War in Ukraine, the refugee crisis, and the Polish housing market

Abstract

The ongoing war in Ukraine has become a global issue and caused a major refugee crisis in Europe. The displacement of millions of people from Ukraine, who entered neighbouring countries, attributed to a housing demand shock in host cities and subsequent increases in rents and prices.

The article investigates the housing market reaction in Poland’s five largest cities caused by the arrival of refugees from Ukraine following the Russian invasion in February 2022. We use a difference-in-difference quasi-experimental scenario to test whether exposure to mass refugee inflow translates to housing market dynamics.

According to our findings, an increase in a city population of 1 pp caused by the inflow of refugees led to a 0.72–0.74% increase in housing rents. Additionally, we found some evidence that the arrival of migrants may have slightly increased apartment prices; however, the impact is smaller than in the case of rents and statistically significant only in selected specifications.

The paper contributes to the literature on the impact of immigration or refugee inflow on housing market dynamics. The rent increases reported in this study may have serious housing policy implications, both in the short and mid-run.

Keywords:

• Housing market
• refugees
• war
• Ukraine
• housing rents
• house prices

1. Introduction

The Russian invasion of Ukraine in February 2022 will likely have profound and global economic and political consequences both in the short and long run. Early economic evidence suggests that war affected the oil market and created spillovers to other financial and commodity markets (Adekoya et al., 2022; Boungou & Yatié, 2022; Tosun & Eshraghi, 2022) as well as concerns about energy security and debate on energy transition (Żuk & Żuk, 2022). However, the effects of the War in Ukraine are more significant at a regional scale. The Russian invasion of Ukraine is the most potent military conflict since the Second World War and has already generated the biggest refugee crisis in Europe. Due to geographical proximity, neighbouring Central and Eastern European countries accommodated most Ukrainian refugees. According to official statistics, more than 8 million people (primarily women and children) have fled from Ukraine since the beginning of the war. Approximately 4.8 million refugees registered for Temporary Protection or equivalent protection in Europe, of which 1.56 million registered in Poland as of 15 February 2023 (UNHCR, 2023). The mass exodus has multiple economic, social, and policy implications for the host country, including the labour market, primary and secondary education system, and housing market. The demand shock created by the mass inflow of refugees is also transmitted to metropolitan residential markets, most likely affecting rents, prices, and housing affordability.

Most refugees who decided to stay in Poland sought housing in major Polish cities: Warsaw, Krakow, Lodz, Wroclaw, and Poznan. The critical determinants of host city choices made by Ukrainian refugees were familiarity with the host cities, the presence of the Ukrainian diaspora, geographic proximity, and labour market possibilities.

The paper aims to evaluate the impact of the inflow of refugees on house prices and rents in five major cities in Poland (Warsaw, Krakow, Wroclaw, Lodz and Poznan). In the empirical part of the paper, we test two hypotheses. Firstly, we hypothesise that the refugee crisis substantially affected rents more than prices. Secondly, we speculate that the effect’s size varied concerning the impact’s magnitude, with cities experiencing the most significant inflow of refugees being affected the most.

The methodological approach is straightforward. Based on a unique dataset on housing offers, we investigate house price and rent dynamics before and after the beginning of the military conflict in Ukraine. We use a Difference-in-Differences (DiD) quasi-experimental setting to evaluate whether the stimuli significantly influenced housing rents and prices. Due to time-varying exposure to refugee inflows, differences between cities, and lack of a pure control group, we apply a recently developed modified DiD framework with continuous treatment that accounts for the fact that all major cities in Poland have been affected by the inflow of Ukrainian refugees.

The paper adds to the housing literature on the consequences of mass migration or refugee inflows in three key areas. Firstly, the research bridges the gap in empirical evidence on the short-term, immediate effect of immigration on housing markets. Compared to prior literature, we focus on the short-term, temporal effect observed in the first months of immigration and study market responses to the positive demand shock. The immediate impact was not covered in the existing literature. Secondly, from a methodological perspective, due to the heterogeneity of the magnitude of the effect of refugee inflow on metropolitan housing markets, we use the DiD approach with a continuous treatment in our empirical setup. In contrast, most of the prior papers in the field used a classic, binary DiD approach.

Last but not least, the dynamics and consequences of the military conflict in Ukraine are unprecedented in scale; therefore, several conclusions drawn from previous studies may not fully apply to the current situation in European countries affected by the refugee crisis. Few papers addressed various housing implications of the War in Ukraine, and none explored the links between the refugee crisis and housing market dynamics. To our knowledge, it is the first paper to directly investigate the impact of the influx of Ukrainian refugees on housing market dynamics. Trojanek and Gluszak (2022) also addressed the causal effect of the War in Ukraine on Poland’s housing; however, due to a lack of data, they did not directly link price and rent dynamics to the magnitude of refugee displacement.

We believe the research findings will show a new perspective on the effect of abnormal events on housing market dynamics and affordability.

The rest of the paper is organised as follows. The following section critically assesses empirical work on the linkages between military conflicts and housing markets. In the methodology section, we address data collection and empirical approach. In the results section, we present findings related to the impact of the refugee crisis on house prices and rents in major cities in Poland. In particular, we seek to find whether the crisis had a particular effect on cities most affected by the mass exodus of Ukrainians. In the last section, we try to evaluate prior research findings. We also identify potential future research themes, both short-term and long-term.

