCON)					1.451330 (0.3215)
d(lPER)						1.562239 (0.0857) *
d(dlRIR)							0.306144 (0.3711)

Notes: ***, **, and * indicate the significance at 1%, 5%, and 10%. The value of the coefficient is outside the brackets. The value of the p-value is inside the brackets.

, providing comprehensive insights into the dynamics and relationships among the variables under investigation.

The ECT parameter coefficient is expected to be negative and significant ref. [55]. The findings reveal that the error correction terms (ECTs) exhibit consistently negative and statistically significant values across all seven models. This proves the presence of long-term relationships among the variables under analysis. Such robust empirical evidence underscores the persistence of equilibrium adjustments over time, clarifying the permanent interdependencies among the analyzed variables.

In the long-run analysis, the empirical findings consistently demonstrate a positive and statistically significant relationship between the variable of primary interest, namely dlRGDP, and lHP across all observed models (except Model 1 when the findings are not statistically significant). In terms of magnitude, the long-run coefficients of dlRGDP range from 2.41 to 10.68 across Models 2 through 7, and are statistically significant at the 1% level. This implies that a 1-percentage-point increase in dlRGDP is associated with an increase in lHP of between 2.41% (Model 6) and 10.68% (Model 4), depending on the model specification and included control variables. These effect sizes are economically meaningful, indicating that an increase in the proportion of remittances within GDP positively impacts house prices within the European Union (EU). Such results underscore the substantive impact of remittance inflows on housing market dynamics, emphasizing their role as a significant driver of house price appreciation within the EU context. Such research results are in line with ref. [25,30,31] who observed a positive and significant relationship between remittance and house prices. Simultaneously, in six out of seven observed models, dlGDP also positively and statistically significantly affects European house prices. Similar results showing that a rise in GDP growth is associated with an increase in housing prices may be found in refs. [56,57]. In a high-growth economy, households often seek better-quality properties due to increased disposable income, improved financial stability, and rising aspirations for enhanced living standards. As individuals experience economic prosperity, they become willing to pay premium prices for properties that offer superior amenities, features, and overall quality.

In Model 1 and Model 2, the level of dlNM also influences house price growth. The positive coefficient associated with dlNM indicates that higher levels of dlNM contribute to an increase in lHP. This delicate insight underscores the complex interaction between economic factors and demographic trends in shaping the dynamics of the European housing market. Refs. [58,59,60] also recognized the pressure of migration on house price growth. The flow of migration exerts a noticeable effect on housing markets in destination areas, causing a notable surge in housing demand and subsequent escalation of house prices. In Model 2, we also found that lR is a positive but not statistically significant variable.

Interestingly, in both Model 3 and Model 6, lTGDP and lEF exhibit a negative and statistically significant impact on lHP. The negative impact of lEF on lHP, as observed in these two models, while not widely documented in the existing literature, can be theoretically supported by the arguments presented by ref. [61] who posits that political uncertainty escalates the risk premium, resulting in increased volatility in stock markets. By analogy, uncertainty in international economic conditions could similarly elevate the risk premium within the housing market, rendering it more volatile and less appealing to potential buyers. Consequently, this heightened volatility and perceived risk could suppress demand, thereby exerting downward pressure on house prices. The adverse effect of a higher lTGDP on lHP contradicts the expectations of the authors based on the research results of refs. [62,63,64] and it is also some sort of common knowledge. However, refs. [65,66] showed that tourism development has no significant impact on house price growth. It is apparent that the upward path of housing prices often coincides with the development of tourism. However, an isolated analysis of the nexus between tourism expansion and housing prices, without due consideration of other contextual variables, can lead to an inflated estimation of the influence of tourism development on housing market dynamics.

Model 4 extends the baseline specifications by introducing dlREER to capture the currency-driven cost effects on lHP. The added dlREER term also crosses the threshold positively and significantly, suggesting that real appreciation of the domestic currency, by reducing the cost of imported construction materials, exerts additional upward pressure on house prices ref. [16]. Together, these results highlight how both external exchange rate movements and inward remittance flows jointly shape the evolution of EU house prices.

Model 5 extends the full specification by incorporating dlPCON. In the long-run analysis, the coefficient of dlPCON indicates that a 1% increase in construction output leads to a 7.05% reduction in lHP, which is consistent with supply-side moderation effects ref. [14].

