The World Bank Economic Review, 37(1), 2023, 147–176 https://doi.org10.1093/wber/lhac024 Downloaded from https://academic.oup.com/wber/article/37/1/147/6874774 by International Monetary Fund / World Bank - IMF user on 11 September 2023 Article Refugees and Housing: Evidence from the Mortgage Market Yusuf Emre Akgündüz ˘ , Yavuz Selim Hacıhasanoglu , and Fatih Yılmaz Abstract This paper investigates the impact of large-scale Syrian refugee inflows on the Turkish housing market. Em- ploying a micro-level data set of the population of mortgaged houses in Türkiye between 2010 and 2017, it identifies the dynamic effects using a difference-in-differences approach. As the regional distribution of Syrian refugees is presumably not exogenous, it is instrumented in the estimations. The instrument is constructed using the distance from Turkish provinces to each Syrian region, while weighting each Syrian region by their pop- ulation and distance to Türkiye compared to other destination countries. The results show that house prices increased in response to the arrival of Syrian refugees. The effects are mostly driven by low-priced housing and faded after 2014. The results further show that construction permits and sales increased, while the average age of purchased houses declined, indicating an increase in supply that may explain the fading-out effect over time. Finally, the findings provide suggestive evidence that houses that are sold after the arrival of refugees decline in size, which further points to a squeeze in the housing market for natives. JEL classification: J15, J61, R31 Keywords: immigration, housing, house prices 1. Introduction The unfortunate increase in worldwide forced displacement over the years has led to a great deal of academic and policy interest in the impacts of refugees on host communities. Apart from the economic effects on local labor markets and firms, a key area of research is the effect on scarce local amenities and infrastructure. The impact of immigration on amenities can in fact be a larger contributing factor to public opposition to immigration (Card, Dustmann, and Preston 2012; Edo et al. 2019). Foremost among scarce amenities is housing and a large population inflow will inevitably affect the housing market in an area. Yusuf Emre Akgündüz (corresponding author) is an Assistant Professor at Sabanci University, Faculty of Arts and Social Sciences, Istanbul, Türkiye his email address is emre.akgunduz@sabanciuniv.edu. Yavuz Selim Hacıhasanoglu ˘ is Deputy Di- rector at the Research and Monetary Policy Department at the Central Bank of the Republic of Türkiye, Istanbul, Türkiye his email address is yavuz.selim@tcmb.gov.tr. Fatih Yılmaz is an Independent Research Economist, Istanbul, Türkiye and a Research Fellow at the Economic Research Forum, Cairo, Egypt; his email address is drfatihy@gmail.com. The contents of this paper do not reflect the views of the Central Bank of the Republic of Türkiye. All remaining errors are our own. © The Author(s) 2022. Published by Oxford University Press on behalf of the International Bank for Reconstruction and Development / THE WORLD BANK. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com 148 ˘ Akgündüz, Hacıhasanoglu, and Yılmaz This paper exploits the rapid arrival of Syrian refugees in Türkiye after the beginning of the Syrian Civil War to study the consequences of forced displacement on host-country housing markets. Rapid refugee Downloaded from https://academic.oup.com/wber/article/37/1/147/6874774 by International Monetary Fund / World Bank - IMF user on 11 September 2023 inflows can cause excess demand in the short run, but also contribute to increasing supply in the long run through a reduction in labor costs. Others channels through which immigration can affect house prices, namely native preferences for living in the same neighborhood as refugees and changes in the income of level of natives, make the relationship between immigration and housing an empirical question. Using a novel administrative data set of mortgage credits by banks, we analyze the relationship between local housing markets and refugee intensity. Using data up to five years after the arrival of Syrian refugees, we identify the effects of refugee inflows on the housing market in both the short and the long term. Our analysis covers the effects on house prices, mortgage loans, sales, construction permits, and house characteristics (i.e. age and size). The data set is comprised of the population of mortgaged house sales that were reported by Turkish banks to the Central Bank of the Republic of Türkiye (CBRT) between 2010 and 2017. Banks report all mortgaged house sales in Türkiye with information on house price, loan, house characteristics like size, and exact location (i.e. the parcel). The data set allows us to control for a rich set of house characteristics when analyzing the impact on price, loan-to-value ratio, and sales, as well as house characteristics. We use a difference-in-differences specification and instrument the distribution of refugees to analyze the effects. In the case of the housing market, the self-selection of refugees into low-cost (high-cost) regions can bias estimates downwards (upwards), and an instrumental variables approach is therefore needed to claim causality. Syrian refugees are heavily concentrated in certain provinces of Türkiye. Our instrumental variables strategy follows the existing literature on the impact of the Syrian refugees in Türkiye (Del Car- pio and Wagner 2015; Tumen 2019; Akgündüz and Torun 2020; Aksu, Erzan, and Kırdar 2022). In par- ticular, the relationship between Syrian refugee intensity in a province and the distance of that province from each Syrian region weighted by each region’s pre-war population shares and distance from Türkiye compared to alternative destination countries allows us to make causal claims on the impact of hosting refugees on the housing market. Our estimates suggest that refugees increased prices in host provinces. Per percentage point increase in the refugee-to-native ratio, house prices increase by more than 2 percent. The effect is almost immediately detectable after Syrian refugees first arrived in Türkiye. However, this effect appears to be temporary as the estimates become smaller and less statistically significant after 2014. Analyzing the heterogeneity of the effects, we find that the increase in prices is mainly led by an increase in the price of houses in low- priced neighborhoods, where the size of the effect on average is as high as 5.5 percent, depending on the specification. The impact on houses in neighborhoods with pre-crisis prices above the province median is weaker. We further show that older houses and houses with multiple rooms had a price premium after the arrival of the refugees, which is consistent with survey reports of house preferences among Syrian refugees. While the ratio of Syrian refugees to natives in a province has a positive effect on house prices, we find a negative effect in districts hosting refugee camps. We document an insignificant effect on loan-to-value ratios, indicating that an increase in mortgage availability due to higher income or loan supply did not cause the increase in house prices. Consistent with the increase in prices, we find an increase in demand as measured by the number of houses sold. The increase is led by a rise in the number of houses sold through mortgages as the share of mortgaged sales in total sales increases. Since Syrian refugees are likely to have limited access to the mortgage market, the rise in sales is mostly due to an increase in the rental income of houses. Consistent with the fading of the effect in later years, we find evidence of a higher number of new construction permits and show that the average age of houses sold on the mortgage market declined. However, these new houses are not necessarily of better quality, as we find some suggestive evidence of a decline in size. We make several contributions to the literature. The first strand of literature that we contribute links immigration to house prices. Here, our contribution is largely due to the suddenness of the Syrian shock The World Bank Economic Review 149 in Türkiye and the level of detail inherent in our data. Several studies have analyzed the impact of immi- gration on house and rental prices (see for example Saiz (2007) on America, Akbari and Aydede (2012) Downloaded from https://academic.oup.com/wber/article/37/1/147/6874774 by International Monetary Fund / World Bank - IMF user on 11 September 2023 on Canada, Gonzalez and Ortega (2013) on Spain, Accetturo et al. (2014) on Italy, Sá (2015) on the UK, and Degen and Fischer (2017) on Switzerland). The results are dependent on the level of analysis. At broader levels of aggregation, i.e. across cities or provinces, immigrants are generally found to increase prices (Gonzalez and Ortega 2013), but there can be negative effects at micro-levels of aggregation when native preferences may lead to moving away from immigrant hosting neighborhoods (Sá 2015).1 Most analyzes use enclave-based (i.e. shift-share) instruments, which Jaeger, Ruist, and Stuhler (2018) caution for potentially biased estimates. Data in all studies are at the level of provinces or districts rather than individual houses or locations, which further makes it necessary to assume that the composition of house characteristics among houses sold remains unaffected.2 Our study of the Syrian refugee crisis in Türkiye improves on the existing identification in the literature, first by exploiting a rapid and exogenous inflow of refugees, and secondly by employing micro-level data that allow us to control for house location at a very fine level and include controls for house characteristics. In addition, the nature of the Syrian refugee shock allows for further insights. Since the number of refugees rose rapidly between 2012 and 2015 and stabilized afterwards, we are able to identify the dynamic effects of an immigration shock on the housing market and distinguish between short- and long-term price responses. A smaller substrand of the literature on immigration and housing prices analyzes the impact of refugee shocks on the housing market. Specifically, three papers estimate the effect of refugee inflows on rental prices: Depetris-Chauvin and Santos (2018) analyze the impact of internally displaced persons on urban rental prices in Colombia and find an increase in low-price rentals and a decrease in high-price rentals, which they link to excess demand for low-price rentals and increased crime in hosting municipalities. Rozo and Sviatschi (2021) analyze the impact of Syrian refugees on the housing market in Jordan and find an increase in housing expenditures and the rental income of individuals who own houses. Most related to our paper is the study of Balkan et al. (2018), who analyze the impact of Syrian refugees on rental prices reported by natives in the Survey of Income and Living Conditions (SILC). The authors report an increase in prices of high-quality rentals and no impact on low-quality rentals. Their analysis is based on the early years of the refugee crisis up to 2013 and assumes that refugee-hosting regions were exogenously affected in 2012 and 2013 in a difference-in-differences setup. In studying house prices and characteristics on a longer horizon, our study complements it. However, we have significant data advantages when identifying the effects. An important distinction between SILC data and our mortgage data is that the SILC data only include information about the respondents’ location at the level of 12 NUTS-1 regions, while the mortgage data allow us to use the more detailed information about Syrian refugee distribution at the level of 81 provinces. Analyzing the effects at the 12-region level makes it difficult to control for regional year-specific shocks, which are important since Turkish regions have very different levels of development.3 In addition, the authors can only identify effects at the binary level, because the number of treated regions is limited at the 12-region level, while our study allows for estimates from incremental increases in the ratio of Syrian refugees to natives. A unique feature of our analysis is the estimation of effects on house characteristics, age and size, and loan-to-value ratios. Since previous studies mainly rely on aggregated regional house price indices, the 1 For a detailed overview of the results in this literature and the difference between within and across region results, see Cochrane and Poot (2021). 2 For example, the main specification in Gonzalez and Ortega (2013) is at the province level which corresponds to 50 Spanish provinces, Sá (2015) at the local authority level which corresponds to 170 local authorities in England and Wales, Accetturo et al. (2014) at the district level for 20 large Italian cities, and Depetris-Chauvin and Santos (2018) at the city level for 13 Colombian cities. 3 Previously, Aksu, Erzan, and Kırdar (2022) and Akgündüz and Torun (2020) have shown that controlling for regional trends can significantly alter results when estimating the effects of Syrian refugees on employment and firm outcomes. 150 ˘ Akgündüz, Hacıhasanoglu, and Yılmaz impact on the characteristics of new housing and mortgages remains understudied. House quality repre- sents an important dimension for overall welfare and loan-to-value ratios can link effects of immigration Downloaded from https://academic.oup.com/wber/article/37/1/147/6874774 by International Monetary Fund / World Bank - IMF user on 11 September 2023 on local economic conditions to the housing market. The remainder of the paper is organized as follows. The next section provides a general overview of the housing market and the arrival of Syrian refugees in Türkiye. A discussion on the supply, demand, and preference channels through which the refugee inflows can affect the housing market is then provided. This is followed by a presentation of the methodology and the description of the data. We then provide a throughly discussion of our results on prices, sales, construction permits, and house characteristics, along with various robustness checks. Finally, we provide conclusions in the last section. 