2. Literature review

Wars create short-term disturbances in the operation of the housing market, but some consequences may be persistent and change the urban housing markets in the long run (Ambrose et al., 2013; Eichholtz, 1997; Meen et al., 2016). Several empirical studies addressed the economic consequences of military conflicts on house prices and rents. Historical research shows that the first three Anglo-Dutch Wars in the seventeenth century directly affected housing rents in Amsterdam. Still, they also affected the city’s economy and demographics in the following decades (Eichholtz et al., 2012). Similar conclusions may be drawn from Karagedikli and Tunçer’s study on the rental market dynamics in Edirne (one of the major cities of the Ottoman Empire) from 1720 to 1814. Empirical evidence shows that the Russo-Turkish War in the eighteenth century generated demographic shocks that contributed to the rent decreases (Karagedikli & Tunçer, 2021). Kholodilin et al. (2021) investigated the impact of the outbreak of World War I on the housing market in St. Petersburg. They documented that significant population growth, triggered by refugee flows, led to a rapid rent increase. The explosive rise in rents was stopped by introducing rent controls in 1915. Negative and positive population shocks were also instrumental to rent dynamics in Berlin during World War I (Kholodilin, 2016).

Knoll et al. (2017) demonstrate World War I and II’s temporal and persistent negative impact on European housing markets. Relatively few economic studies explored the consequences of military conflicts on housing markets. Scarce evidence indicates that ongoing military operations or risk of terrorism have a detrimental effect on house values in the affected locations. Arbel et al. (2010) analysed the dynamics of housing prices in Jerusalem before and after the Second Intifada. They investigated residential sale prices in the Gilo settlement in Jerusalem from 1997 to 2008. The results demonstrate a delayed but statistically significant reduction in house prices (10–12% depending on the model) in the area under Palestinian heavy machine gun fire. The price reduction persisted for another five years after the uprising. Another Jerusalem study (Hazam & Felsenstein, 2007) indicates that the proximity of the Separation between the Israeli part of Jerusalem and the Palestinian Authority territory negatively affects house prices (due to the increased risk of terrorist attacks). Other studies on the impact of terrorist threats on the real estate market confirm these findings. Abadie and Dermisi (2008) found that after 9/11 the attack on the WTC in New York, an increase in vacancies was observed in office buildings located in the vicinity of Chicago’s most famous skyscrapers (Willis Tower, Hancock Center) due to the potential threat of a terrorist attack. Empirical research in Northern Ireland also confirms the impact of terror threats on house prices (Besley & Mueller, 2009). More recent evidence from the Syrian civil war (started in 2011) demonstrates that military conflicts decrease housing affordability, alter urban development patterns, and foster chaotic urban development and the emergence of informal settlements (Wind & Ibrahim, 2020). Strategic challenges related to the future housing policy that will provide sustainable incentives for the renewal of housing stock in Ukraine after the war were discussed by Shcherbyna (2022).

Direct housing market consequences of wars in areas affected by military operations are well documented. However, war conflicts, revolts, and political turmoil cause mass migration and refugee flow to other regions and countries, creating spillover effects on local and regional housing markets. Massive inflows of refugees, asylum-seekers and migrants generate demand shocks affecting rents and house prices. The evidence of those indirect effects blends within a broader category of research on the impact of mass immigration on the housing market. Evidence from the United States suggests that the immigration shock caused by the exodus of Cuban immigrants in 1980 (Mariel boatlift) contributed to the significant rent increase in Miami, but only in the case of low-quality housing stock (Saiz, 2003). The results were confirmed in a follow-up study investigating the impact of immigration in 306 metropolitan areas in the US. They found a positive effect of immigration on housing, both rents and house prices (Saiz, 2007). A positive but limited influence of immigration on house prices in Canada was reported by Akbari and Aydede (2012).

Recent evidence from the Syrian refugee crisis shows that immigration caused rent increases in Jordan, especially in low-income areas (Alhawarin et al., 2021). Evidence from Turkey shows that immigration from Syria may also contribute to increases in the high-quality residential sector due to the housing sorting mechanism (Balkan et al., 2018). Authors argue that the local population reacted to the refugee crisis by relocating from the districts affected by the mass inflow of migrants to more expensive neighbourhoods, resulting in substantial rent and price increases in the latter case. Similar conclusions come from earlier European studies. The empirical evidence from Italy (Accetturo et al., 2014) and the UK (Lastrapes & Lebesmuehlbacher, 2020; Sá, 2015) suggest that immigration hurts urban areas affected due to segregation effects and mobility of local population who relocates from immigrant-dense districts to other parts of the city. Negative attitudes towards immigrants, refugees and asylum seekers also manifest at the micro-level. Recent studies indicate that refugee housing projects generate negative externalities at the neighbourhood level (van Vuuren et al., 2019) or asylum seekers’ reception centres (Daams et al., 2019) can have a detrimental impact on house prices in proximity. Research based on a recent large influx of refugees to Germany in 2014 and 2015 suggests that the location of shelters for refugees leads to rental prices decline in the neighbourhood (Hennig, 2021) and can fuel negative sentiments and reactions (Endrich, 2023; Entorf & Lange, 2023).

The economic literature on the recent Ukrainian refugee crisis remains relatively underdeveloped. Few empirical studies addressed the effects of the war in Ukraine in 2022 on housing demand, house prices and rents in Poland. Trojanek and Gluszak (2022) demonstrated that the War in Ukraine and the subsequent mass exodus of refugees initially resulted in a swift reaction in the rental market in Polish cities but did not significantly affect house prices. Their study focused on an aggregated time series describing the housing market in Warsaw and Krakow.

To our knowledge, few (if any) housing-oriented studies addressed the ongoing refugee crisis’s direct impact on housing market dynamics in neighbourhood countries after Russia invaded Ukraine. The paper attempts to fill the gap in the empirical evidence on the topic.

3. Methods and data

3.1. Housing market data

In Poland, neither the Statistical Office (SO) nor the National Bank (NBP) publishes monthly housing and rental price information. Moreover, depending on the city, the Property Price Registry housing sales data are only available with a substantial delay (R. Hill et al., 2023). Furthermore, renting agreements are generally private, and no official rent microdata is available. As a result, based on transactions alone, it is impossible to keep track of continuous developments in monthly periods in the housing and rental markets in Poland.