Model 6 completes the supply-side dimension by adding dlPER to the full specification. In the long-run analysis, the coefficient of dlPER indicates that a 1% increase in new permits ultimately leads to a 3.09% reduction in lHP, consistent with the moderating effect of anticipated supply expansions on long-term price levels ref. [14].

Conversely, in Model 7, lFGDP negatively and statistically significantly impacts lHP, indicating that a higher lFGDP tends to be associated with a lower lHP. This counterintuitive result likely reflects a supply-side effect, where FDI directed into housing development expands the available stock and thus puts downward pressure on prices. Empirically, Ref. [18] find that FDI inflows into Malaysia between 1999 and 2015 had a statistically significant negative impact on residential property prices. This result is not consistent with economic theory, which suggests that higher levels of lFGDP relax domestic lending standards and bolster credit supply through complementarities between FDI and domestic financial markets ref. [67], stimulating housing demand and subsequently driving up prices—as evidenced by panel–VAR analysis showing that foreign real estate investments significantly increase house prices in 21 emerging economies [68].

The final model, Model 7, augments the previous specifications by also introducing dlRIR. dlRIR has a substantial and persistent negative impact on lHP levels in the long run. A permanent 1% increase in dlRIR is associated with a 0.5771% reduction in the long-run level of lHP. Higher real borrowing costs discourage mortgage demand and raise the user cost of housing, thereby pulling down equilibrium house prices. This result is consistent with ref. [17], who find that in Central and Eastern European markets, real interest rates are a key fundamental driver of long-run house price dynamics.

Interesting insights emerge when examining short-term results. Across all observed equations, lRGDP exhibits a negative and statistically significant impact on lHP, except in the case of the first and sixth models where the results are insignificant. Furthermore, as far as dlGDP is concerned, it also, in the short term, negatively and statistically significantly affects European lHP in the five models. In the short term, it is necessary to highlight only lEF, which in the third model has a positive and statistically significant effect on lHP. In Model 4, the short-term coefficient of dlREER is negative and statistically significant, implying that a 1% of the domestic currency is associated with an immediate 22.03% reduction in house price growth, underscoring the strong near-term dampening effect of currency strength on the housing market. Model 6 is also relevant to mention here. In the short-run dynamics, the first dlPER implies that a 1% rise in dlPER is associated with an immediate 1.56% increase in lHP. Although only marginally significant, this result suggests that increased permit activity can temporarily boost market expectations and price momentum. Upon analyzing the short-term coefficient results of the variables, it can be concluded that their impact is less significant than that of the long-term coefficients. Additionally, the strength of these effects grows with an extended temporal duration.

The next step was to check causality among the variables; in order to do so, ref. [39] was applied. In

Table 8. Causality analysis.

Null Hypothesis	W-Stat	Z-Bar Stat	p-Value
DLRGDP ≠> LHP	10.1673	2.68502	0.0073 ***
LHP ≠> DLRGDP	6.79207	0.93119	0.3518
LFGDP ≠> LHP	7.10939	1.97251	0.0486 **
LHP ≠> LFGDP	6.62553	1.60675	0.1081
LTGDP ≠> LHP	6.69681	1.48641	0.1372
LHP ≠> LTGDP	4.78965	0.15095	0.8800
LR ≠> LHP	6.66510	1.63666	0.1017
LHP ≠> LR	4.58611	0.06510	0.9481
DLNM ≠> LHP	27.6530	11.7709	0.0000 ***
LHP ≠> DLNM	4.31231	−0.35734	0.7208
LEF ≠> LHP	10.4613	4.50630	7 × 10−6 ***
LHP ≠> LEF	9.10044	3.47761	0.0005 ***
DLGDP ≠> LHP	4.92780	−0.03752	0.9701
LHP ≠> DLGDP	7.47153	1.28425	0.1991
DLREER ≠> LHP	7.85907	1.48562	0.1374
LHP ≠>DLREER	5.87119	0.45268	0.6508
DLPCON ≠> LHP	8.66833	1.90612	0.0566 *
LHP ≠> DLPCON	8.35452	1.74306	0.0813 *
LPER ≠> LHP	15.0457	7.97180	2 × 10−15 ***
LHP ≠> LPER	8.40911	2.95501	0.0031 ***
DLRIR ≠> LHP	5.98032	0.50939	0.6105
LHP ≠> DLRIR	8.87891	2.01554	0.0438 **

Note: ***, **, and *, respectively, denote significance at the 1%, 5%, and 10% levels. In the parentheses, the p-values of each test are reported. The symbol “≠>” denotes “does not linearly Granger-cause”.