2. Syrian Refugees and the Housing Market in Türkiye Syrian refugees first arrived in Türkiye at the end of April 2011. The legal status of Syrian refugees in Türkiye is designated as temporary guests rather than refugees. The initial numbers were small but rose rapidly to 880,000 by the end of 2013. The number reached around 3.4 million by the end of 2017. The early years of the crisis, 2012–2015, differ somewhat in that the rise in the number of Syrian refugees is higher. In December 2012, 53 percent of refugees were located in camps (UNHCR 2012). By 2013, the ratio had fallen to less than 25 percent (UNHCR 2013). Also in 2012–2013, most refugees were located in provinces close to the Syrian border and began to disperse across Türkiye in 2014 (Tumen 2016). Many field studies from face-to-face interviews with Syrian refugees and local citizens have been carried out since the beginning of the refugee crisis, several of which are summarized in table 1. A common finding in surveys is that Syrian refugees live in low-income neighborhoods in Türkiye (see the last column of table 1). Cost is likely to be the driving factor in the preference of refugees to settle in low-income areas. Given the rapid increase in the demand for housing, particularly low-cost housing, surveys report that house rental and sale prices significantly increased in Turkish provinces bordering Syria.4 Moreover, accommodation costs are an important share of Syrian households’ budgets, which they appear to limit by living with large families (i.e. extended families or multiple families) in big houses (with more bedrooms). Living with multiple families in large houses lowers the fixed costs of accommodation. According to the field studies, it is quite common to find large Syrian refugee households, consisting of 10–15 people, living in the same house. Given the high cost of relatively newer buildings (with central heating, better insulation, etc.), many Syrian refugees also prefer to live in old houses without a centralized heating system. This preference to minimize costs by sharing the fixed costs of accommodation raises the demand for especially large housing units with typically three or more bedrooms and/or relatively older units. Foreign nationals have been able to buy property in Türkiye without any special permit since 2012.5 However, special legislation applies to Syrian nationals. Following the unification of Hatay province with Türkiye in 1939, property purchases by Syrians in Türkiye (and all foreign nationals’ property purchases in Hatay Province only) have been regulated. Following the revision of the law in 2012, Syrians became eligible for property purchases in Türkiye with a special permit from the relevant state agency. In practice, only wealthy Syrian refugees may utilize this allowance. The alternative, which is also common practice among Syrian refugees, is to purchase a property through their relatives who are legal citizens of Türkiye, or through the ownership of their firms. Foreign nationals, including Syrians, can start a business (and thus establish a firm) in Türkiye, which can legally own housing assets, just like any other assets.6 4 These findings are also widely reflected in the media outlets, where several newspaper articles noted the increase in house rental and sale prices in the Turkish provinces bordering Syria. 5 See Land Registry Law with order number 644 article 35 legislated in May 2012. 6 The number of firms established by Syrians rose starting from 2013 (See Figure Figure 3 in TEPAV, 2018). Over the period of 2013–2017, about 6,300 new firms were established by Syrians, while the corresponding number was 278 for The World Bank Economic Review 151 Table 1. Regional Surveys on Syrian Refugee Population in Türkiye and Their Brief Findings Title Authors Year Region Method Findings Downloaded from https://academic.oup.com/wber/article/37/1/147/6874774 by International Monetary Fund / World Bank - IMF user on 11 September 2023 Syrians in Istanbul and A. Kavas, I. 2019 Istanbul Face to face interviews Syrian refugees mostly prefer Post-War Syrian Ghettos Avsar, O. to live in low-cost (in Turkish) Kadkoy, and neighborhoods of Istanbul. E. C. Bilgic They tend to live in extended families or multiple families in the same house. Rental rate and house prices increased in Syrian-refugee-populated neighborhoods. Exploring the Locational S. Savrav and 2019 Ankara Face to face interviews Syrian refugees live mostly in Preferences of Syrian N. a. Sat low-cost neighborhoods of Migrants in Ankara (in Ankara, where house prices Turkish) and rental rates are relatively cheaper. A Study on Syrians under K. Alptekin, D. 2018 Konya Face to face interviews (with Syrian refugees live mostly in Temporary Protection in A. Ulutas, and NGOs) low-cost neighborhoods of Konya D. U. Gunduz Konya. Syrian Refugees in the P. Budak, M. 2017 Elazig Face to face interviews Syrian refugees raised rental Perception of Local S. Demir, M. rates in the province. People and Social Tan, and M. Impacts: Case of Elazig Sati (in Turkish) Spatial Distribution and M. E. Sonmez 2016 Gaziantep Local census, income and Syrian refugees live mostly in Futurity of Syrian house price and rental rate low-cost neighborhoods of Refugees in the City of data Gaziantep. Gaziantep (in Turkish) Impact of Syrian H. Ozturkler 2015 Border Face to face interviews Syrian refugees raised rental Refugees Economic on and T. Göksel provinces rates in all the provinces and Türkiye: A Synthetic house sales in most provinces. Modelling (in Turkish) Source: Authors’ collection from various survey sources. Note: Syrians, just like any other foreign nationals, are allowed to obtain loans from banks in Türkiye. General credit provision conditions that apply to Turkish citizens—e.g., proving satisfactory stable income and providing collateral—would make it difficult to utilize this option for the majority of Syrian refugees. For natives, monthly mortgage payments from a loan are capped at half the monthly income. The biggest obstacle to credit markets is that most Syrian refugees are largely employed informally and therefore cannot formally document their income.7 Nevertheless, some wealthier Syrians, business owners or legal workers, may in practice obtain mortgage loans. Limited access to credit markets makes it unlikely that the majority of Syrians can purchase houses through the mortgage market. As a result, rentals appear to be a more economically viable alternative. It is therefore worth noting that an increase in house demand may not be due to house purchases by Syrian the 2010–2012 period. Following Istanbul with the highest number, most of these firms are located in Syrian-refugee- populated provinces of Türkiye bordering Syria such as Kilis, Hatay, and Gaziantep. See Akgündüz, van den Berg, and Hassink (2018) for an analysis of the effects of Syrian refugee inflows on firm entry. 7 Syrians were first allowed to receive work permits in January 2016. Between the years 2016 and 2018, only 28 thousand work permits were granted, as opposed to the more than 4 million Syrians living in Türkiye. 152 ˘ Akgündüz, Hacıhasanoglu, and Yılmaz Figure 1. Hedonic Price Index 100 100 Downloaded from https://academic.oup.com/wber/article/37/1/147/6874774 by International Monetary Fund / World Bank - IMF user on 11 September 2023 80 80 60 60 40 40 2010 2011 2012 2013 2014 2015 2016 2017 House Price Index (Gaziantep, Adiyaman, Kilis) House Price Index (Türkiye) Source: Central Bank of the Republic of Türkiye Statistics Portal. Note: The figure plots the annual average of the monthly data provided for the NUTS-2 region (Gaziantep, Adiyaman, and Kilis) with the highest ratio of Syrians to native population (17.1 percent) relative to the country average as of 2016. refugees. It would rather be motivated by the significant increase in the demand for rentals. Given the rapid shift in housing demand for relatively low-cost districts, house prices would be expected to adjust to reflect the increase in the rental rate. Nevertheless, rises in rental prices will place upward pressure on house prices as their investment value goes up. The impact of Syrian refugees is visually evident in the evolution of house prices in the border region. Figure 1 plots the house price index reported by the CBRT for the NUTS-2 region consisting of Gaziantep, Kilis, and Adiyaman. As this region is on the Syrian border, it has been hosting refugees since 2012 and had the highest ratio of Syrians to the native population in 2016. When compared to the house price index for Türkiye, fig. 1 shows that prices started to rise in 2012 and did not return to the country level until 2017. 3. Literature and Theory There are four channels through which the arrival of refugees can affect the housing market. First, the increase in the population represents a simple increase in housing demand. Second, the entry of Syrian refugees into the labor market can reduce construction costs and therefore raise housing supply. Third, the arrival of Syrian refugees can affect local economic activity and therefore the availability of mortgages for natives and the demand for housing. Finally, natives’ preferences towards living in proximity to refugees and decisions to migrate can affect the housing market. Demand: The most straightforward impact of refugees on the housing market is an increase in demand. By 2017, the ratio of refugees to natives had reached 5 percent in Türkiye and this number is as high as 100 percent in provinces like Kilis (fig. 2). Even within a given hosting province, refugees are likely to be concentrated in lower-income areas given that they are employed mostly informally. The growth in population density in regions hosting refugees will have boosted housing demand either through direct Figure 2. Syrian Refugees to Population Ratio by Province (2016) Kırklareli Sinop Bartın Kastamonu Zonguldak Artvin Ardahan The World Bank Economic Review Edirne Karabük Samsun Tekirdağ İstanbul Rize Kocaeli Düzce Ordu Trabzon Yalova Sakarya Çankırı Amasya Giresun Bolu Çorum Tokat Gümüşhane Kars Bayburt Çanakkale Bursa Bilecik Erzurum Balıkesir Kırıkkale Iğdır Ankara Yozgat Eskişehir Sivas Erzincan Ağrı Kütahya Kırşehir Tunceli Bingöl Muş Nevşehir Manisa Elazığ Afyonkarahis Aksaray Kayseri İzmir Uşak Malatya Bitlis Van Diyarbakır Konya Batman Isparta Niğde Kahramanmara Adıyaman Siirt Denizli Aydın Burdur Osmaniye Şanlıurfa Mardin Şırnak Hakkari Adana Gaziantep Karaman (.3,1] Muğla Kilis (.1,.3] (.05,.1] Antalya Mersin Hatay (.01,.05] (.0005,.01] [0,.0005] Source: Authors’ calculations from Ministry of Interior and Turkish Statistics data. Note: The figure displays the map of refugee-to-population ratios by Turkish provinces as of 2016. 153 Downloaded from https://academic.oup.com/wber/article/37/1/147/6874774 by International Monetary Fund / World Bank - IMF user on 11 September 2023 154 ˘ Akgündüz, Hacıhasanoglu, and Yılmaz purchases by Syrians or the increase in rental prices, which can in turn increase the demand for housing investment.8 The demand effect should be particularly evident in cases of large-scale immigrant inflows, Downloaded from https://academic.oup.com/wber/article/37/1/147/6874774 by International Monetary Fund / World Bank - IMF user on 11 September 2023 which is what we observe in Türkiye with the rapid increase in the number of Syrian refugees. Similarly in Spain, Gonzalez and Ortega (2013) attribute a quarter of the growth in Spanish housing prices between 2000 and 2010 to the increase in the immigrant population. The arrival of refugees even in small numbers in the earlier years of the crisis can lead to a perception of a housing squeeze, which can increase demand in hosting regions. Supporting this claim in Sweden, Tyrcha (2020) reports a strong impact on the perception of housing availability from the arrival of refugees. The reaction of the housing market to the increase in demand will depend mainly on the curvature of the supply curve. If the supply of housing is relatively inelastic in hosting provinces, the increase in demand will lead to a rise in prices, while a more flat supply curve will lead to stable prices but increases in the number of new housing units. Since the construction of new housing units takes time and the arrival of refugees is unexpected, we would predict the short-run effects of the increase in population to be more pronounced on housing prices. The existing literature largely confirms the expectation that the supply curve is inelastic in the short run and elastic in the long run (Harter-Dreiman 2004). Even in the long run, the housing supply curve may never be completely flat due to factors like topography or government regulations and permits on construction (Saiz 2010; Diamond 2017). Supply: Since Syrian refugees are mostly employed informally, studies on their labor-market effects gen- erally find a decline in wages and employment of informally employed and low-skilled natives (Ceritoglu et al. 2017; Aksu, Erzan, and Kırdar 2022). These findings are in line with Akgündüz and Torun (2020), who find that the task content of the average native employee experienced a reduction in manual task content in favor of more complex tasks. Analyzing the Türkiye Demographic and Health Survey in 2018, Demirci and Kirdar (2023) show that Syrian refugees are relatively younger, less educated, and tend to