Several problems related to the availability and validity of housing sales and rental information may be tackled using listings (asking) data. Using housing listings as a source of information in housing studies has a long and established tradition (Pollakowski, 1995). In recent years, the interest in using offer prices/rents has increased substantially (Diewert & Shimizu, 2016; Hahn et al., 2022; Loberto et al., 2018; Pfeifer & Steurer, 2022; Trojanek et al., 2023). Several studies indicate that the offer data may be an adequate and reliable substitute when transaction data are unavailable, or its quality is compromised (Anenberg & Laufer, 2017; Kolbe et al., 2021; Lyons, 2019; Shimizu et al., 2016; Wang et al., 2020).

The study was based on offer data collected monthly from the popular announcement portal (gratka.pl). The data on asking prices (221,458 observations) and asking rents (112,260 observations) for the five biggest cities in Poland (Warsaw, Krakow, Lodz, Wroclaw and Poznan) were gathered from January 2021 to May 2022. Warsaw is Poland’s capital and the country’s largest housing market, with many new developments and housing demand fueled by strong economic growth. Krakow is the second largest city in Poland and a popular destination for investors and property buyers due to its historical heritage and a growing economy (thriving due to business process outsourcing, shared services and the IT industry). Lodz is located in the centre of Poland and has transformed an industrial city into a cultural and business centre. Many loft conversions and industrial spaces are turned into residential units compared to other cities. Wroclaw is one of the fastest-growing cities in Poland, an academic and cultural centre with a thriving housing market. Poznan is a city in western Poland with a growing economy and a stable housing market. In all these cities, the rental market is robust, and the demand is generated by young professionals, expatriates, and students, especially in Warsaw, Krakow and Wroclaw.

To ensure validity and accuracy, the listing data requires adequate and sequential preparation and auditing (Trojanek et al., 2023). First, we removed observations with missing information on key variables. Then, based on detailed information about the location (street), the offer was assigned to a district. In the next step, duplicate offers were removed, leaving in the database the ad with the lowest price in a given month. Finally, we excluded outliers from the dataset by restricting apartment size and keeping only apartments with a usable area between 20 and 200 square meters. Identifying atypical observations is crucial and necessary to properly carry out further stages of analysis (Su & Tsai, 2011) because outliers may lead to biased results and inappropriate interpretations (Rousseeuw & Hubert, 2011; Winson-Geideman & Krause, 2016).

Given the above, we address this issue in our research based on Cook’s distance. We independently computed hedonic models of log price for each market in each city, considering the available key features (spread to the city centre, size, construction technology, age and quality of apartments and floor where they were located) to remove outliers from the dataset. After all, the dataset decreased to 35,950 sale offers and 28,221 rental offers after the cleaning procedures. Table A1 in the Appendix presents the data characteristics before and after cleaning.

The monthly housing and rental price indexes in 2021–2022 were estimated using the time-dummy hedonic method (R. J. Hill et al., 2018; R. Hill & Trojanek, 2022; Waltl, 2016, 2019; Zhang & Yi, 2017). To control changes in the structure and quality of apartments for sale/rent in monthly periods and determine the dynamics, we used several covariates that account for locational and structural quality heterogeneity (distance from the city centre, area, age and the construction technology of the building, height of the building, apartment’s location on the floor, and quality of individual apartments). Tables A2 and A3 in the Appendix present all control variables used in the study.

3.2. Refugees data

The exact measurement of the number of refugees staying in Poland is problematic. For obvious reasons, the residence and period of stay are not always registered, and official statistics do not cover the whole phenomenon. Nevertheless, records confirmed that the economic and societal challenges and burdens related to the inflow of refugees from Ukraine are not equally distributed in space. Poland has been affected disproportionately, receiving more than 1.5 million refugees during the first month of the War (Kardas et al., 2022). This mass inflow of refugees was mainly observed in the first weeks after the Russian invasion until mid-March 2022 (Figure 1). In the first year of military operations, approximately 9 million Ukrainians entered Poland. Not all of them stayed in Poland; however, as of January 2023, according to official statistics, 976 thousands citizens of Ukraine have been assigned a Polish personal identification number (PESEL). Most of the refugees from Ukraine were women and children. According to the survey conducted by the Centre for Analysis and Research of the Union of Polish Metropolises, finding accommodation is a most crucial issue, at least for 28% of Ukrainians who completed the questionnaire (Wojdat et al., 2022).

Figure 1. Daily border traffic between Ukraine and Poland from 24 February to 31 May 2022.

The housing situation of Ukrainians who decided to stay in the 12 biggest cities in Poland during the first two months after the Russian invasion of Ukraine was diversified, but 98% remained in the city where they registered for PESEL. It is safe to say that despite self-organised help, hospitality, and limited public support, many refugees paid for accommodation in the first phase of their stay in Poland. Approximately 19% of refugees were renting an apartment on the market, 23% shared apartments with host/friends from Ukraine, 20% lived with Polish hosts, and 18% lived in apartments without hosts The rest stayed at refugee collective assistance facilities or hotels (Wojdat et al., 2022) (Figure 2).

Figure 2. The share of Ukrainians in the population of selected cities in Poland (May 2022).

The value of the treatment variable (REF) should match the observed refugee inflow in that particular city and the month since the beginning of the war in Ukraine. The estimates on the number of Ukrainian refugees staying in Polish cities were obtained (Wojdat & Cywiński, 2022; Wojdat et al., 2022). The estimates of the number and location of Ukrainians residing in Poland are based on a geotrapping algorithm. The method involves using information from mobile devices and GPS geospatial tools to track and record the movement and activities of individuals in specific geographic areas. This spatio-temporal data was later used to assess the number of Ukrainian refugees moving to and residing in selected Polish cities. Since we only have access to the refugee flow data covering the first three months since the beginning of the war (March, April and May 2022), it is impossible to compare them directly to compatible statistics before the Russian invasion. To calculate the net inflow of Ukrainian refugees, we compare the geotrapping estimate to the number of Ukrainians registered in the Social Insurance Institution in Poland (ZUS) at the end of December 2021. The basic summary of the data on Ukrainian refugees in Krakow, Lodz, Poznan, Warsaw and Wroclaw is presented in Table 1.