, the results are presented.

Based on the pairwise Dumitrescu and Hurlin ref. [39] results obtained in Table 7, several causal relationships can be confirmed. As a result of the analysis, a unidirectional causal relationship was found from dlRGDP to lHP, from lFGDP to lHP, from lNM to lHP, and from lHP to dlRIR. Bidirectional causality was found between lEF and lHP, between lHP and lPER, and between lHP and dlPCON. From the first causal link, it can be deduced that an increased dlRGDP simultaneously generates additional demand for houses and thus exerts upward pressure on prices. However, these results contradict the findings of ref. [33], who found no causality in their study on property prices in Nairobi, Kenya. Furthermore, these findings challenge the claims of ref. [11], who asserts that remittances contribute to increased housing supply and thus lower housing prices. Nevertheless, ref. [27] confirmed the impact of remittances on the real estate market in Jordan in his research. In addition, ref. [26] demonstrated the causal link between remittances and the escalation of housing prices in their study. The second causal link states that lFGDP affects lHP by boosting economic growth, which increases incomes and demand for housing, which in turn influences prices. Furthermore, ref. [69] confirmed this relationship by proving that housing prices increase as a result of FDI inflows. The third causality relationship from lNM to lHP implies that demographic inflows directly influence property market dynamics beyond traditional demand-side factors. Specifically, surges in net migration elevate local housing demand, placing upward pressure on prices, particularly when supply remains constrained. Ref. [70] confirms that immigration shocks significantly enhance the predictive accuracy of house price models, especially VAR specifications, highlighting migration as a critical signal for housing market analysis. A last unidirectional causal relationship is detected from lHP to real interest rates dlRIR. This implies that developments in the property market may influence central bank policy responses. In particular, rapid increases in housing prices can signal economic overheating, prompting monetary authorities to raise real interest rates. Ref. [13] find that even minor policy adjustments in low-rate environments can trigger substantial corrections in housing markets, highlighting the importance of housing prices as a policy signal.

The first bidirectional causal relationship is found between lEF and lHP. Greater economic freedom, characterized by less government intervention and stronger property rights, leads to higher property prices. This bidirectional relationship suggests that economic freedom can drive up housing prices. Conversely, housing prices can be a sign of economic prosperity, which attracts investment and further strengthens economic freedom. Currently, there are no papers that deal with causality between economic freedom and housing prices besides ref. [71] who found that economic freedom is positively related to housing values. A second bidirectional causal relationship is found between lPER and lHP. This suggests that greater building activity, proxied by permit approvals, drives housing supply, which influences pricing dynamics. Conversely, increasing housing prices can encourage new construction, reflected in rising permit issuance. This interaction is partially supported by ref. [14], who examine the dual role of demand and supply factors in shaping housing prices in Europe, although they report that supply-side indicators often lack strong long-run significance. The test also confirms a third bidirectional causal relationship between dlPCON and lHP. Increased economic activity in the construction sector may indicate housing booms, driving price growth. In turn, rising housing prices can stimulate construction through wealth effects and increased investment. This cyclical linkage aligns with ref. [15], who emphasizes the role of monetary policy and real estate cycles in shaping both housing production and consumption behavior.

In contrast, no significant causality is observed between dlREER and lHP. This indicates that fluctuations in external competitiveness do not directly affect domestic residential property values. Ref. [16] support this view, arguing that exchange rate dynamics play a limited role in housing markets compared to internal factors such as credit growth and income levels. Ref. [17] similarly note that exchange rate dynamics exert only a marginal effect on housing prices in these regions, especially when compared to internal drivers like income growth and credit expansion.