work physically in primarily informal private labor markets. Given that construction has a relatively high informality share in Türkiye, a large influx of informal Syrian refugee workers into the construction sector may well reduce the costs in hosting regions. Such an effect would be in line with Bratsberg and Raaum (2012), who find a decline in wages in the construction sector due to immigrant employment. The decline in costs due to the employment of Syrian refugees can then shift the supply curve to the right, particularly in the long run. The supply channel would predict and explain the negative effects of immigration on house prices found in long-run studies like Akbari and Aydede (2012). Unlike the demand effect, which should be stronger in neighborhoods where refugees live, the supply effect is likely to benefit the entire province due to the larger size of regional labor markets compared to neighborhoods. Economic activity and domestic credit conditions: An indirect mechanism by which Syrian refugees can affect house prices is through their impact on local economies. For instance, Altındag, ˘ Bakis, and Rozo (2020) report an increase in the use of firm inputs and Akgündüz et al. (2022) find a rise in firm sales, entry, and exports after the arrival of Syrian refugees in hosting provinces. Besides the positive impact of the arrival of Syrians on the local economic activity, no overall adverse effects on average wages of natives were found (Aksu, Erzan, and Kırdar 2022). Put differently, Syrian refugees appear to substitute for natives in informal labor, while complementing the formal employment of natives. These findings suggest that the disposable income of natives in hosting provinces may have increased. A permanent increase in income would improve natives’ credit worthiness and thus raise their borrowing capacity. An increase in borrowing capacity would apply to mortgages as well, as a simple outcome of the widespread mortgage market practice in Türkiye that caps monthly mortgage loan payments to half of the borrowers’ monthly income. While the degree of the impact would depend on the capacity and willingness of financial 8 We assume that the increase in population immediately translates into an income effect. This assumption is justified by previous studies that find limited effects from Syrian refugees on formally employed natives. While only formally employed natives can realistically buy houses through the mortgage system, the negative effects on informal native employment are also limited (Aksu, Erzan, and Kırdar 2022). The World Bank Economic Review 155 institutions to increase mortgage loans in hosting regions, the increase in access to mortgage loans can trigger a rise in loan-to-value ratios and thus prices (Mian and Sufi 2009; Barone et al. 2021).9 Downloaded from https://academic.oup.com/wber/article/37/1/147/6874774 by International Monetary Fund / World Bank - IMF user on 11 September 2023 Preferences: Native preferences towards living in the same neighborhood as Syrian refugees can sig- nificantly affect the demand for these neighborhoods. If there is a strong internal migration response to avoid living in the same regions as Syrian refugees, the demand effect from the arrival of refugees can be tempered by a decline in demand from natives. Natives may be substitutable in the labor force with refugees, which would decrease their economic prospects and lead to out-migration from local labor mar- kets. Previous studies show limited evidence for a large internal migration effect from the arrival of Syrian refugees across Turkish provinces (Del Carpio and Wagner 2015; Aksu, Erzan, and Kırdar 2022). While there is evidence showing that native workers migrate out due to a decline in their employment prospects, the arrival of additional public employees to provide services such as education and health care appears to offset these effects in overall internal migration (Akgündüz, Altan, and Bagır˘ 2020). The net effect on the housing market from internal migration across provinces is therefore likely to be limited. Internal migration may also occur between neighborhoods within provinces hosting refugees. A com- monly cited cause is an increase in crime, although the impact of immigration on crime is generally es- timated to be weak (Bianchi, Buonanno, and Pinotti 2012; Ousey and Kubrin 2018). Even without a tangible cause, like economic prospects or crime, ethnic preferences may play a role. Tabellini (2020) finds that immigration can trigger hostile political reactions despite a lack of economic effects. If natives are unwilling to live in the same neighborhood as refugees, they may migrate to other neighborhoods within the province. As such, we may find an increase in demand in neighborhoods that do not host any refugees. The preference to live away from refugees can even push prices downwards, as shown in the case of the UK by Sá (2015), who attributes the decline in house prices in immigrant-hosting neigh- borhoods to the out-migration of high-income natives from these neighborhoods. In the long run, such preferences can even lead to the formation of new residential areas with limited ethnic mixing (Moraga, Ferrer-i Carbonell, and Saiz 2019). Overall, the theoretical discussion and the literature findings present a mixed picture of the expected effects of refugees on the housing market of hosting provinces. While increased demand is likely to push prices up, this effect may be offset in the longer run through a decline in construction costs. In addition, increased economic activity could change the availability of mortgage loans. The effects at the neigh- borhood level will also vary depending on whether the natives exhibit an aversion to living in the same neighborhood as refugees. 4. Methodology Our empirical strategy is difference-in-differences based on the province level variation of Syrian refugees. In the baseline model, we estimate equation (1) using OLS, where i refers to each individual house sold, j to province, and t to the year and month of sale. Since we expect the effects to vary over time, we define the refugee–native ratios separately for each year, Ratiok jt . The ratio variable for a given year is equal to the refugee–native ratios at the province level for that year and is set equal to 0 in 2010 and 2011.10 Our primary parameter of interest is β k , which shows the effect of hosting refugees each year. In addition, we estimate the effect of a refugee camp in a housing district that varies over time at the district level with the parameter γ . In all models, we include year fixed effects given by Tt and province fixed effects α j , as 9 See Follain and Dunsky (1997) for more direct evidence on a positive relationship between income and mortgage loans. 10 As discussed in the data section, we exclude the months after April in 2011 and the whole of 2012 in our baseline analysis due to a lack of data at the province level. 156 ˘ Akgündüz, Hacıhasanoglu, and Yılmaz well as month fixed effects given by mijt : 2017 Downloaded from https://academic.oup.com/wber/article/37/1/147/6874774 by International Monetary Fund / World Bank - IMF user on 11 September 2023 yi jt = a0 + βk Ratiok jt + γ Campi jt + Xi jt + α j + Tt k=2013 + mi jt + λi + RT jt + ei jt . (1) Since our data include detailed information at the house level, we control for the changes in the composi- tion of houses sold by including a vector of house characteristics X, which includes the log of size, the log of age, the number of rooms (by type) of the house, and whether there is security, an elevator, or a pool in the building.11 In some specifications, we further control for the heterogeneity in location by including neighborhood fixed effects λi . We can also further control for parcel-level fixed effects, which are close to being equivalent to building fixed effects, but this significantly reduces the variation in the sample since the identification is then based on sales of houses from the same parcel in different years. We interact the ratio variable with house characteristics to test heterogeneous effects in some models and add interactions between year fixed effects and the given house characteristic in these models. Since Syrian refugees are, particularly before 2014, concentrated in regions bordering Syria, the difference-in-differences approach relies on the existence of parallel trends between provinces with high Syrian to native ratios and others. Regional differences in development between eastern and western Türkiye make it unlikely for the parallel trends assumption to hold. The growing literature on the eco- nomic effects of Syrian refugees have tackled this problem by including regional year fixed effects or time trends (Akgündüz and Torun 2020; Aksu, Erzan, and Kırdar 2022).12 We follow a similar approach and include NUTS-1-level region-year fixed effects, which are shown in equation (1) as RTjt . The effects are then identified through time variation within the 12 NUTS-1 regions of Türkiye. However, the resulting specification is demanding and may lead to overfitting. As robustness tests, we replace the 12 NUTS-1 region-year fixed effects with 5 region-year fixed effects where the regions are defined as West, South, North, East, and Central Anatolia. Finally, we test the specifications using linear time trends at the level of 26 NUTS-2 regions. While region-year fixed effects are more flexible in capturing time variation, the NUTS-2-level linear time trends allow for finer regional controls. Throughout the analysis, we further choose what is considered to be the most conservative option to deal with serial correlation and cluster the standard errors at the level of 81 provinces (Cameron and Miller 2015). The Syrian refugee crisis in Türkiye provides a shock with plausibly exogenous timing. However, after 2013, Syrian refugees spread across Türkiye rather than staying in the border regions. Their province selection may well be endogenous if they choose to settle in provinces with cheaper and more accessible housing or provinces with a higher rate of economic growth, which would in turn lead to differences in the trend of house prices. To deal with the potential endogeneity problem caused by self-selection, we use the instrument based on the distance from each province to the Syrian border proposed by Aksu, Erzan, and Kırdar (2022), which modifies the distance-based instrument of Del Carpio and Wagner (2015).13 Since 11 The inclusion of house characteristics as controls may be interpreted as undesirable due to the “bad control” problem if these characteristics are affected by the arrival of refugees. However, our purpose is to estimate the effects on the hedonic price rather than the average price of houses. Nevertheless, we include a specification in the tables where house characteristics are omitted. 12 In their main specifications, Aksu, Erzan, and Kırdar (2022) use 5-region time trends while Akgündüz and Torun (2020) use 12 NUTS-1 region-year fixed effects to control for regional trends in employment and firm outcomes. 13 The instrument of Del Carpio and Wagner (2015) multiplies the total number of Syrian refugees in Türkiye with the weighted distance from each Turkish region to a Syrian region, further weighted by the pre-war share of the population in each Syrian region. To test the robustness of our results, we repeated the analysis using the distance-based instrument employed by Akgündüz and Torun (2020), who construct their instrument at the NUTS-3 level following Del Carpio and Wagner (2015) and find very similar results. The similarity of the second-stage results is unsurprising given that the correlation between the two instruments is 0.93. The World Bank Economic Review 157 Table 2. Cross-Sectional First-Stage Estimates at the Province Level 2013 2014 2015 2016 2017 Downloaded from https://academic.oup.com/wber/article/37/1/147/6874774 by International Monetary Fund / World Bank - IMF user on 11 September 2023 A: No regional controls Instrument 0.004** 0.003** 0.005*** 0.005*** 0.005*** (0.0015) (0.0013) (0.0018) (0.0020) (0.0018) F-stat 6.22 6.08 6.78 6.95 7.62 N 81 81 81 81 81 B: With 12 region fixed effects Instrument 0.004** 0.004** 0.005** 0.006** 0.005** (0.0018) (0.0016) (0.0022) (0.0024) (0.0021) F-stat 5.27 5.17 5.69 5.69 6.16 N 81 81 81 81 81 Source: Authors’ estimations. Note: Dependent variables are the ratio of Syrian refugees to the native population at the level of 81 provinces for each year. The instrument is the weighted distance from a province to Syrian regions for each year. Panel B includes NUTS-1-level region fixed effects. Standard errors are heteroskedasticity robust. ***p < 0.01, **p < 0.05, *p < 0.1. we estimate the impact of the ratio of Syrian refugees to native population for each year, we construct the instrument, IVTj , for each year as well, according to equation (2). There are 12 regions (governorates) in Syria and ds,T is the minimum distance from each to neighboring countries: Türkiye, Lebanon, Iraq, and Jordan. The variable π s gives the pre-war population share of each Syrian region, STt stands for the total number of Syrian refugees in these four countries in a given year t, while dp,s is the distance from each Turkish province to the Syrian border. In essence, the instrument is the distance-weighted number of refugees in each province for a given year. We later test the heterogeneity of effects of Syrian refugee inflows on house prices for houses located in high- and low-price neighborhoods, older houses, and large houses. When including interaction effects in our models to estimate the heterogeneity of Syrian-to-native ratio effects by house characteristics, we generate instruments for those interactions by interacting the house characteristics with the instrument in the