Table 1. Ukrainians in selected cities in Poland before and after the Russian invasion of Ukraine.

City	Dec 2021*			March 2022**		April 2022**		May 2022**
	Population	Ukrainians	%	Ukrainians	%	Ukrainians	%	Ukrainians	%
Krakow	780,0	30,7	3.8%	177,6	18.5%	196,5	20,1%	180,9	18.8%
Łodz	672.2	16,4	2.4%	85,7	11.3%	96,8	12,6%	66,6	9.0%
Poznan	532,0	25,8	4.6%	84,6	13.7%	106,2	16,6%	99,2	15.7%
Warsaw	1794,2	103,8	5.5%	266,9	12.9%	326,5	15,4%	352,1	16.4%
Wroclaw	641,9	43,9	6.4%	187,3	22.6%	247,6	27,8%	250,2	28.0%

Note: * Source: The data on Ukrainians come from The Social Insurance Institution ZUS (2021) Online: https://www.zus.pl/baza-wiedzy/statystyka/opracowania-tematyczne/cudzoziemcy ** Source: The data on Ukrainians come from Wojdat et al. (2022) Online: https://metropolie.pl/fileadmin/news/2022/07/Urban_hospitality_update.pdf.

We investigate the impact of the refugee crisis on the housing market using monthly data from January 2021 to May 2022. We construct a time variable concerning the beginning of the war in Ukraine in February 2022 (t = 0); therefore, t ranges from −13 (13 months before the start of the Russian invasion) to +3 (three months after). Finding a city in Poland unaffected by the refugee crisis is impossible. To some degree, all cities were affected by refugee inflow; thus, we assume that the treatment variable is continuous, not binary. The treatment exposure may be measured as a relative inflow of refugees from Ukraine. The number will be proportional to the number of refugees finding accommodation in a given city (i) at a given time (t) and inversely proportional to the size of the city. In the paper, we assume that the number of Ukrainian refugees fleeing from the country would be 0 before the refugee crisis began after Russia invaded Ukraine on 24 February 2022 ($t\le 0)$. Although one could argue that some Ukrainians might have left their country before the beginning of the war, during the military build-up at the end of 2021 or the start of 2022, the official data suggest that no massive cross-border activity was registered before the military operation. We believe the assumption realistically reflects the actual data.

For each of the three months after the beginning of military operations in Ukraine (March 2022, April 2022 and May 2022, t > 0), we calculate the net inflow of Ukrainian refugees by subtracting the baseline number of Ukrainians that were registered in the Social Insurance Institution in Poland (ZUS) before the war (UKR_0i) from the geotrapping estimate of the number of Ukrainians in city s at month t (UKR_ts). Finally, we calculate the treatment variable REF_ti, by dividing the net inflow of Ukrainian refugees by the city’s official population before the war ${\mathit{\text{POP}}}_{0s}$ using the latest available census data. (1) $${\mathit{\text{REF}}}_{\mathit{\text{ts}}}=\left\{\begin{array}{c}0,t\le 0\\ \frac{{\mathit{\text{UKR}}}_{\mathit{\text{ts}}}-{\mathit{\text{UKR}}}_{0s}}{{\mathit{\text{POP}}}_{0s}},t>0\end{array}\right.$$(1)

The value treatment variable REF can be interpreted as a relative increase in the pre-war city population due to the inflow of Ukrainian refugees. The values of the continuous treatment variable REF are presented in the Appendix (Table A4). The following section, 3.3, describes the quasi-experimental setting.

3.3. Econometric approach

The Russian invasion of Ukraine started on 24 February 2022 and caused an unprecedented outflow of refugees that crossed borders and entered neighbouring states – Poland, Slovakia, Hungary, and Moldova. Based on available data and geographic distances, we assume that the timing of migration shock for the host country is the same for all Polish cities in our sample. Thus, in our specification, we divide our sample into periods preceding the Russian invasion, starting from January 2021 and ending in February 2022 (before) and succeeding periods from March 2022 to May 2022 (after).

The research investigates the causal effect of Ukrainian refugees’ inflow on rents and prices in selected Polish cities. We use hedonic regression DiD with continuous treatment, a generalisation of the traditional binary treatment approach. In a canonical representation of this quasi-experimental method, we typically track the outcome of interest in two groups – treated (subject to certain interventions) and control group (not subjected to the treatment) in two time periods – before and after the application of the treatment. The DiD with continuous treatment (that accounts for variation in treatment intensity) enables us to evaluate causal effects when there is no untreated comparison unit. In our case, it stems from the simple fact that all major cities in Poland (including those selected in our study) have been affected by the inflow of Ukrainian refugees after the Russian invasion of Ukraine. It is, therefore, reasonable to assume that all units (cities) were subject to a treatment to some extent.

Quasi-experimental methods, including DiD, have been widely used in urban and housing studies to evaluate the impact of significant events on selected housing indicators (housing construction, house prices). In particular, the approach is useful when investigating the causal influence between creating (dis)amenities and property values in proximity (Cheung et al., 2018; Fink & Stratmann, 2015; Gibbons & Machin, 2005; Trojanek & Gluszak, 2018). Another popular application is the market response to introducing certain housing market policies and other economic, sports and political events (Coates & Matheson, 2011). Quasi-experimental methods were also used in prior research on immigration effects on urban housing markets.