6. Conclusions

The widespread review of the literature highlights diverse findings regarding the relationship between remittances and house prices, drawing from studies across regions such as Kenya, Ghana, Latin America, Nepal, Pakistan, Ecuador, and Colombia. Authors like refs. [20,26] reveal a positive correlation between remittances and housing prices, emphasizing the stimulating effect of remittance inflows on real estate growth. Conversely, studies by refs. [32,33] challenge this correlation in Nairobi, Kenya. Insights from Nepal and Pakistan by refs. [8,10,21,28,34] confirm the significant impact of remittances on housing demand and prices. However, research in rural Ecuador by ref. [35] and Colombia by ref. [11] presents nuanced perspectives, indicating varied effects of remittances on housing markets. Overall, the literature underscores the complex interdependence of economic conditions, policy frameworks, and regional dynamics in shaping the relationship between remittances and real estate prices.

The main focus of this study was to analyze the impact of remittance on the EU’s house prices. Parallelly, based on previous research, some other important variables are included in the analysis. The research findings of this study highlight the complicated intertwining of economic factors that include the relationships between remittance, migration, other included economic factors, and housing market prices within the European Union context. The analysis reveals permanent steadiness adjustments among the variables over time, as evidenced by consistently negative and statistically significant error correction terms (ECTs). Particularly, an increase in dlRGDP positively influences lHP within the EU, alongside a significant positive effect of dlGDP on lHP. Moreover, higher levels of lNM contribute to increased lHP, highlighting the relationship between economic factors and demographic trends. Contrary to expectations, the study shows that a higher lTGDP negatively impacts lHP in the EU. This finding challenges prior research and common knowledge, suggesting that an isolated examination of tourism development may overestimate its influence on housing market dynamics. Despite the parallel rise in housing prices with tourism development, nuanced analysis considering contextual variables is crucial for accurate assessment. Economic freedom, as represented by lEF, negatively influences lHP. Although empirical evidence remains limited, theory suggests that sensitive global economic uncertainty elevates the housing market’s risk premium and increases price volatility. The resulting erosion of buyer confidence suppresses demand and thus drives down equilibrium house prices. Additionally, higher levels of lR positively impact lHP, reflecting increased investor confidence and access to credit, stimulating housing demand and driving prices upward.

A negative impact on lHP is also observed for dlPCON, lPER, and dlRIR. These extended models provide strong empirical support for the relevance of external, supply-side, and financial determinants in shaping housing price dynamics across EU countries. Conversely, increases in dlPCON and lPER significantly reduce long-run lHP, confirming the expected moderating effects of supply-side expansion. Lastly, dlRIR is shown to have a persistent and economically meaningful negative effect on lHP, highlighting the critical role of monetary conditions in housing affordability. Collectively, these findings underscore the multidimensional nature of housing price formation in the EU and offer valuable insights for policymakers aiming to balance affordability, investment, and economic growth. Furthermore, dlREER positively influences lHP. Specifically, dlREER exerts upward pressure on lHP by lowering import costs for construction inputs, reinforcing the link between external competitiveness and domestic real estate markets.

Furthermore, pairwise causality analysis confirms causal links from dlRGDP, dlFGDP, and lNM to lHP. Conversely, lHP affects lPER and dlRIR. Bidirectional causality is observed between lEF and lHP and dlPCON and lHP, highlighting the reciprocal relationship between those variables.

These findings contribute to a deeper understanding of the complex interactions shaping housing market dynamics within the EU context, informing policymakers, researchers, and market participants alike. However, further research is necessary to explore the refinements of these relationships and their implications for housing market policies and interventions. Academic community and policymakers often overlook the use of remittances to finance housing investment, treating them as unproductive or with little or no impact on development ref. [72]. However, refs. [29,73] has shown in his studies that investments in housing not only improve the living conditions and well-being of migrant households, but also serve as a reasonably safe investment and an opportunity to generate additional income, which contributes to a better future existence. Therefore, such housing investments have direct and indirect effects that can favorably influence the social and economic development of the areas from which the migrants originate. Investments in housing can open up further opportunities for investment and knowledge transfer in their home communities, supported by family and community networks. Consequently, given the size and continuous dynamics of migration flows, migrant investments in real estate can have a significant impact on local real estate markets. These migration incentives may contribute to increased demand for real estate in certain areas, which may result in higher real estate prices. Also, such investments can shape the structure of the real estate market and affect the availability and affordability of housing for the local population. Therefore, it is important to take into account the role of migrants and their investments in real estate when analyzing and planning policies related to the real estate market.