following equation: 12 1 π ds,T s STt IVT j = . (2) s=1 ds,T 1 + 1 ds,L + d1s,J + 1 ds,I d p,s The Aksu, Erzan, and Kırdar (2022) instrument we employ differs from the formulation by Del Carpio and Wagner (2015) in two ways. First, it reweights the prewar population shares of Syrian regions according to their distance from the four refugee destinations. As such, the weight on regions close to Türkiye increases. For example, the weighted pre-war population of Aleppo exceeds 40 percent, as opposed to the actual population share of 21.6 percent. The reweighting according to distance from destination countries is justified by the 2018 Türkiye Demographic and Health Survey data, which show that over 58 percent of Syrian refugees in Türkiye originate from Aleppo. This change may therefore improve the relevance of the instrument. Second, it uses the total number of Syrian refugees rather than Syrian refugees in Türkiye. Since the size and timing of the inflow to Türkiye may itself be endogenous, using the total number of Syrian refugees appears to be the safer option to satisfy the exclusion restriction assumption. The baseline specification shown by equation (1) requires the instrument to be relevant for the distri- bution of refugees across Turkish provinces for each year between 2013 and 2017. Therefore, the instru- ment needs to be relevant across provinces within each year. To test the relevance of the instrument, in table 2 we present cross-sectional estimates of the first stage at the level of 81 provinces for all years. Panel A presents the bivariate relationships while panel B includes NUTS-1-level region fixed effects. In both panels, the instrument is strongly correlated with the refugee distribution and the coefficient esti- mate remains stable. The F-statistics are lower than the 10 threshold at the province level. However, for 158 ˘ Akgündüz, Hacıhasanoglu, and Yılmaz our baseline estimation of price effects with house characteristics, NUTS-1 region-year fixed effects and neighborhood fixed effects, the first stage F-statistics are 22.6, 29.12, 68.59, 58.37, and 51.52 in order Downloaded from https://academic.oup.com/wber/article/37/1/147/6874774 by International Monetary Fund / World Bank - IMF user on 11 September 2023 from 2013 to 2017. 5. Data Our primary data set is collected by the CBRT to monitor price changes in the Turkish housing market. During the approval stage of mortgage credits, banks use real estate appraisal companies’ valuation re- ports that are logged to and are used by the CBRT to generate and publicly report the Residential Property Price Index (RPPI) at the NUTS-2 level every month.14 We have access to the micro-data that are used to generate regional aggregates. Covering the years 2010–2017, the data set contains the prices of all houses sold with mortgage financing in Türkiye, which corresponds to nearly 40 percent of all house sales in 2013. There is little information about how similar mortgaged houses are to other houses in terms of their characteristics and price levels. The price index generated by the CBRT has a 0.98 correlation with the house price index of the Turkish Statistical Institute during the period of analysis, suggesting that price changes in the mortgage market are representative of the overall price changes in the house market. While the data set also includes valuation reports from mortgage applications for houses that were not sold, we exclude these observations from the data, as they may not reflect the actual transaction prices. Valuation reports include a rich set of observable house characteristics. In particular, we have the following information for each house: price, loan, building quality, construction year, gross area (in square meters), number of rooms, bathrooms, and balconies, and heating system types. In addition, we also have information about whether the building has security, car parking, a swimming pool, and an elevator. We construct loan-to-value ratios by dividing the mortgage loan obtained from the bank by the price of the house.15 The valuation reports further contain detailed geographical information about the house at the province, county, neighborhood, and parcel level. We use geographic information to include location fixed effects at the neighborhood and parcel levels.16 We merge this micro-level house price data with the number of Syrian refugees across 81 provinces in Türkiye from 2013 to 2017. As the Syrian refugees first arrived in Türkiye in late April 2011,17 we assume the number of Syrian refugees from 2010 until April 2011 to be 0 in the analysis. We drop the months after April in 2011 and 2012 from the analysis given that the initial numbers of the refugees were negligible, are unavailable at the province level, and a high proportion were housed in camps. The cross- provincial data on the number of Syrian refugees from 2013 to 2017, obtained from the Ministry of the Interior’s Immigration Management Directorate, are end-of-year values. Assuming end-of-year values to be representative of the whole year would be misleading, as the monthly figures from the UNHCR display significant variation in a given year. Considering this, we adjust the end-of-year values to represent cross- provincial annual averages using the monthly series from the UNHCR.18 In addition to the non-camp refugee population, we also track the location of camps across the years at the district level, as some 14 The RPPI can be accessed through the official statistics website of the CBRT: https://evds2.tcmb.gov.tr/index.php? 15 To avoid data mistakes and outliers, we exclude all loan-to-value ratios above 0.9. Legally, the maximum loan-to-value ratio during the sample period was 0.8. The number of observations where the loan-to-value ratio exceeds the 0.9 limit is less than 0.01 percent of the sample. 16 Parcel-level information is close to being equivalent to the building in which a flat or house is located in the Turkish real estate system. The title deeds of residences identify the location of the house based on the parcel information. Supporting the idea that parcel fixed effects can be treated as building fixed effects, we found that regressing house size in square meters on parcel fixed effects results in R-squared values above 0.70 in our sample. 17 For further details on the early chronological developments of the Syrian refugee crises in Türkiye, see Dinçer et al. (2013). 18 This adjustment follows Aksu, Erzan, and Kırdar (2022), where each province (end-of-the-year) value is multiplied by the ratio of the monthly average number of Syrians in Türkiye to the end-of-the-year number of Syrians in Türkiye. The World Bank Economic Review 159 refugee populations continued to live in these camps throughout the sample period. The camp location dummy is constructed based on UNHCR reports for Türkiye and updated for each year, except for 2015, Downloaded from https://academic.oup.com/wber/article/37/1/147/6874774 by International Monetary Fund / World Bank - IMF user on 11 September 2023 at the district level.19 The monthly average annual total number of Syrian refugees was 402,843 in 2013, 862,732 in 2014, 1,896,295 in 2015, 2,732,498 in 2016, and 3,112,170 in 2017. While the numbers rise rapidly, the share of refugees in each province is relatively stable, with high correlation at the province level over the years. We construct the ratio of the Syrian refugees to the native population obtained from the Turkish Statistical Institute. The province level population figures do not include the number of Syrian refugees. Refugees are legally designated as temporary guests rather than residents, while the population counted by the Turkish Statistical Institute is based on the residents’ registry. Merged with the Syrian-to-native population ratios, our data set is comprised of 2.18 million observa- tions. While the variation is at the level of 81 provinces, being able to include neighborhood and parcel fixed effects allows for fine-level controls of location and building heterogeneity. There are more than 25,000 neighborhoods and over 400,000 unique parcels in our data.20 Neighborhoods are entered into the data set as a text variable and there exist a number of small typos, which we corrected by grouping neighborhood entries based on their Levenshtein distance.21 When fixed effects for these two variables are included in a specification, only locations that are observed in at least two years can be included in the sample. While including neighborhood fixed effects results in a marginal drop in the sample size, the resulting sample size is 1.6 million when parcel fixed effects are included. The summary statistics for prices, total sales, and house characteristics are presented in tables 3, 4, and 5 respectively. The price variable is normalized for all houses by dividing it by the size of the house in square meters. To avoid outliers driving the results, we winsorize the (per square meter) price variable at 1 percent and 99 percent thresholds for each province-year group. Between 2010 and 2017, house prices in Türkiye increased from 1,147 to 2,111 Turkish lira per square meter, which corresponds to an 84 percent increase in nominal terms. There is significant regional variation behind average house prices (see fig. 3). Since there is considerable variation in inflation in Türkiye, we deflate house prices by CPI to eliminate the increase in house prices due to changes in the inflation rate. Table 3 shows the log of CPI-adjusted mean values of housing prices by year. We calculate the price per square meter of each house (in Turkish lira), and use its log transformation as our dependent variable. Table 4 summarizes the characteristics of all the houses we use in our main sample. It shows that the average size of the houses in our data is 113 square meters and the average building year is 2005. Means for security staff, elevator, parking area, and swimming pool show the percentage of houses having these facilities. These variables are included in most specifications as controls for building characteristics. In order to directly test the impact of refugee inflows on housing supply and demand, we use data from the Turkish Statistical Institute on house sales, mortgaged house sales, and new construction permits. These data are provided at the year-province level and the number of construction permits is further divided into residential and non-residential categories. A single permit can be given for a multi-dwelling building, implying that permits should be interpreted as the number of buildings that begin construction rather than the number of individual dwellings. We take the ratios of house sales and new construction 19 These reports are accessible on UNHCR’s website refugees in Türkiye: https://www.unhcr.org/turkey.html. They are not published at regular intervals. We used daily reports published on 9 November 2012, 18 December 2013, 12 September 2014, 5 December 2016, and 2 October 2017. 20 These figures are based on the number of fixed effects included in specifications with neighborhood and parcel fixed effects. Only location fixed effects with at least two observations from different years can be included in the estimations. 21 We used the Stata add-on of Barker and Pöge (2017) to implement the grouping. Not correcting the typos in the neigh- borhood name entries and treating each unique entry as a separate neighborhood makes negligible difference to the results. 160 Figure 3. Average House Prices Kırklareli Sinop Bartın Kastamonu Zonguldak Artvin Ardahan Edirne Karabük Samsun Tekirdağ İstanbul Rize Kocaeli Düzce Ordu Trabzon Yalova Sakarya Çankırı Amasya Giresun Bolu Çorum Tokat Gümüşhane Kars Bayburt Çanakkale Bursa Bilecik Erzurum Balıkesir Kırıkkale Iğdır Ankara Yozgat Eskişehir Sivas Erzincan Ağrı Kütahya Kırşehir Tunceli Bingöl Muş Nevşehir Manisa Elazığ Afyonkarahis Aksaray Kayseri İzmir Uşak Malatya Bitlis Van Diyarbakır Konya Batman Isparta Niğde Kahramanmara Adıyaman Siirt Denizli Aydın Burdur Osmaniye Şanlıurfa Mardin Şırnak Hakkari Adana Gaziantep Karaman (6.9,7.3] Muğla Kilis (6.7,6.9] (6.5,6.7] Antalya Mersin Hatay (6.45,6.5] (6.43,6.45] [6.35,6.43] Source: Authors’ calculations from Mortgage Registry data for the years 2010–2017. Note: The figure displays the map of the average of (log) meters squared prices by Turkish provinces as of 2016. ˘ Akgündüz, Hacıhasanoglu, and Yılmaz Downloaded from https://academic.oup.com/wber/article/37/1/147/6874774 by International Monetary Fund / World Bank - IMF user on 11 September 2023 The World Bank Economic Review 161 Table 3. Summary Statistics—Yearly Prices Year Mean p50 SD p10 p90 N Downloaded from https://academic.oup.com/wber/article/37/1/147/6874774 by International Monetary Fund / World Bank - IMF user on 11 September 2023 2010 6.800 6.768 0.447 6.310 7.357 269,742 2011 6.762 6.721 0.428 6.298 7.303 101,429 2013 6.862 6.791 0.421 6.389 7.429 356,313 2014 6.894 6.815 0.441 6.404 7.497 321,876 2015 6.929 6.840 0.470 6.414 7.577 345,965 2016 6.958 6.879 0.471 6.432 7.604 391,605 2017 6.941 6.875 0.464 6.407 7.574 411,612 Source: Authors’ calculations from the mortgage registry data. Note: The price variable is defined per meter squared. It is then log transformed and winsorized at the 1% and 99% thresholds for each province-year group. Data for 2011 are limited to the months January–April. Table 4. Summary Statistics—Building Characteristics Mean p50 SD p10 p90 N Year of construction 2005.967 2010 11.058 1989 2016 2,191,905 Age (log) 1.597 1.386 1.154 0 3.258 2,184,825 Size (m2) 113.604 108 85.917 67 160 2,198,542 Size (log) 4.663 4.682 0.363 4.204693 5.075 2,198,542 Building quality 2.362 2 0.573 2 3 2,198,512 Number of living rooms 1.010 1 0.142 1 1 2,198,542 Number of rooms 2.735 3 0.830 2 4 2,198,542 Number of bathrooms 1.252 1 0.480 1 2 2,198,533 Number of balconies 1.568 2 0.796 1 2 2,198,185 Security staff 0.096 0 0.294 0 0 2,198,542 Heating system 2.182 2 0.872 1 4 2,198,508 Elevator 0.441 0 0.496 0 1 2,198,530 Parking area 0.442 0 0.497 0 1 2,198,542 Swimming pool 0.061 0 0.240 0 0 2,198,542 Source: Authors’ calculations from the mortgage registry data by CBRT. Note: For the age variable, which is log transformed, we add 1 to avoid dropping new houses. These variables constitute the vector of building controls in the regression results. permits to the population at the province level for easier interpretation.22 The relevant summary statistics are provided by table 5.23 Not all houses are sold through mortgages in Türkiye, which has implications for the interpretation of our results. Around a third of total houses sold in Türkiye were sold through mortgages during the years of our analysis, according to the Turkish Statistical Institute data. There is considerable province- level heterogeneity in the ratio of mortgaged house sales to total house sales. For interpretation, the key is whether mortgaged sales are a randomly selected sample of all sales. There is no available data to compare house characteristics of mortgaged and other sales. Nevertheless, we can speculate that if there is a selection, banks prefer to loan to higher-value houses since they are worth more as collateral. As such, if the impact of immigration on high- (low-) value houses is greater, our estimates will be larger (smaller) than the average effect on all houses. 