Since its creation, the method has evolved. In the last ten years, significant progress has been made in adapting the standard version of DiD to other scenarios to evaluate the effect of the treatment over multiple periods, (ii) variation in treatment size, treatment timing (Callaway et al., 2021). Modified DiD specification allowed for continuous treatment, multiple periods before and after intervention (de Chaisemartin & D’Haultfœuille, 2018; de Chaisemartin & D’Haultfœuille, 2020; de Chaisemartin et al., 2019). In particular, continuous treatment allows treatment evaluation in scenarios where treatment is distributed to all units, and the true control group does not exist. To account for the heterogenous treatment, we estimate a difference-in-difference model with continuous treatment effect based on cross-sectional data on housing rents(prices): (2) $$y_{\mathit{\text{ist}}}={\gamma }_s+{\gamma }_t+x_{\mathit{\text{ist}}}\beta +{+\mathit{\text{REF}}}_{\mathit{\text{ts}}}\delta +{\epsilon }_{\mathit{\text{ist}}}$$(2)

Where i refers to the individual housing offer, s denotes a city-level index, and t is a monthly time index. In Equation (2)(2) $$y_{\mathit{\text{ist}}}={\gamma }_s+{\gamma }_t+x_{\mathit{\text{ist}}}\beta +{+\mathit{\text{REF}}}_{\mathit{\text{ts}}}\delta +{\epsilon }_{\mathit{\text{ist}}}$$(2) , ${\gamma }_s$ are city-level fixed effects, effects, and ${\epsilon }_{\mathit{\text{ist}}}$ is the error term. Dependent variable $y_{\mathit{\text{ist}}}$ is a natural logarithm of offer rent (or price) for an apartment observed in a given city and month. Covariates $x_{\mathit{\text{ist}}}$ are locational and structural characteristics of housing offers (see Appendix Table A2). Coefficient $\beta$ captures the effect of a particular attribute on rent or price. Finally, ${\mathit{\text{REF}}}_{\mathit{\text{ts}}}$ is a continuous treatment variable that varies at the city and time levels. Coefficient δ captures the average treatment effect on the treated (ATET). The treatment variable is based on a relative change in the population of a given city subject to the inflow of refugees. That allows us to estimate the more precise measure of house price and rent elasticity related to immigration inflow. The timing of the treatment is similar in all cities. The empirical results are presented and discussed in the following section: 4.

4. Results and discussion

4.1. Apartments’ rents

We use a difference-in-differences model based on cross-sectional housing market data to evaluate the impact of the mass inflow of refugees on rents in five cities in Poland. The total sample size was 28,221 housing rental offers. In the repeated cross-sectional hedonic regression model, we controlled for selected locational and structural characteristics of residential properties. That allowed us to account for apartments’ quality heterogeneity. The monthly hedonic rent indices in the study period are presented in Figure 3.

Figure 3. Monthly hedonic rent indices from January 2021 till May 2022.

Graphical evaluation reveals that trends in apartment prices were approximately parallel before the Russian invasion (before the treatment period). Furthermore, although we could not directly test for parallel trends and anticipatory effects using the Granger causality test (no untreated control group available), visual inspection suggests that rent dynamics have changed since the beginning of the War in Ukraine. Cities affected the most by the mass inflow of refugees (Wroclaw, Krakow) experienced the highest rent increases shortly after.

We apply a quasi-experimental DiD setting with continuous treatment to assess this effect directly. Our treatment variable represents the different exposures to refugee inflow across cities and over time. We use two specifications of the model. The dependent variable is a natural logarithm of apartment offer rent in the first. In the second model, the dependent is a natural logarithm of apartment offer rent per sqm. The results of DiD model estimation are presented in Table 2.

Table 2. Difference-in-differences hedonic regression for offer apartment rents in selected cities in Poland.

	Dep. var. is ln offer rent.		Dep. var. is ln offer rent/m2
	(1)		(2)
	Coef.	P > t	Coef.	P > t
ATET
REF	0.742	0.000***	0.716	0.000***
controls
cc	−0.045	0.000***	−0.049	0.000***
area	0.014	0.000***	−0.001	0.596
age	−0.002	0.002**	−0.002	0.001**
quality	0.143	0.004**	0.146	0.005**
floor	−0.015	0.034*	−0.029	0.001**
tech	0.053	0.081	0.043	0.158
bldtype	0.020	0.031*	0.012	0.048*
month
2021m2	−0.010	0.024*	−0.008	0.036*
2021m3	−0.021	0.022*	−0.014	0.037*
2021m4	−0.011	0.677	−0.010	0.563
2021m5	−0.011	0.511	−0.013	0.353
2021m6	0.010	0.353	0.008	0.208
2021m7	0.023	0.093	0.022	0.079
2021m8	0.039	0.043*	0.043	0.019*
2021m9	0.050	0.104	0.049	0.064
2021m10	0.078	0.022*	0.078	0.010*
2021m11	0.075	0.027*	0.074	0.022*
2021m12	0.093	0.015*	0.084	0.019*
2022m1	0.117	0.009**	0.113	0.014*
2022m2	0.110	0.013*	0.107	0.015*
2022m3	0.067	0.003**	0.071	0.006**
2022m4	0.142	0.003**	0.136	0.006**
2022m5	0.167	0.009**	0.160	0.009**
intercept	6.796	0.000***	3.743	0.000***

Note: ATET estimates adjusted for covariates, group effects, and time effects. Std. err. adjusted for 5 clusters in the city. Repeated cross-sectional. N = 28,221. *p < 0.05; **p < 0.01; ***p < 0.001.

The estimation results are intuitive and in line with a priori expectations regarding the impact of control variables on rents. Hedonic regression estimates reveal that all considered housing attributes have logical signs. Offer rents were positively affected by the quality of the apartment (quality) and negatively affected by the distance from the city centre (cc) and building age (age). Offer rents of apartments located on the ground floor or last floor of the building were significantly higher than in the case of apartments situated on middle floors; other stay equal (floor). Along with expectations, net apartment area (area) was positively related to total rent but negatively related to rent per square meter (bigger apartments being relatively cheaper). Offer rents in apartments located in prefabricated buildings (mostly constructed in the 1970s and 1980s) seem lower than in traditionally made buildings (tech). However, the coefficients were not statistically significant in that case.