22 The estimated effects are qualitatively similar when we use the log of new construction permits. 23 When analyzing the effects on these variables, we cannot exclude the months after April in 2011. We, therefore, assume that the number of refugees in 2011 is 0. This may lead to an underestimation of the effects if the initial refugee inflow in 2011 affected the housing supply or sales. 162 ˘ Akgündüz, Hacıhasanoglu, and Yılmaz Table 5. Summary Statistics—Total Sales at Month-Province Level Mean p50 SD p10 p90 N Downloaded from https://academic.oup.com/wber/article/37/1/147/6874774 by International Monetary Fund / World Bank - IMF user on 11 September 2023 Sales / population 0.0117 0.0107 0.0075 0.0028 0.0209 560 Mortgages / population 0.0037 0.0034 0.0025 0.0007 0.0069 560 Mortgages / sales 0.3107 0.3158 0.0853 0.2116 0.4037 560 Residence permits / population 0.0114 0.0108 0.0061 0.0041 0.0192 560 Non-residential permits / population 0.0001 0.0000 0.0001 0 0.0001 560 Source: Authors’ calculations from the Turkish Statistical Institute data. Note: The summary statistics are for the period 2010–2017 with the exception of 2012. Residence and non-residential permits refer to the number of new construction permits issued. Sales refers to all house sales in a province, while mortgages refers to the number of mortgaged house sales in the province. 6. Results 6.1. Impact on House Prices The results showing the effects of the arrival of Syrian refugees on house prices in a province are presented in table 6. Panel A shows OLS estimates while panel B shows 2SLS estimates. In the first column, we only include year, month, and province fixed effects. In column 2, a vector of house characteristics is added. In column 3, regional trends are controlled for by using 12 region-year fixed effects. Column 4 includes neighborhood fixed effects while column 5 replaces the neighborhood fixed effects with parcel-level (or building-level) fixed effects. In all columns and both panels, the estimated coefficients are positive, suggesting that house prices in- creased in response to the arrival of Syrian refugees. Also common to all estimates is the decline in the size of the coefficients over time. By 2016, some point estimates even turn negative. The average coefficient es- timates across the columns are shown by fig. 4. There is limited difference in coefficients between OLS and IV estimates, which points to a limited role for endogeneity and self-selection after controlling for region- year fixed effects. This effect is larger when we control for the location of the house in the province, indicating a degree of heterogeneity in the effects across houses. As expected, standard errors tend to be smaller in the OLS estimates. According to the 2SLS estimates in column 4, which is our preferred baseline specification with region-year and neighborhood fixed effects, we find that a 1 percentage point increase in the Syrian-to-native population ratio is associated with a 2.2 percent and a 1.2 percent in- crease in house prices in 2013 and 2014 respectively. However, neither OLS nor IV estimates are precisely estimated or consistently statistically significant at conventional levels. The largest change in the estimated coefficients is observed when region-year fixed effects are intro- duced in column 3. This shift in estimates suggests that there are significant regional trends. To test the robustness of the estimates to different region-year controls, we present estimates in table 7, where we replace 12 NUTS-1-level region-year fixed effects with 5 region-year fixed effects in columns 1 to 3 and with NUTS-2-level linear time trends in columns 4 to 6. The results do not qualitatively change. However, the effects are more precisely estimated. It is therefore reasonable to conclude that specifications with the NUTS-1 region-year fixed effects provide relatively conservative estimates of the effects. The decline in the size of the estimated coefficients over the years may be due to a fading out of the initial effect or the change in the dispersion of Syrian refugees across Turkish provinces in later years. To test whether there is an actual fade-out of the effects, we fix the ratio of Syrian refugees to natives to 2013 and estimate yearly coefficients for all years except 2011, which is set as the baseline year. We use our baseline specification corresponding to column 4 in table 6. Similar to the baseline estimates, we find that the positive effects fade out over time after 2015 as shown in table 8. We further see a large point estimate already in 2012, indicating that the effect of the refugee inflow is indeed immediate. On the other hand, The World Bank Economic Review 163 Table 6. Impact of Syrian Refugees on House Prices (1) (2) (3) (4) (5) Downloaded from https://academic.oup.com/wber/article/37/1/147/6874774 by International Monetary Fund / World Bank - IMF user on 11 September 2023 A: OLS Ratio-2013 4.1681** 3.9752** 1.0382 1.3157 1.9635** (1.7353) (1.7368) (1.0465) (1.0349) (0.9395) Ratio-2014 2.4880* 2.3432* 0.4793 0.7390 1.2750 (1.3223) (1.3000) (0.8258) (0.8778) (0.8631) Ratio-2015 1.0229 0.9879 0.0492 0.1603 0.4752 (0.6270) (0.6343) (0.3723) (0.4153) (0.4306) Ratio-2016 0.5081 0.4890 −0.0936 0.0104 0.2638 (0.4557) (0.4678) (0.3409) (0.3713) (0.3603) Ratio-2017 0.2620 0.2762 −0.0970 −0.0067 0.2694 (0.4202) (0.4372) (0.3639) (0.4010) (0.3758) Camp district −0.0322 −0.0215 −0.0123 −0.1130* −0.1395** (0.0670) (0.0572) (0.0551) (0.0657) (0.0687) B: 2SLS Ratio-2013 3.6329 3.3026 1.9770 2.2203 3.2928** (2.3373) (2.4638) (1.7715) (1.8279) (1.6367) Ratio-2014 2.1730 1.8860 1.2380 1.2139 2.1009 (2.0940) (2.0841) (1.5905) (1.5763) (1.5185) Ratio-2015 0.3842 0.3506 0.3468 0.3299 0.8146 (1.0740) (1.0563) (0.8025) (0.8088) (0.8681) Ratio-2016 −0.2742 −0.2937 −0.0730 −0.0445 0.3397 (0.7842) (0.7853) (0.6567) (0.6881) (0.7230) Ratio-2017 −0.6497 −0.6052 −0.2243 −0.2133 0.2250 (0.6825) (0.6807) (0.6388) (0.6827) (0.7293) Camp district −0.0112 −0.0010 −0.0107 −0.0970 −0.1359 (0.0735) (0.0641) (0.0586) (0.0945) (0.0959) N 2,198,542 2,184,341 2,184,341 2,162,167 1,570,824 Month + + + + + Year + + + + + Province + + + + + Building controls — + + + + 12 region-year FE — — + + + Neighborhood FE — — — + — Building FE — — — — + Source: Authors’ estimations. Note: Ratio is the ratio of Syrian refugees to the native population at the level of 81 provinces. The camp district variable is a dummy indicating whether the district the house is located in contains a refugee camp. In all IV estimates, the weighted distance from a province to Syrian regions for each year are used as instruments for the ratio variable. Controls included in each column are shown at the bottom of the table. Building controls include age, whether the house is new, size, number of rooms (by type), a measure of building quality provided by the expert evaluation, the type of the heating system, and whether the building has an elevator, pool, or safety staff. Standard errors are clustered at province level. ***p < 0.01, **p < 0.05, *p < 0.1. the estimated coefficients for 2010 are close to 0, which lends support to the validity of the parallel trends assumption in the differences-in-differences setup.24 Apart from the effect of Syrian refugees at the province level, we estimate the effect of refugee camps at the district level within provinces. Estimates in columns 4 and 5 in table 6 are more reliable in esti- mating the effect of camp presence at the district level since the previous three columns do not include district-level fixed effects. The point estimates become large and negative once neighborhood fixed effects 24 When we included the months after April 2011 in the analysis, the 2010 coefficient became negative, indicating that there is already an effect on house prices from refugee inflows in the latter half of 2011. 164 Figure 4. Average Estimated Price Effects by Year (a) (b) (c) Source: Authors’ estimations. Note: Panel (a) shows the average coefficients estimated from five specifications in table 6 for each year. Panels (b) and (c) display the average coefficients from five specifications for low-price (b) and high-price (c) houses in table 9. ˘ Akgündüz, Hacıhasanoglu, and Yılmaz Downloaded from https://academic.oup.com/wber/article/37/1/147/6874774 by International Monetary Fund / World Bank - IMF user on 11 September 2023 The World Bank Economic Review 165 Table 7. Impact of Syrian Refugees on House Prices—Alternative Controls (1) (2) (3) (4) (5) (6) Downloaded from https://academic.oup.com/wber/article/37/1/147/6874774 by International Monetary Fund / World Bank - IMF user on 11 September 2023 A: OLS Ratio-2013 2.2613* 2.9076* 3.2321** 3.0312* 3.6202** 3.9619** (1.3466) (1.4682) (1.2321) (1.5351) (1.5396) (1.5358) Ratio-2014 1.4779 2.0722* 2.4382** 1.7763 2.2035* 2.5153** (1.0665) (1.2355) (1.1130) (1.0939) (1.1677) (1.2277) Ratio-2015 0.6775 0.9489 1.1858* 0.6801 0.8774* 1.0387* (0.5791) (0.6861) (0.6381) (0.4309) (0.4965) (0.5367) Ratio-2016 0.3540 0.5885 0.7722 0.2293 0.3916 0.4988 (0.4631) (0.5549) (0.4993) (0.2552) (0.3064) (0.3284) Ratio-2017 0.2207 0.4407 0.6517 −0.0131 0.1632 0.2836 (0.4350) (0.5229) (0.4557) (0.1983) (0.2470) (0.2693) Camp district −0.0181 −0.1278* −0.1459* 0.0036 −0.0396 −0.0645 (0.0569) (0.0758) (0.0783) (0.0587) (0.0682) (0.0721) B: 2SLS Ratio-2013 2.7544 3.0857 3.9287** 4.2203** 5.7111*** 6.3211*** (2.0119) (2.1522) (1.7660) (1.7320) (1.5635) (1.2931) Ratio-2014 1.8451 1.9539 2.7383 2.7389* 3.7118*** 4.2171*** (1.7732) (1.8842) (1.6520) (1.4004) (1.3401) (1.1388) Ratio-2015 0.6845 0.7371 1.1932 0.8352 1.4164** 1.6281*** (0.9307) (0.9995) (0.9602) (0.5708) (0.5931) (0.5381) Ratio-2016 0.1920 0.2521 0.6170 0.1249 0.6432* 0.7256** (0.7463) (0.8275) (0.7828) (0.3491) (0.3557) (0.3125) Ratio-2017 −0.0132 0.0349 0.4555 −0.1801 0.4281 0.5054* (0.6967) (0.7920) (0.7581) (0.3169) (0.2942) (0.2574) Camp district −0.0155 −0.0946 −0.1337 0.0001 −0.0584 −0.0938 (0.0608) (0.1076) (0.1067) (0.0597) (0.0991) (0.1062) N 2,184,341 2,162,167 1,570,824 2,184,341 2,162,167 1,570,824 Month + + + + + + Year + + + + + + Province + + + + + + Building controls + + + + + + 5 region-year FE + + + — — — 26 region-year trend — — — + + + Neighborhood FE — + — — + — Building FE — — + — — + Source: Authors’ estimations. Note: Ratio is the ratio of Syrian refugees to the native population at the level of 81 provinces. The camp district variable is a dummy indicating whether the district the house is located in contains a refugee camp. In all IV estimates, the weighted distance from a province to Syrian regions for each year are used as instruments for the ratio variable. In column 4, the F-statistics for the first stage of each ratio variable in order from 2013 to 2017 are 22.64, 29.12, 68.59, 58.37, and 51.52. Controls included in each column are shown at the bottom of the table. Building controls include age, whether the house is new, size, number of rooms (by type), a measure of building quality provided by the expert evaluation, the type of the heating system, and whether the building has an elevator, pool, or safety staff. Standard errors are clustered at province level. ***p < 0.01, **p < 0.05, *p < 0.1. are introduced, indicating a decline of around 10 percent in house prices in districts hosting refugee camps. Similar to the effect of refugee inflows, the point estimates are not precisely estimated and are marginally insignificant at conventional levels. These estimates suggest that there is a preference to live away from refugee camps within the province. 6.2. Heterogeneity of Price Effects We know from surveys of Syrian refugees that they are concentrated in low-price areas. If refugees are generally staying in low-price housing, there will be heterogeneity in the demand effect, with weaker effects 166 ˘ Akgündüz, Hacıhasanoglu, and Yılmaz Table 8. Impact of Syrian Refugees on House Prices—Event Study Approach (1) (2) Downloaded from https://academic.oup.com/wber/article/37/1/147/6874774 by International Monetary Fund / World Bank - IMF user on 11 September 2023 OLS 2SLS Ratio-2010 0.0884 −0.3233 (0.3372) (0.3381) Ratio-2012 1.5068* 1.7898 (0.8598) (1.1521) Ratio-2013 1.7785 1.9835 (1.3916) (1.9827) Ratio-2014 1.9254 1.8025 (1.8707) (2.6578) Ratio-2015 1.3539 0.9504 (1.8780) (2.8857) Ratio-2016 0.6260 −0.4410 (2.0763) (3.4299) Ratio-2017 0.1729 −1.2639 (2.1518) (3.6394) Camp district −0.1185* −0.1024 (0.0609) (0.0885) N 2,397,742 2,397,742 Source: Authors’ estimations. Note: Ratio is the ratio of Syrian refugees to the native population at the level of 81 provinces. The camp district variable is a dummy indicating whether the district the house is located in contains a refugee camp. In columns 2 and 4, 2011 observations after April are excluded from the sample. In all IV estimates, the weighted distance from a province to Syrian regions for each year are used as instruments for the ratio variable. All regressions include year, month, neighborhood fixed effects, and controls for building characteristics—corresponding to column 4 baseline, table 4. Standard errors are clustered at province level. ***p < 0.01, **p < 0.05, *p < 0.1. on high-cost housing and stronger effects on low-price housing. We test this hypothesis empirically in table 9. Rather than estimating a single parameter for each year, we interact the ratio variable with an indicator variable (high price) that is equal to 1 for parcels above the median price for a province in 2011.25 Table 9 presents the results for the five specifications shown in table 6 in two panels for OLS and IV estimations. The results in all specifications show that the positive effects on prices are stronger for houses located in a parcel that had a price level below the median in 2011. According to the IV estimates, the 2013–2014 impact on low-cost housing prices ranges between 2.5 percent and 7 percent per percentage point increase in the refugee-to-native ratio. Similar to the main results, the effects decline over time. Figure 4 shows the average of the five specifications over the years. The IV results start at 5.5 percent in 2013 and decline to 0.5 percent by 2017. Interactions between refugee ratios and above-median-price variables show that high-price houses are affected less by the arrival of Syrian refugees. Figure 4 shows the mean of the sum of main and interaction effects for each year. According to the OLS estimates, there is initially a positive effect for high-price houses, but the effect is less than half the impact on low-price houses. The size of the effect starts at around 2 percent in 2013 and declines to 0 over the years. For the IV model, the effects are even smaller and turn negative in later years. Surveys of Syrian refugees explicitly report that refugees tend to live in older houses due to price and location. We capture a similar price effect when we allow for heterogeneity based on the age of the house. In table 10, we estimate the interaction between older houses aged more than 4 and the ratio of Syrian refugees to the native population in a given year. While the interaction effects have a smaller size than 25 We limit the high–low-price neighborhood definition to the start of the refugee inflow, which limits the sample size for this exercise. Specifically, high-price neighborhoods are defined as having a price above the median neighborhood-level average price in 2011. The World Bank Economic Review 167 Table 9. Impact of Syrian Refugees on House Prices—Price Heterogeneity (1) (2) (3) (4) (5) Downloaded from https://academic.oup.com/wber/article/37/1/147/6874774 by International Monetary Fund / World Bank - IMF user on 11 September 2023 A: OLS Ratio-2013 5.8940*** 5.3929*** 2.8561** 2.5453** 2.5347** (1.5942) (1.6653) (1.1601) (1.1051) (1.0163) Ratio-2014 3.8559** 3.7481** 1.3975 1.5737 1.7482* (1.6017) (1.5936) (1.1765) (1.1267) (1.0400) Ratio-2015 1.7719** 1.7659** 0.5173 0.4753 0.4927 (0.7679) (0.7961) (0.5413) (0.5579) (0.5502) Ratio-2016 1.1718** 1.1912** 0.2893 0.2540 0.2734 (0.5223) (0.5571) (0.4952) (0.4766) (0.4671) Ratio-2017 0.7763 0.8113 0.3407 0.3425 0.3984 (0.4810) (0.5103) (0.5193) (0.4994) (0.4792) High x Ratio-2013 −3.2745*** −2.4939*** −2.3339*** −1.0547** −0.4611 (0.8388) (0.7534) (0.7690) (0.4223) (0.4609) High x Ratio-2014 −1.9764 −1.9555* −1.4811* −0.7182* −0.2051 (1.2442) (1.0526) (0.8440) (0.4012) (0.3788) High x Ratio-2015 −1.1326** −1.0956*** −0.8308*** −0.3776 −0.0608 (0.4802) (0.4028) (0.2945) (0.2618) (0.1896) High x Ratio-2016 −0.8604** −0.8794** −0.7429*** −0.3122* 0.0419 (0.3892) (0.3532) (0.2302) (0.1645) (0.1342) High x Ratio-2017 −0.5201 −0.5262* −0.6325*** −0.3109** −0.0036 (0.3169) (0.3080) (0.2371) (0.1303) (0.1530) Camp district −0.1726* −0.1413 −0.1272* −0.1674** −0.1793** (0.0875) (0.0851) (0.0702) (0.0662) (0.0757) B: 2SLS Ratio-2013 7.6348*** 7.0760*** 5.0155* 4.1778** 3.9611** (2.0924) (2.2840) (2.5482) (1.9805) (1.5845) Ratio-2014 5.4224*** 5.2482*** 3.7808* 2.9272 2.6898 (1.8877) (1.8936) (2.1480) (1.8075) (1.6314) Ratio-2015 1.8299 1.8360 1.5283 1.1551 1.0240 (1.1206) (1.1252) (1.0451) (0.9398) (0.9313) Ratio-2016 0.6899 0.7234 0.8468 0.5150 0.3939 (0.9106) (0.9190) (0.8490) (0.7753) (0.7898) Ratio-2017 0.0297 0.1108 0.6413 0.4073 0.3115 (0.8024) (0.7868) (0.8956) (0.8564) (0.8523) High x Ratio-2013 −7.7715** −6.1630** −5.8346** −2.9874** −1.1973 (3.4225) (2.9514) (2.8201) (1.3363) (0.7793) High x Ratio-2014 −6.6272*** −5.8197*** −5.0189** −2.4191*** −0.3984 (2.3504) (2.0502) (2.0978) (0.8705) (0.4811) High x Ratio-2015 −2.8593*** −2.4837*** −2.1488*** −1.0922*** −0.2389 (0.8620) (0.7129) (0.8068) (0.3073) (0.2710) High x Ratio-2016 −2.0191*** −1.7941*** −1.6878** −0.7481*** −0.0702 (0.6395) (0.5331) (0.6403) (0.2584) (0.2679) High x Ratio-2017 −1.7048** −1.4678*** −1.5581** −0.7010*** −0.0674 (0.6552) (0.5376) (0.6063) (0.2561) (0.3326) Camp district −0.1087 −0.0889 −0.1215 −0.1666** −0.1867** (0.0950) (0.0976) (0.0801) (0.0786) (0.0886) N 440,732 436,855 436,855 431,513 359,278 Building controls — + + + + 12 region-year FE — — + + + Neighborhood FE — — — + — Building FE — — — — + Source: Authors’ estimations. Note: Ratio is the ratio of Syrian refugees to the native population at the level of 81 provinces. The camp district variable is a dummy indicating whether the district the house is located in contains a refugee camp. In all IV estimates, the weighted distance from a province to Syrian regions and their interactions with high-price houses for each year are used as instruments for the ratio of Syrian to the native population ratio and the interaction between the ratio and high-price housing. High-priced houses are defined as houses located in parcels that had a price above the province median in 2010–2011. Controls included in each column are shown at the bottom of the table. Building controls include age, whether the house is new, size, number of rooms (by type), a measure of building quality provided by the expert evaluation, the type of the heating system, and whether the building has an elevator, pool, or safety staff. The interactions between high-priced housing and year are added as a control in all models. Month, year, and province fixed effects are controlled in all the models. Standard errors are clustered at the province level. ***p < 0.01, **p < 0.05, *p < 0.1. 168 ˘ Akgündüz, Hacıhasanoglu, and Yılmaz Table 10. Impact of Syrian Refugees on House Prices—Age Heterogeneity (1) (2) (3) (4) (5) Downloaded from https://academic.oup.com/wber/article/37/1/147/6874774 by International Monetary Fund / World Bank - IMF user on 11 September 2023 A: OLS Ratio-2013 4.0249** 3.7685** 0.9122 1.0994 1.7809* (1.8702) (1.8542) (1.1216) (1.0842) (0.9427) Ratio-2014 2.0684 2.0019 0.1802 0.5078 1.1123 (1.2822) (1.2610) (0.7981) (0.8736) (0.8147) Ratio-2015 0.8774 0.8551 −0.0383 0.0638 0.4104 (0.6216) (0.6223) (0.3830) (0.4169) (0.4158) Ratio-2016 0.3805 0.3728 −0.1714 −0.0683 0.2171 (0.4533) (0.4613) (0.3574) (0.3784) (0.3542) Ratio-2017 0.1423 0.1624 −0.1858 −0.0895 0.2342 (0.4160) (0.4346) (0.3761) (0.4064) (0.3770) Old x Ratio-2013 0.1774 0.4380 0.3825 0.5309 0.2840* (0.8588) (0.7731) (0.7783) (0.3707) (0.1600) Old x Ratio-2014 1.0390 0.9830* 1.0387 0.6390** 0.2882 (0.7572) (0.5826) (0.6487) (0.2423) (0.1876) Old x Ratio-2015 0.2982 0.3890* 0.3023 0.2563** 0.0731 (0.2819) (0.2324) (0.2257) (0.1098) (0.1012) Old x Ratio-2016 0.3025 0.3461* 0.2721 0.2068** 0.0593 (0.2175) (0.1818) (0.1817) (0.0960) (0.0643) Old x Ratio-2017 0.2786* 0.2953** 0.2551* 0.1814** 0.0155 (0.1650) (0.1471) (0.1474) (0.0826) (0.0647) Camp district −0.0335 −0.0213 −0.0118 −0.1135* −0.1391** (0.0660) (0.0566) (0.0546) (0.0660) (0.0691) B: 2SLS Ratio-2013 4.2445* 3.4096 2.3280 1.9289 2.8312* (2.5195) (2.6662) (2.0922) (1.9529) (1.6437) Ratio-2014 2.2711 1.7790 1.2504 0.9702 1.7432 (2.0227) (2.0385) (1.6864) (1.6001) (1.4833) Ratio-2015 0.3900 0.2756 0.2916 0.1984 0.6733 (0.9682) (0.9948) (0.8297) (0.8078) (0.8530) Ratio-2016 −0.2667 −0.3430 −0.1198 −0.1334 0.2902 (0.6861) (0.7330) (0.6834) (0.6970) (0.7222) Ratio-2017 −0.6386 −0.6593 −0.2765 −0.2803 0.2199 (0.5901) (0.6394) (0.6528) (0.6901) (0.7406) Old x Ratio-2013 −1.9419 −0.6406 −1.0364 0.6371 0.6812 (2.0768) (1.3434) (1.4466) (0.4513) (0.5076) Old x Ratio-2014 −0.7648 −0.0189 −0.1451 0.5705** 0.6288*** (1.4215) (0.8801) (0.8941) (0.2347) (0.1997) Old x Ratio-2015 −0.2241 0.0972 0.1083 0.3218*** 0.2357*** (0.7090) (0.4634) (0.3954) (0.1206) (0.0794) Old x Ratio-2016 −0.1823 0.0272 0.1047 0.2156* 0.0402 (0.5720) (0.3835) (0.3182) (0.1161) (0.0693) Old x Ratio-2017 −0.1022 0.0600 0.1265 0.1329 −0.0944 (0.3998) (0.2430) (0.1964) (0.0836) (0.0592) Camp district −0.0139 −0.0011 −0.0109 −0.0973 −0.1353 (0.0730) (0.0636) (0.0582) (0.0943) (0.0960) N 2,198,542 2,184,341 2,184,341 2,162,167 1,570,824 Building controls — + + + + 12 region-year FE — — + + + Neighborhood FE — — — + — Building FE — — — — + Source: Authors’ estimations. Note: Ratio is the ratio of Syrian refugees to the native population at the level of 81 provinces. The camp district variable is a dummy indicating whether the district the house is located in contains a refugee camp. In all IV estimates, the weighted distance from a province to Syrian regions and their interactions with old houses for each year are used as instruments for the ratio of Syrian to the native population and the interaction between the ratio and old houses. Old houses are defined as houses older than four years. Controls included in each column are shown at the bottom of the table. Building controls include age, whether the house is new, size, number of rooms (by type), a measure of building quality provided by the expert evaluation, the type of the heating system, and whether the building has an elevator, pool, or safety staff. The interactions between large houses and year are added as a control in all models. Month, year, and province fixed effects are controlled in all the models. Standard errors are clustered at province level. ***p < 0.01, **p < 0.05, *p < 0.1. The World Bank Economic Review 169 the interaction with the price of the house, we still find a positive additional effect in the order of 0.2 percent to 0.6 percent for old houses once neighborhood fixed effects are introduced. In line with the Downloaded from https://academic.oup.com/wber/article/37/1/147/6874774 by International Monetary Fund / World Bank - IMF user on 11 September 2023 heterogeneity with respect to price, the interaction effects with age disappear in later years. A third source of heterogeneity we empirically test is related to the survey findings that Syrians share houses with multiple rooms. We may expect a higher demand for houses with multiple rooms to in- crease the demand for large houses. Since we directly observe the number of rooms in a house, we simply generated a variable that indicated whether a house has more than three rooms (the median value) and interacted it in our specifications with the ratio of Syrian to the native population. The results of these estimations are shown in table 11. While the interactions are not quite as large as the interactions with high-price housing, we find an additional positive effect for houses with more than three rooms. The ad- ditional effect for houses with multiple rooms is larger and more consistently statistically significant at conventional levels in the IV panel, indicating an effect of about 0.5 percent. Unlike the main effect, the interaction coefficients remain large and positive until 2017. Since mortgages are largely attainable for natives, a change in the composition of house characteristics can be informative about native choices in the housing market. If there is a preference for older and larger houses among Syrian refugees that drives the prices up, natives may start buying younger and smaller houses. Specifically, we show the effects at the house level for two outcomes in table 12, where log- transformed age in panel A and size (in log-transformed square meters) in panel B are dependent variables. Since house characteristics are now the dependent variables, we do not present any specifications with the vector of house characteristics as controls. The first panel shows the effects on average house age, where the estimates are uniformly negative. The effects are statistically significant and large, at around 3 percent in 2013 and 2014 when the price effects were large. A decline in average house age may be indicative of two mechanisms. First, if new natives are buying newly constructed houses due to an increase in supply, building age can fall. Second, natives may be moving to newer neighborhoods in response to Syrians settling in urban and older neighborhoods. Both explanations are consistent with the increase in prices observed in 2013 and 2014 and the additional premium for older houses. We find more limited evidence for an effect on size. Panel B shows the estimates for the size effects. While they are largely negative, few estimated parameters are statistically significant. A negative effect would be in line with the price premium for the larger houses we explored. The decline in the size of housing purchased with mortgages. The results on prices and house characteristics are in line with a heterogeneous demand shock in the housing market. According to survey evidence, Syrians tend to live in older, low-price houses and often rent rooms rather than complete houses. Consistent with the increased demand, we find that the overall price effect is almost entirely driven by low-price houses and there is an additional premium on houses with multiple rooms. Consistent with a preference from Syrian refugees for older and larger houses, natives appear to buy younger and smaller houses. While not always statistically significant, we also find a negative effect on house prices in districts hosting refugee camps, which would be consistent with a preference to live away from refugee camps. On the other hand, we find smaller effects on house prices in higher-priced locations, indicating that there is no role for the mobility of natives towards high-price housing because of a preference to avoid living in the same neighborhoods as Syrian refugees. Since the demand shock appears to be heterogeneous, an implication of the results is that the variation of house prices in refugee-hosting provinces would rise after their arrival. This implication is confirmed in table 13, where we regress the standard deviation of province level prices26 on the ratio of Syrian refugees at the province-year level. In line with the effects on prices, we find a positive effect on the standard deviation of prices at the province level that fades away over time. 26 Specifically, this is the standard deviation of our CPI-adjusted, log-transformed, and winsorized price variable. 