The estimates from econometric models allow us to assess the average treatment effect on the treated (ATET), which is an effect of the population increase in a given city due to the inflow of refugees on apartment rental prices, while others stay equal. Overall, the results suggest the significant positive impact of refugee inflow (REF) on apartment rents. The 1pp increase in city population due to refugees’ arrival causes approximately 0.72–0.74% increase in apartment rents (depending on specification). There is compelling empirical evidence that the relative size of refugee inflow translates to rent increases.

4.2. Apartments’ prices

Similarly to our previous empirical exercise on housing rents, we use a DiD model to evaluate the impact of the mass inflow of refugees on apartment prices in five cities in Poland. The total sample size was 35,950 apartment sale offers. In the repeated cross-sectional hedonic regression model, we controlled for selected locational and structural characteristics of residential properties. Again, the hedonic approach allowed us to control for housing quality heterogeneity. The monthly hedonic price indices in the study period are presented in Figure 4.

Figure 4. Monthly hedonic price indices from January 2021 till May 2022.

Graphical assessment reveals mostly parallel trends in apartment prices in the pretreatment period. We also have not found anticipatory effects before February 2022. Generally, we conclude that price trends in all the cities investigated were similar, and the market reaction did not proceed with the Russian invasion.

We investigate the issue further, as previously relaxing the assumption about equality of treatment intensity. We use two specifications of the model. In the first specification, the dependent variable is a natural logarithm of apartment offer price, whereas in the second, the dependent is a natural logarithm of apartment offer price per sqm. The estimated results of the DID model with continuous treatment REF are presented in Table 3.

Table 3. Difference-in-differences hedonic regression for offer apartment prices in selected cities in Poland.

	Dep. var. is ln offer price.		Dep. var. is ln offer price/m2
	(1)		(2)
	Coef.	P > t	Coef.	P > t
ATET
REF	0.255	0.047*	0.234	0.086
controls
cc	−0.041	0.000***	−0.045	0.000***
area	0.014	0.000***	−0.001	0.195
age	−0.002	0.035*	−0.001	0.080
quality	0.058	0.002**	0.060	0.002**
floor	0.006	0.154	0.001	0.822
tech	0.080	0.003**	0.085	0.003**
bldtype	0.012	0.118	0.015	0.049*
month
2021m2	0.008	0.100	0.005	0.214
2021m3	0.014	0.095	0.013	0.054
2021m4	0.031	0.000***	0.03	0.000***
2021m5	0.034	0.000***	0.037	0.000***
2021m6	0.048	0.000***	0.049	0.000***
2021m7	0.061	0.000***	0.061	0.000***
2021m8	0.071	0.000***	0.075	0.000***
2021m9	0.088	0.000***	0.089	0.000***
2021m10	0.096	0.000***	0.094	0.000***
2021m11	0.111	0.000***	0.110	0.000***
2021m12	0.120	0.000***	0.116	0.000***
2022m1	0.137	0.000***	0.134	0.000***
2022m2	0.140	0.000***	0.136	0.000***
2022m3	0.118	0.000***	0.118	0.001***
2022m4	0.123	0.002**	0.122	0.002**
2022m5	0.136	0.001**	0.136	0.002**
intercept	12.431	0.000***	9.324	0.000***

Note: ATET estimate adjusted for covariates, group effects, and time effects. Std. err. adjusted for 5 clusters in the city. Repeated cross-sectional. N = 35,950. *p < 0.05; **p < 0.01; ***p < 0.001.

Hedonic regression estimates reveal that most key attributes we controlled for in the empirical model significantly impact offer prices. The estimation results generally go along with theoretical expectations. The increase in the quality of apartments positively influenced their prices. Offer prices decreased with the distance from the city center (cc) and building age (age). Apartment net area (area) increased the overall asking price but negatively related to the offer price per square meter (bigger apartments being relatively cheaper). Offer prices were significantly lower in apartments located in prefabricated buildings compared to their otherwise similar counterparts found in traditionally constructed buildings (tech). On the other hand, the location on the floor did not significantly affect the offer prices (floor).

The results suggest a positive linkage between refugee inflow (REF) and house prices. The 1 pp increase in city population due to refugees’ arrival translates to a 0.23–0.26% increase in apartment prices (depending on specification). The coefficient is smaller than for housing rents and statistically significant only in one specification examined (p-value slightly below 0.05). The evidence on the relative impact of refuges inflow on house prices is relatively weak, especially compared to their effect on housing rents. We conclude that the results are ambiguous and must be treated with caution.

4.3. Discussion

The overall empirical results concerning the regional housing market reaction to the refugee crisis in Poland require further discussion. The research adds to the previous study on the immediate impact of war on house prices in Poland (Trojanek & Gluszak, 2022). Using an extended study period and data on refugees and the housing market in more cities, we found that 1 pp of a host city’s population caused by the mass inflow of Ukrainian refugees increased the average rents by 0.72–0.74%. The significant increase in housing rents after the Russian invasion of Ukraine was mainly caused by the increase in demand generated by a massive wave of refugees from Ukraine (surveys suggest that approximately 19% of refugees entered the rental market). On the other hand, the rent increase was also caused by a supply decrease. Surveys suggest that approximately 18% of Ukrainian refugees found free accommodation in houses and flats belonging to Polish citizens. As a result, a significant share of housing units that could have been potentially offered on the market has been temporarily removed from the stock. In comparison, the 1 pp increase in the city’s population related to the inflow of refugees resulted in a 0.23–0.26% increase in house prices in the short run. The estimated results are generally in line with prior research on immigration shocks; however, they are relatively weaker than those reported by Saiz (2007) – especially for house prices, where elasticities were closer to 1. This finding requires further discussion.

The fact that the effect of mass immigration on prices is significantly smaller than rent can be explained by tenure choice theory. Economic literature suggests a link between mobility and housing tenure choices (Boehm, 1981; Broulíková et al., 2020). Generally, there is a positive relation between the length of expected stay and the probability of buying a house, mainly due to relatively high transaction costs related to acquiring property rights (Haurin & Gill, 2002). People tend to rent when they do not plan to settle, and the results from the survey distributed among Ukrainian refugees suggest that the vast majority of refugees do not plan to stay permanently in Poland. They were only transiting through Poland, so they would rather enter the rental market than buy apartments. Secondly, the literature suggests (Ha et al., 2021) that the probability of investing in housing is lower for long-distance movers due to high search costs and the possibility of making ill-informed decisions. That explanation also seems to apply to many Ukrainian refugees who found temporary accommodation in Polish cities, which, in many cases, they have never visited before. Regardless of their economic status and plans to stay in Poland for an extended period, it is doubtful that refugees will buy housing immediately upon their arrival to Poland (in the first three months). Both those mechanisms explain why housing rents increased significantly more than prices in a short period after the Russian invasion of Ukraine.

There are several limitations to the research. Firstly, the estimates of market responses to refugees’ inflow are based on offered rents and offered prices. Using housing offers has one advantage over sales data related to information collection. Sales data is typically lagged and only available several months after the transaction has been recorded. Offer prices allowed us to observe the evolution of the housing market in real time and get up-to-date results. Additionally, rental contracts are not registered; thus, there is no statistical information on market rents. Therefore, offer rents are the only feasible data source that can be used in empirical research.

We only operated at the aggregate city level. Our estimates refer to overall house rents and prices in response to the demand shock caused by the massive inflow of Ukrainian refugees. Unfortunately, there is no detailed public information on the location of Ukrainian refugees (within city districts or neighbourhoods) that would allow for tracking intra-city rent and price dynamics at the submarket level (as in Alhawarin et al., 2021; Balkan et al., 2018). Provided the precise geo-referenced data on refugees’ mobility patterns are available, a more detailed spatially and nuanced investigation of housing submarkets responses will be possible.

Several papers addressed the linkages between immigration and the housing market in the long run (Saiz, 2003). We investigated the immediate, short-run impact of mass refugee shock on housing markets in major cities in Poland. Therefore, we followed rent and price adjustments three months after the Russian invasion and did not investigate further market reactions. In the mid-term perspective, housing markets will likely respond with supply adjustments, subject to the course of military operations in Ukraine and migration plans of refugees currently staying in major cities in Poland. According to the survey conducted at the end of April 2022, most of the Ukrainian refugees planned to return to Ukrainian immediately (39%) or several months after the cessation of hostilities in Ukraine (20%), but 16% wanted to stay in Poland for at least one additional year. Recent data and reports indicate that since August 2022, Poland has witnessed an exodus of Ukrainians who either migrated to other European countries (mainly Germany) or returned to their homes in Ukraine (Andrews et al., 2023; Centre for East European Studies, 2023; Kohlenberger et al., 2022). The links between the dynamics of movements of Ukrainian refugees and their mid-term effects on housing markets in Poland and other European countries are beyond the scope of this paper but remain an interesting direction for further research.

The research raises several housing policy questions. Despite housing market disturbances caused by the mass inflow of Ukrainian refugees to Poland, there was a consensus amongst major political parties in Poland on economic, legal and housing support for Ukrainian citizens. At the beginning of the refugee crisis, on 12 March 2022, the Polish parliament almost unanimously adopted a law on assistance to Ukrainian refugees (Lesinska, 2023). The political support followed a general positive attitude of Polish citizens towards Ukrainians that was particularly high at the beginning of the War in Ukraine. In the survey conducted by Ipsos in May 2022, approximately 67% of Poles expressed support for the prolonged stay of Ukrainian refugees in Poland (Ptak, 2023). On the other hand, negative attitudes towards refugees have also been observed. Reports suggest that xenophobia, anti-Ukrainian resentment and hate speech were primarily spread by far-right politicians, radical activists and publicists and facilitated by social media (Helsinki Foundation for Human Rights, 2023). The negative sentiments may increase, however, because demand shock caused by waves of Ukrainian adds to the previous housing affordability crisis caused by the COVID-19 pandemic, especially in case of vulnerable households (Kajta et al., 2022). Although a recent German study demonstrated that immigrants do not pay rent premiums compared to domestic citizens (Eilers et al., 2021), scattered evidence from Poland suggests that refugees from Ukraine may face substandard living conditions and are subject to segregation. Often, finding adequate housing became problematic for Ukrainians who compete with Polish citizens (primarily young households) on the rental market. Moreover, several other recent studies from various developed countries suggest that mass immigration can encourage xenophobic attitudes and translate into multiple discriminatory practices in the rental market (Auspurg et al., 2023; Verhaeghe & De Coninck, 2022; Ziersch et al., 2023). As a result, effective public policy and non-standard solutions related to temporary and permanent housing are required to increase housing affordability in major cities in Poland (Mędrzecka-Stefańska et al., 2022) and mitigate potential disturbances caused by mass immigration in future.

5. Conclusions

The Russian invasion of Ukraine in February 2022 resulted in a global political, economic and humanitarian crisis. On a regional scale, the ongoing military operations and the permanent risk of missile attacks and aerial warfare in Ukraine caused the largest refugee crisis in Europe since World War II. This study is a premiere direct attempt to assess the immediate impact of the displacement of refugees caused by the recent phase of the Russo-Ukrainian conflict on major metropolitan housing markets in Poland affected by the unprecedented inflow of people from Ukraine. Except for the early paper of Trojanek and Gluszak (2022), there is little (if any) empirical evidence of the impact of the War in Ukraine on housing rents and prices in Europe. We bridge the gap in the empirical evidence by using a novel quasi-experimental approach.