170 ˘ Akgündüz, Hacıhasanoglu, and Yılmaz Table 11. Impact of Syrian Refugees on House Prices—Size Heterogeneity (1) (2) (3) (4) (5) Downloaded from https://academic.oup.com/wber/article/37/1/147/6874774 by International Monetary Fund / World Bank - IMF user on 11 September 2023 A: OLS Ratio-2013 4.1258** 3.9659** 1.0322 1.2427 1.9848** (1.6962) (1.6786) (1.0016) (1.0105) (0.9337) Ratio-2014 2.5356* 2.3534* 0.4959 0.7035 1.2551 (1.3172) (1.2738) (0.8084) (0.8582) (0.8645) Ratio-2015 1.0361 0.9911 0.0540 0.1346 0.4669 (0.6277) (0.6281) (0.3669) (0.4079) (0.4305) Ratio-2016 0.5142 0.4888 −0.0928 −0.0115 0.2568 (0.4628) (0.4646) (0.3380) (0.3688) (0.3604) Ratio-2017 0.2597 0.2627 −0.1023 −0.0295 0.2661 (0.4251) (0.4321) (0.3601) (0.4005) (0.3767) Large x Ratio-2013 0.6318 0.1767 0.3292 0.5905 −0.0623 (0.8872) (0.6674) (0.6717) (0.3835) (0.2002) Large x Ratio-2014 −0.0316 0.0155 0.0696 0.3288 0.2437 (0.5611) (0.4175) (0.4316) (0.2900) (0.1502) Large x Ratio-2015 0.1307 0.0605 0.0789 0.2582* 0.1288* (0.2622) (0.1839) (0.1979) (0.1341) (0.0657) Large x Ratio-2016 0.1499 0.0776 0.0509 0.2016** 0.0851 (0.2091) (0.1522) (0.1582) (0.0892) (0.0648) Large x Ratio-2017 0.1944 0.1449 0.0737 0.1766** 0.0483 (0.2156) (0.1807) (0.1640) (0.0850) (0.0704) Camp district −0.0297 −0.0204 −0.0116 −0.1107* −0.1408** (0.0664) (0.0561) (0.0540) (0.0657) (0.0690) B: 2SLS Ratio-2013 3.4493 3.3592 1.9659 2.1065 3.2569** (2.3474) (2.3819) (1.7097) (1.7823) (1.6177) Ratio-2014 2.0798 1.9052 1.2167 1.1056 2.0345 (2.1002) (2.0339) (1.5472) (1.5271) (1.5032) Ratio-2015 0.3419 0.3481 0.3329 0.2727 0.7866 (1.0871) (1.0403) (0.7863) (0.7865) (0.8582) Ratio-2016 −0.3243 −0.3120 −0.1000 −0.1047 0.3109 (0.8035) (0.7756) (0.6419) (0.6688) (0.7126) Ratio-2017 −0.7080 −0.6466 −0.2643 −0.2782 0.1968 (0.6945) (0.6663) (0.6175) (0.6612) (0.7154) Large x Ratio-2013 2.0375 0.0461 0.3270 0.9647* 0.4769 (1.5753) (0.9804) (0.8855) (0.5086) (0.5403) Large x Ratio-2014 1.2791 0.2376 0.3509 0.8767* 0.7386* (1.1430) (0.6382) (0.6480) (0.5167) (0.3757) Large x Ratio-2015 0.7637 0.3066 0.2995 0.5271** 0.3756* (0.5856) (0.3033) (0.3196) (0.2558) (0.1928) Large x Ratio-2016 0.7860* 0.4283* 0.3666 0.5097** 0.3167** (0.4652) (0.2574) (0.2719) (0.2216) (0.1496) Large x Ratio-2017 0.8584* 0.5833* 0.4920 0.5551** 0.2983** (0.4641) (0.3213) (0.3251) (0.2171) (0.1461) Camp district −0.0099 −0.0009 −0.0105 −0.0948 −0.1383 (0.0729) (0.0630) (0.0574) (0.0945) (0.0969) N 2,198,542 2,184,341 2,184,341 2,162,167 1,570,824 Building controls — + + + + 12 region-year FE — — + + + Neighborhood FE — — — + — Building FE — — — — + Source: Authors’ estimations. Note: Ratio is the ratio of Syrian refugees to the native population at the level of 81 provinces. The camp district variable is a dummy indicating whether the district the house is located in contains a refugee camp. In all IV estimates, the weighted distance from a province to Syrian regions and their interactions with large houses for each year are used as instruments for the Syrian to the native population ratio and the interaction between the ratio and large houses. Large houses are defined as houses with more than the sample median of three rooms. Controls included in each column are shown at the bottom of the table. Building controls include age, whether the house is new, size, number of rooms (by type), a measure of building quality provided by the expert evaluation, the type of the heating system, and whether the building has an elevator, pool, or safety staff. The interactions between large houses and year fixed effects are added as controls in all models. Month, year, and province fixed effects are controlled in all the models. Standard errors are clustered at province level. ***p < 0.01, **p < 0.05, *p < 0.1. The World Bank Economic Review 171 Table 12. Impact on House Characteristics OLS 2SLS Downloaded from https://academic.oup.com/wber/article/37/1/147/6874774 by International Monetary Fund / World Bank - IMF user on 11 September 2023 (1) (2) (3) (4) (5) (6) A: Age Ratio-2013 −0.8047 −2.3546 −2.0422 2.4269 −4.4003** −3.6226 (1.3213) (1.7567) (1.7710) (2.4171) (2.1368) (2.2094) Ratio-2014 −2.0536** −2.3883* −1.7025 0.2579 −4.1848*** −3.1648*** (1.0257) (1.3112) (1.1767) (1.7910) (1.0286) (1.0953) Ratio-2015 −1.0561** −1.2871** −0.9065* 0.0323 −2.1298*** −1.3762*** (0.4801) (0.6110) (0.4987) (0.8807) (0.4267) (0.4733) Ratio-2016 −0.6192 −0.7223 −0.4877 0.2850 −1.3723*** −0.8560 (0.4150) (0.5223) (0.4858) (0.6300) (0.4920) (0.5297) Ratio-2017 −0.2974 −0.5295 −0.3815 0.7343 −1.0175* −0.5988 (0.4336) (0.5377) (0.4844) (0.6538) (0.5363) (0.4863) Camp district −0.0315 −0.0565 0.0366 −0.0554 −0.0492 0.0530 (0.1772) (0.1704) (0.0708) (0.1803) (0.1710) (0.0749) N 2,184,825 2,184,825 2,162,643 2,184,825 2,184,825 2,162,643 B: Size Ratio-2013 −1.0768 −0.3410 −0.7163 −0.5217 −0.9950 −1.5235* (0.6697) (0.6252) (0.6270) (1.1262) (0.9779) (0.8122) Ratio-2014 −0.9166* −0.2737 −0.6403 −0.3335 −0.7443 −1.0513 (0.5497) (0.4528) (0.4301) (0.9253) (0.7955) (0.6490) Ratio-2015 −0.5010* −0.1957 −0.3055* −0.1544 −0.3263 −0.4125 (0.2927) (0.2382) (0.1824) (0.4715) (0.4468) (0.3290) Ratio-2016 −0.4487* −0.1635 −0.2294 −0.1507 −0.1728 −0.2172 (0.2693) (0.2534) (0.1906) (0.4021) (0.4481) (0.3309) Ratio-2017 −0.4306 −0.1448 −0.2141 −0.1814 −0.1936 −0.2035 (0.2740) (0.2564) (0.1932) (0.3939) (0.4482) (0.3214) Camp district −0.0488 −0.0495 −0.0178 −0.0557* −0.0493 −0.0214 (0.0299) (0.0298) (0.0225) (0.0319) (0.0307) (0.0259) N 2,198,542 2,198,542 2,176,283 2,198,542 2,198,542 2,176,283 Month + + + + + + Year + + + + + + Province + + + + + + 12 region year FE — + + — + + Neighborhood — — + — — + Source: Authors’ estimations. Note: Ratio is the ratio of Syrian refugees to the native population at the level of 81 provinces. The camp district variable is a dummy indicating whether the district the house is located in contains a refugee camp. In all IV estimates, the weighted distance from a province to Syrian regions for each year are used as instruments for the ratio of Syrian refugees to native population. Age and size (in meters squared) variables are log transformed. Controls included in each column are shown at the bottom of the table. Standard errors are clustered at province level. ***p < 0.01, **p < 0.05, *p < 0.1. 6.3. House Sales, Mortgage Share, and Construction In table 14 we provide results for the effects of refugee inflows on house sales, mortgaged house sales, the share of mortgaged house sales, construction permits for residential buildings and for non-residential buildings. All variables are scaled by the province populations and all specifications include province, year, and NUTS-1 region-year fixed effects. Panel A shows OLS and panel B the 2SLS estimates. Consistent with a positive demand shock, we find an increase in sales following the inflow of refugees. Similar to the price effects, the impact on the number of sales is largest in 2013 and 2014, where we find increases of 2 to 2.5 percent. Nevertheless, the positive effect on house sales remains large and statistically significant in later years as well. Columns 2 and 3 further show that the increase is largely through mortgages and 172 ˘ Akgündüz, Hacıhasanoglu, and Yılmaz Table 13. Impact on Province Level Standard Deviation of Prices OLS 2SLS Downloaded from https://academic.oup.com/wber/article/37/1/147/6874774 by International Monetary Fund / World Bank - IMF user on 11 September 2023 (1) (2) (3) (4) Ratio-2013 0.5276*** 0.5339*** 0.5002* 0.7956** (0.1649) (0.1665) (0.2556) (0.3984) Ratio-2014 0.6972*** 0.6441*** 0.6500*** 0.7275*** (0.1180) (0.1125) (0.1904) (0.2633) Ratio-2015 0.1568*** 0.1292*** 0.1313 0.1803 (0.0592) (0.0479) (0.0917) (0.1382) Ratio-2016 0.0990** 0.1006** 0.0806 0.1481 (0.0441) (0.0433) (0.0659) (0.1122) Ratio-2017 0.0870** 0.0907** 0.0712 0.1301 (0.0381) (0.0443) (0.0590) (0.0990) N 567 560 567 560 Year + + + + Province + + + + 12 region-year FE — — + + Source: Authors’ estimations. Note: Ratio is the ratio of Syrian refugees to the native population at the level of 81 provinces. Controls included in each column are shown at the bottom of the table. Standard errors are clustered at province level. ***p < 0.01, **p < 0.05, *p < 0.1. that the share of mortgages in all house sales rises following the arrival of Syrian refugees. The increase in the share of mortgages is largest in 2013, where a percentage point increase in the refugee-to-native ratio raises the share of mortgages by 0.7 percent. While the effect size declines in later years, it remains positive and statistically significant. Since the share of mortgages rises after the arrival of Syrian refugees, effects on the house characteristics of houses sold through mortgages may be due to a change in composition. Furthermore, the effect on the share of mortgaged houses highlights the relevance of controlling for house characteristics, including location, in isolating the impact on prices when using the sample of mortgaged house sales. While the results from the first three columns are consistent with a rapid increase in demand for houses in host regions, we measure whether there is a response in supply in columns 4 and 5 by estimating the effects on the number of construction permits given for residential and non-residential buildings. Between 2013 and 2015, we find a statistically significant and positive effect across the presented specifications. The effect starts as high as an increase of 0.07 new construction permits for each percentage point increase in the Syrian-to-native population ratio in 2013. The effect size declines but remains statistically signif- icant until 2015.27 The impact disappears in 2016–2017, when the total number of refugees in Türkiye is relatively stable. The impact on new construction permits confirms that the inflow of refugees trig- gered new construction and increased housing supply in hosting regions. On the other hand, there is no equivalent impact on non-residential construction permits, where the estimated effects are close to 0. The increase in residential without an accompanying increasing non-residential construction further confirms that the former effect is in response to an increase in residential house demand rather than a decline in construction costs. Table 15 shows the impact of Syrian refugee inflows on loan-to-value ratios of mortgages issued. If the Syrians’ impact on local economic activity and local credit markets is strong enough, there may be a positive feed to the loan-to-value ratios during the years of housing price increases observed in table 6. Unlike the estimated effects on prices, we find statistically insignificant and negative effect on 27 This result overlaps with Cengiz and Tekgüç (2021) who find an increase in construction permits in hosting regions with data up to 2015. The World Bank Economic Review 173 Table 14. Impact on Sales—Mortgages and Construction (1) (2) (3) (4) (5) Downloaded from https://academic.oup.com/wber/article/37/1/147/6874774 by International Monetary Fund / World Bank - IMF user on 11 September 2023 Total Mortgaged Share of Residential Non-residential sales sales mortgages construction construction A: OLS Ratio-2013 0.0217*** 0.0077*** 0.7043** 0.0627*** −0.0001 (0.0027) (0.0015) (0.3127) (0.0063) (0.0002) Ratio-2014 0.0263*** 0.0069*** 0.4369*** 0.0285*** −0.0001 (0.0015) (0.0005) (0.1491) (0.0065) (0.0001) Ratio-2015 0.0128*** 0.0041*** 0.2728*** 0.0090*** −0.0000 (0.0009) (0.0003) (0.0721) (0.0017) (0.0000) Ratio-2016 0.0082*** 0.0017*** 0.1487*** −0.0023* 0.0000 (0.0008) (0.0004) (0.0532) (0.0013) (0.0000) Ratio-2017 0.0059*** 0.0021*** 0.1918*** −0.0032 −0.0001 (0.0011) (0.0004) (0.0537) (0.0020) (0.0001) B: 2SLS Ratio-2013 0.0205* 0.0113** 1.2880* 0.0715*** −0.0004 (0.0111) (0.0056) (0.6563) (0.0157) (0.0004) Ratio-2014 0.0215*** 0.0077*** 0.7041** 0.0279** −0.0004 (0.0070) (0.0017) (0.3144) (0.0131) (0.0004) Ratio-2015 0.0097*** 0.0042*** 0.3846** 0.0117*** −0.0001 (0.0035) (0.0009) (0.1473) (0.0044) (0.0001) Ratio-2016 0.0069*** 0.0025** 0.2661** −0.0024 −0.0000 (0.0020) (0.0010) (0.1272) (0.0030) (0.0001) Ratio-2017 0.0062*** 0.0035** 0.3121** −0.0012 −0.0004 (0.0020) (0.0014) (0.1254) (0.0047) (0.0004) N 560 560 560 560 560 Source: Authors’ estimations. Note: Ratio is the ratio of Syrian refugees to the native population at the level of 81 provinces. The dependent variables are the ratio of sales, mortgaged sales, residential and non-residential construction permits to population. The share of mortgages is the ratio of mortgaged sales to all sales. The regressions are estimated at the province-year level. In all IV estimates, the weighted distance from a province to Syrian regions for each year are used as instruments for the ratio of refugee to native population. Age and size (in meters squared) variables are log transformed. All specifications include province, year, and 12 region-year (NUTS-1) fixed effects. Standard errors are clustered at province level. ***p < 0.01, **p < 0.05, *p < 0.1. loan-to-value ratios that ranges between −0.05 and −0.3 percentage points between 2013 and 2015. The estimated effects turn positive, though are small, in 2017. The estimates on loan-to-value ratios suggest that the rise in housing prices does not seem to be driven by an increase in the income of natives. In fact, the negative estimates suggest that the mortgage credits did not keep up with the increase in prices be- tween 2013 and 2015 when the price effect of refugee inflows was largest. In other words, the increase in house prices with higher Syrian population is not driven by credit market adjustments due to an increase in economic activity and income. 