Using novel data on house rents and prices, we found a slight price increase and a substantial increase in short-term rents after the invasion. Using the difference-in-difference approach, we observed that the relative size of the inflow of Ukrainian refugees translates to house rents. We estimate that a 1 pp. increase in city population due to refugees’ arrival translates to a 0.72–0.74% increase in apartment rents. The estimates suggest that the inflow of refugees may have caused apartment price increases, however, they are inconclusive. The relative effect is smaller and not statistically significant at the conventional level.

To our best knowledge, despite the gravity of the war in Ukraine, there are no housing studies on the relationship between the refugee inflow initiated in February 2022 and housing market dynamics. Additionally, the paper contributes to the growing body of empirical evidence on the impact of mass migration or refugee inflow on housing markets targeted by newcomers. The findings have essential housing policy implications both in the short and long run, as a sharp decrease in housing affordability most likely requires well-targeted and effective public intervention. The latter goes beyond temporal protection and emergency instruments applied in the short term in most European countries.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Correction Statement

This article has been republished with minor changes. These changes do not impact the academic content of the article.

Additional information

Funding

The National Science Centre of Poland supported this work under Grant number 2021/43/B/HS4/01213. The open access was co-financed by the Minister of Science of the Republic of Poland under the “Regional Initiative for Excellence” Programme within the project “Poznań University of Economics and Business for Economy 5.0: Regional Initiative—Global Effects (RIGE)” and by the subsidy granted to the Krakow University of Economics (WAP Programme).

Notes on contributors

Michal Gluszak

Michal Gluszak is an Associate Professor and Vice - Director of Institute of Economics at Krakow University of Economics. He works at the Department of Real Estate and Investment Process. He specializes in Microeconomics, Public Economics, Housing Economics, Urban Economics and Real Estate Economics.

Radoslaw Trojanek

Radoslaw Trojanek is an Associate professor at the Poznań University of Economics and Business in the Department of Microeconomics and a property valuer, authorisation no. 5048. For over twenty years, he has compiled data on listings and housing market transactions in Poland. His research interest is in the broadly defined real estate market, particularly the housing market.

Appendix

Table A1. The housing dataset before and after cleaning.

	Sale offers		Rental offers
	Raw	After cleaning	Raw	After cleaning
Krakow
Mean price/m2	11,449.29	11,236.76	45.96	47.09
Std price/m2	3,258.64	2,837.15	13.70	11.96
Observations	48,638	8,766	28,278	9,087
Lodz
Mean price/m2	6,559.95	6,581.75	40.56	41.99
Std price/m2	1,571.39	1,303.51	14.68	8.89
Observations	15,036	1,854	6,446	1,447
Poznan
Mean price/m2	8,505.75	8,370.93	41.40	43.55
Std price/m2	2,056.16	1,835.97	12.36	10.52
Observations	12,430	1,958	7,123	1,701
Warsaw
Mean price/m2	13,098.69	12,920.33	60.58	63.08
Std price/m2	3,987.85	3,276.83	21.08	20.83
Observations	114,766	19,960	57,562	13,368
Wroclaw
Mean price/m2	9,566.15	9,288.46	49.85	52.18
Std price/m2	2,393.32	1,978.75	16.37	13.75
Observations	30,588	3,412	12,851	2,618

Source: Own research.

Table A2. Control variables used in hedonic rent models (N = 28,221).

Symbol	Description	Mean	SD	Min	Max
cc	Distance to the city center in km	3.86	2.65	0.06	19.41
area	Area of dwelling in square meters	53.61	24.88	15.06	198
age	Age of the building in years	20.47	28.62	0	278
quality	Quality of the apartment based on information in listings. Values from 1 to 3. Notation: 1– low quality, 2 – medium quality, 3 – high quality	2.79	0.41	1	3
floor	Location on floor (0-ground floor or last floor, 1-otherwise)	0.76	0.42	0	1
tech	Building construction technology (0-prefabracate, 1-traditional)	0.95	0.22	0	1
bldtype	Height of the building (0-up to 4 stories, 1-five stories and higher)	0.61	0.49	0	1
month	Month fixed effects 17-time dummy variables used in the global model. If the apartment was offered in a given month), it takes the value 1; otherwise, it takes 0.

Source: Own research.

Table A3. Control variables used in hedonic price models (N = 35,950).

Symbol	Description	Mean	SD	Min	Max
cc	Distance to the city center in km	5.18	2.94	0.06	15.85
area	Area of dwelling in square meters	57.54	22.87	15.45	199.89
age	Age of the building in years	38.88	30.19	0	522
quality	Quality of the apartment based on information in listings. Values from 0 to 3. Notation: 0 – apartment to renovation, 1 – low quality, 2 – medium quality, 3 – high quality	1.91	1.14	0	3
floor	Location on floor (0-ground floor or last floor, 1-otherwise)	0.47	0.49	0	1
tech	Building construction technology (0-prefabracate, 1-traditional)	0.78	0.41	0	1
bldtype	Height of the building (0-up to 4 stories, 1-five stories and higher)	0.46	0.49	0	1
month	Month fixed effects 17-time dummy variables used in the global model. If the apartment was offered in a given month), it takes the value 1; otherwise, it takes 0.

Source: Own research.

Table A4. Values of continuous treatment variable REF post-intervention.

Symbol	March 2022	April 2022	May 2022
Krakow	0.181	0.205	0.185
Lodz	0.101	0.117	0.073
Poznan	0.105	0.144	0.132
Warsaw	0.086	0.117	0.131
Wroclaw	0.209	0.297	0.301

Source: Own research.