7. Conclusion We investigated the impact of Syrian refugee inflows in Türkiye on the housing market using data on the population of house mortgages. Since the mortgage market is in practice mostly available for natives, our results capture the impact on housing prices for natives. We present findings on house prices, loan-to-value ratios, house characteristics, mortgaged sales, and construction. We find an increase in prices that is concentrated in low-price housing. The prices for low-price hous- ing increased by over 5 percent per percentage increase in the Syrian-to-native population ratio during the years immediately after the refugee crisis. Average prices have also increased by up to 3 percent per 174 ˘ Akgündüz, Hacıhasanoglu, and Yılmaz Table 15. Impact of Syrian Refugees—Loan-to-Value Ratios (1) (2) (3) (4) (5) Downloaded from https://academic.oup.com/wber/article/37/1/147/6874774 by International Monetary Fund / World Bank - IMF user on 11 September 2023 A: OLS Ratio-2013 −0.2160 −0.2844 −0.0159 −0.0446 −0.1590 (0.1580) (0.2085) (0.1445) (0.1722) (0.1309) Ratio-2014 −0.1612 −0.2168 0.0039 −0.0277 −0.1022 (0.1142) (0.1492) (0.0950) (0.0924) (0.0917) Ratio-2015 −0.0422 −0.0797 0.0302 0.0192 −0.0057 (0.0405) (0.0640) (0.0410) (0.0413) (0.0332) Ratio-2016 −0.0066 −0.0444 0.0208 0.0168 0.0108 (0.0309) (0.0518) (0.0470) (0.0464) (0.0393) Ratio-2017 0.0454* 0.0011 0.0590 0.0503 0.0263 (0.0271) (0.0479) (0.0421) (0.0476) (0.0461) Camp district 0.0054 0.0008 −0.0002 0.0025 0.0150*** (0.0068) (0.0060) (0.0059) (0.0086) (0.0056) B: 2SLS Ratio-2013 −0.1031 −0.1308 −0.0469 −0.1098 −0.1712 (0.2295) (0.3276) (0.2156) (0.2041) (0.1517) Ratio-2014 −0.1150 −0.1227 0.0067 −0.0323 −0.1468 (0.1831) (0.2461) (0.1530) (0.1357) (0.1327) Ratio-2015 0.0409 0.0253 0.0605 0.0575 0.0181 (0.0777) (0.1102) (0.0505) (0.0386) (0.0324) Ratio-2016 0.1054* 0.0869 0.0891** 0.0892** 0.0879** (0.0632) (0.0869) (0.0411) (0.0367) (0.0421) Ratio-2017 0.1508*** 0.1173 0.1250*** 0.1286*** 0.1014* (0.0571) (0.0779) (0.0384) (0.0377) (0.0524) Camp district 0.0028 −0.0021 −0.0013 −0.0040 0.0097 (0.0072) (0.0068) (0.0061) (0.0101) (0.0080) N 2,174,262 2,160,097 2,160,097 2,138,110 1,550,055 Month + + + + + Year + + + + + Province + + + + + Building controls — + + + + 12 region-year FE — — + + + Neighborhood FE — — — + — Building FE — — — — + Source: Authors’ estimations. Note: Ratio is the ratio of Syrian refugees to the native population at the level of 81 provinces. The camp district variable is a dummy indicating whether the district the house is located in contains a refugee camp. In all IV estimates, the weighted distance from a province to Syrian regions for each year are used as instruments for the ratio variable. Controls included in each column are shown at the bottom of the table. Building controls include age, whether the house is new, size, number of rooms (by type), a measure of building quality provided by the expert evaluation, the type of the heating system, and whether the building has an elevator, pool, or safety staff. Standard errors are clustered at province level. ***p < 0.01, **p < 0.05, *p < 0.1. percentage increase in refugee-to-population ratio in the same period. Given that Syrian refugees are heav- ily concentrated and the ratio of Syrians to natives reached over 20 percent in multiple border provinces by 2015, the effect on house prices appears very significant. We further document that these effects faded over time, with estimated effects that are statistically insignificant and small by 2017. The results point to a large demand shock and an inelastic supply curve in the short run, which adjusted to the inflow in later years. This interpretation is supported by an increase in new residential construction permits and a decrease in the average age of houses sold in provinces hosting refugees. There is further no corresponding increase in loan-to-value ratios to indicate a rise in the availability of mortgages. The effects we observed The World Bank Economic Review 175 in the Turkish housing market indicate a similar response to refugees in the housing market to the findings of Rozo and Sviatschi (2021) for Jordan. Downloaded from https://academic.oup.com/wber/article/37/1/147/6874774 by International Monetary Fund / World Bank - IMF user on 11 September 2023 Our results should be interpreted alongside other studies of the impact of Syrian refugees on the Turkish economy. Our results underline the difficulties in maintaining amenities and infrastructure quality when migration inflows are sudden and concentrated in specific regions. The sharp increase in low-priced hous- ing is especially worrying from a welfare perspective given that the hardest-hit segments of the working native population were informal workers with low wages (Ceritoglu et al. 2017; Akgündüz and Torun 2020; Aksu, Erzan, and Kırdar 2022). Although the effects in the housing market are temporary, they lasted several years, which can adversely affect low-income natives in host provinces. Data Availability Statement The data underlying this article were provided by the Central Bank of the Republic of Türkiye by permis- sion. Interested researchers need to obtain permission from the Central Bank of the Republic of Türkiye to work with the data set. The authors would be willing to share any relevant codes to assist in replication efforts. References Accetturo, A., F. Manaresi, S. Mocetti, and E. Olivieri. 2014. “Don’t Stand So Close to Me: The Urban Impact of Immigration.” Regional Science and Urban Economics 45: 45–56. Akbari, A. H., and Y. Aydede. 2012. “Effects of Immigration on House Prices in Canada.” Applied Economics 44(13): 1645–58. ˘ 2021. “Immigration and Inter-Regional Job Mobility: Evidence from Syrian Akgündüz, Y. E., A. Altan, and Y. K. Bagır. Refugee Inflows in Türkiye.” Working Papers 1461, Economic Research Forum. ˘ S. M. Cilasun, and M. G. Kirdar. 2022. “Consequences of a Massive Refugee Influx on Akgündüz, Y. E., Y. K. Bagır, firm Performance and Market Structure.” CReAM Discussion Paper Series 2203, Centre for Research and Analysis of Migration (CReAM), Department of Economics, University College London. Akgündüz, Y. E., and H. Torun. 2020. “Two and a Half Million Syrian Refugees, Tasks and Capital Intensity.” Journal of Development Economics 145: 102470. Akgündüz, Y. E., M. van den Berg, and W. Hassink. 2018. “The Impact of the Syrian Refugee Crisis on Firm Entry and Performance in Türkiye.” World Bank Economic Review 32(1): 19–40. Aksu, E., R. Erzan, and M. G. Kırdar. 2022. “The Impact of Mass Migration of Syrians on the Turkish Labor Market.” Labour Economics: 102183. ˘ O., O. Baki¸ Altındag, s, and S. V. Rozo. 2020. “Blessing or Burden? Impacts of Refugees on Businesses and the Informal Economy.” Journal of Development Economics 146: 102490. Balkan, B., E. O. Tok, H. Torun, and S. Tumen. 2018. “Immigration, Housing Rents, and Residential Segregation: Evidence from Syrian Refugees in Türkiye.” IZA Discussion Papers 11611. Institute for the Study of Labor (IZA). https://ideas. repec. org/p/iza/izadps/dp11611. html. Barker, M., and F. Pöge. 2017. “Strdist: Stata module to Calculate the Levenshtein Distance, or Edit Distance, between Strings.” Barone, G., F. David, G. De Blasio, and S. Mocetti. 2021. “How Do House Prices Respond to Mortgage Supply?” Journal of Economic Geography 21(1): 127–40. Bianchi, M., P. Buonanno, and P. Pinotti. 2012. “Do Immigrants Cause Crime?” Journal of the European Economic Association 10(6): 1318–47. Bratsberg, B., and O. Raaum. 2012. “Immigration and Wages: Evidence from Construction.” Economic Journal 122(565): 1177–205. Cameron, A. C., and D. L. Miller. 2015. “A Practitioner’s Guide to Cluster-Robust Inference.” Journal of Human Resources 50(2): 317–72. Card, D., C. Dustmann, and I. Preston. 2012. “Immigration, Wages, and Compositional Amenities.” Journal of the European Economic Association 10(1): 78–119. 176 ˘ Akgündüz, Hacıhasanoglu, and Yılmaz Cengiz, D., and H. Tekgüç. 2022. “Is It Merely a Labor Supply Shock? Impacts of Syrian Migrants on Local Economies in Türkiye.” ILR Review: 75(3): 741–68. Downloaded from https://academic.oup.com/wber/article/37/1/147/6874774 by International Monetary Fund / World Bank - IMF user on 11 September 2023 Ceritoglu, E., H. B. G. Yunculer, H. Torun, and S. Tumen. 2017. “The Impact of Syrian Refugees on Natives’ Labor Market Outcomes in Türkiye: Evidence from a Quasi-Experimental Design.” IZA Journal of Labor Policy 6(1): 5. Cochrane, W., and J. Poot. 2021. “Effects of Immigration on Local Housing Markets.” Economic Geography of Cross- Border Migration: 262–92. Degen, K., and A. M. Fischer. 2017. “Immigration and Swiss House Prices.” Swiss Journal of Economics and Statistics 153(1): 15–36. Del Carpio, X. V., and M. C. Wagner. 2015. The Impact of Syrians Refugees on the Turkish Labor Market. Policy Research Working Paper Series 7402, The World Bank. Demirci, M., and M. G. Kirdar. 2023. “The Labor Market Integration of Syrian Refugees in Türkiye.” World Devel- opment, 162: 106138. Depetris-Chauvin, E., and R. J. Santos. 2018. “Unexpected Guests: The Impact of Internal Displacement Inflows on Rental Prices in Colombian Host Cities.” Journal of Development Economics 134: 289–309. Diamond, R. 2017. “Housing Supply Elasticity and Rent Extraction by State and Local Governments.” American Economic Journal: Economic Policy 9(1): 74–111. sci, and E. Ö. Çarmıklı. 2013. Türkiye and Syrian Refugees: The Dinçer, O. B., V. Federici, E. Ferris, S. Karaca, K. Kiri¸ Limits of Hospitality. International Strategic Research Organization (USAK). Edo, A., Y. Giesing, J. Öztunc, and P. Poutvaara. 2019. “Immigration and Electoral Support for the Far-Left and the Far-Right.” European Economic Review 115: 99–143. Follain, J. R., and R. M. Dunsky. 1997. “The Demand for Mortgage Debt and the Income Tax.” Journal of Housing Research: 155–199. Gonzalez, L., and F. Ortega. 2013. “Immigration and Housing Booms: Evidence from Spain.” Journal of Regional Science 53(1): 37–59. Harter-Dreiman, M. 2004. “Drawing Inferences about Housing Supply Elasticity from House Price Responses to Income Shocks.” Journal of Urban Economics 55(2): 316–37. Jaeger, D. A., J. Ruist, and J. Stuhler. 2018. “Shift-Share Instruments and the Impact of Immigration (No. w24285).” National Bureau of Economic Research. Mian, A., and A. Sufi. 2009. “The Consequences of Mortgage Credit Expansion: Evidence from the US Mortgage Default Crisis.” Quarterly Journal of Economics 124(4): 1449–96. Moraga, J. F.-H., Carbonell A. Ferrer-i, and A. Saiz. 2019. “Immigrant Locations and Native Residential Preferences: Emerging Ghettos or New Communities?” Journal of Urban Economics 112(C): 133–151. Ousey, G. C., and C. E. Kubrin. 2018. “Immigration and Crime: Assessing a Contentious Issue.” Annual Review of Criminology 1: 63–84. Rozo, S. V., and M. Sviatschi. 2021. “Is a Refugee Crisis a Housing Crisis? Only If Housing Supply Is Unresponsive.” Journal of Development Economics 148: 102563. Sá, F. 2015. “Immigration and House Prices in the UK.” Economic Journal 125(587): 1393–424. Saiz, A. 2007. “Immigration and Housing Rents in American Cities.” Journal of urban Economics 61(2): 345–71. ———. 2010. “The Geographic Determinants of Housing Supply.” Quarterly Journal of Economics 125(3): 1253–96. Tabellini, M. 2020. “Gifts of the Immigrants, Woes of the Natives: Lessons from the Age of Mass Migration.” Review of Economic Studies 87(1): 454–86. TEPAV. 2018. Syrian Entrepreneurship and Refugee Start-ups in Türkiye: Leveraging the Turk- ish Experience: Final Report 2018. TEPAV. https://www.tepav.org.tr/upload/files/1566830992- 6.TEPAV_and_EBRD___Syrian_Entrepreneurship_and_Refugee_Start_ups_in_Turkey_Lever....pdf Tumen, S. 2016. “The Economic Impact of Syrian Refugees on Host Countries: Quasi-Experimental Evidence from Türkiye.” American Economic Review 106(5): 456–60. ———. 2019. “Refugees and ‘Ative Flight’ from Public to Private Schools.” Economics Letters 181: 154–9. Tyrcha, A. 2020. “Migration and Perceptions of Housing Availability in Sweden.” Papers in Regional Science 99(4): 945–975. UNHCR. 2012. “UNHCR Türkiye Syrian Refugee Daily Sitrep December 2012.” Daily report, The United Nations Refugee Agency. ———. 2013. “UNHCR Türkiye Syrian Refugee Daily Sitrep December 2013.” Daily report, The United Nations Refugee Agency.