Addressing Vacancy and Affordability Challenges: An Analysis of the Impact of Vacant House Tax on Urban Housing Prices Title Page Dr. Jian Liang Senior Lecturer of Property and Economics School of Economics and Finance Queensland University of Technology E-mail Address: [email protected] #Dr. Kang Mo Koo Associate Professor Department of Economics Yonsei University, Mirae Campus 1 Yonseidae-gil, Wonju Gangwon 26493 Republic of Korea E-mail address: [email protected] # Corresponding author
Addressing Vacancy and Affordability Challenges: An Analysis of the Impact of Vacant House Tax on Urban Housing Prices Abstract In response to the rising number of unoccupied residential properties and the growing affordability crisis in urban centers, certain cities have introduced the Vacant House Tax—a levy based on a percentage of property value for vacant holdings. While it is well-established that vacant, abandoned, and foreclosed properties exert downward pressure on local housing prices, the specific effects of such a vacancy tax system on housing prices within the relevant area remain uncertain. Drawing on transactional and policy data sourced from the Greater Melbourne area in Australia, this study delves into the post-implementation shifts in housing prices following the introduction of the Vacant Residential Land Tax. Our investigation reveals a notable decline of 2%–3% in housing prices as a direct result of the vacancy tax. This decrease is consistent between transactional and rental housing prices. Keywords: housing price, vacancy rate, vacancy tax, rent price JEL Code: D1, G1, H2, R2, R3, R5
Introduction As urban environments have suffered from overurbanization and there is a high density of population in urban areas, more attention is being paid to the issue of housing availability and affordability. One of the issues that urban policymakers have faced is vacant houses that have not been fully utilized, as the vacant units are not continuously occupied by residents or leased out to tenants to ease the affordability problem. To disentangle the issue of housing affordability through the channel of vacant houses, a few cities have implemented the vacancy tax policy that imposes extra tax on the houses that have been left vacant for a significant period. Even though the issue of vacancy in the urban context has been widely discussed thus far, the main focus has been on the increase of incidents such as fire or crime in areas with high vacancy rate1 (Branas et al., 2018; Chen and Rafail, 2020; Schachterle et al., 2012; Walter et al., 2024; Zhang and McCord, 2014). While the purpose of vacancy tax is to maximize the utilization of spatial assets that remain unoccupied for various reasons, it can also affect housing prices by affecting the merit of owning houses as an investment property or a vacation home. There are extant papers that study the impact of vacancy tax on the changes in vacancy rates after the introduction of vacancy tax. However, only a few studies have investigated the impact of such a tax policy on housing prices. To fill in the gap in the literature, this study investigates the impact of the implementation of vacancy tax on housing price in the case of Melbourne, Australia, and extends the scope of investigation to the rental rate and changes in vacancy rate. Thus, we comprehensively highlight the empirical outcome of vacancy tax on price, rent, and vacancy. 1 It has been also discussed that vacant properties and land can affect the physical and mental health of the residents living in the neighborhood (Garvin et al., 2013; Kvik et al., 2022).
To preview the main findings of this study, we find that the implementation of vacancy tax lowers the housing price level by 2.1%–3.4% depending on the buffer area from the boundary of the area that is subject to the policy. The negative impact on the price indicates that the new policy affected the sentiment of property investors at least for 2 years after the implementation. Additionally, the rental rate of the leased properties is lower after the introduction of the vacancy tax policy. We conjecture that the lower rental rate resulted from the change in the supply level in the rental market, as landlords are incentivized to lease out their properties sooner to avoid the tax levy. This conjecture is confirmed by the changes in the vacancy rate. We find that the overall vacancy rate in the pertinent area reduces by 0.2%. With these findings, we conclude that the vacancy tax policy is an effective measure to enhance the affordability and availability of residential properties, at least in the context of the Greater Melbourne area. Robustness tests with various model specifications reconfirms the efficacy of the vacant house tax. The next section introduces the background of the vacant house policy, with a few examples of similar policies in other countries. In the following section, we review the literature regarding the impact of tax on the housing price and the issues of vacant properties in the urban context. We then describe the database and model specification, followed by the main results of the analysis and the additional analysis including geographic falsification analysis, robustness test with entropy balancing, regression discontinuity analysis, and investigation of the changes in vacancy rates. In the subsequent section, the entire paper is concluded with a discussion on the policy implications. Background There are different contexts of vacant houses in urban research. The first thread relates to houses that have been left vacant and abandoned. The social problem resulting from abandoned houses is more pronounced in aging societies, such as Japan and Korea. Sadayuki et al. (2020)
investigate the negative externality of vacant houses in Toshima municipality in Tokyo and find that vacant houses negatively affect the rental price by 0.9%–1.7%; this negative impact is exacerbated when the physical aspect of vacant houses do not meet the general standard, such as slanted buildings with poor maintenance. However, the phenomena here is related to the decreasing population in the research area. On the other hand, other cities face the same issue of vacancy through different channels, where the urban density and demand for housing is high. Due to the high demand for investment in residential properties, or ownership by foreigners, a few countries, such as France, Canada, and Australia, have introduced a vacancy tax policy to reduce the vacancy rate and increase housing affordability. The city of Vancouver approved the Empty Home Tax (EHT) and enacted the vacancy tax bylaw in November 2016. According to this bylaw, residential properties that are left vacant for over six months are subject to the EHT. Based on the report produced by the city of Vancouver, 53% of the households in Vancouver were renters. Accordingly, the focus of the vacancy tax was to enhance the affordability in the rental housing market. After the implementation of the new tax scheme, the market witnessed a significant improvement in the rental market. Figure 1 presents the change in the vacancy rate and tenancy rates of residential units in Vancouver. [Figure 1 here] Since the introduction of the vacancy tax policy, the vacancy rate continuously dropped from 1.36% to 0.71%. On the other hand, the figure reveals a significant increase in the proportion of tenanted units. The tenancy rate, which is the ratio of tenanted units over the total residential units, increased from 25% to 29% in four years. Although this figure simply reveals the change in vacancy and tenancy rates without controlling for other variables, the policy appears to have worked to a certain extent. Additionally, it is notable that the tax rate has increased from 1% of
the assessed value when the policy was first implemented to 1.25% in 2020 and 3% in 2021. Even though the city of Vancouver provided for eight exceptional cases2 for the exemption of the vacancy tax, strict regulation on vacant properties and increasing tax rate appear to have affected the behavior of landlords, which resulted in a decrease in the vacancy rate. Moreover, the tax revenue collected from the EHT program increased from 38 million Canadian dollars to 67 million Canadian dollars (Housing Vancouver, 2022, 2021, 2020). This tax revenue was allocated to support affordable housing initiatives that can enhance the housing supply in Vancouver. In France, tax on vacant houses was introduced before Canada. A new tax policy on vacant housing was passed in July 1998 and became effective from January 1, 1999, onward. This new tax policy, called taxe sur les logements vacants or TLV) added a new layer of tax burden on homeowners who left their units vacant for over two years. The owners with vacant units for over two consecutive years are required to pay 10% of the rental value. Moreover, the tax rate increases with a longer term of vacancy, with units vacant for three to four years subject to a tax rate of 12.5%, and those vacant for longer to a tax rate of 15% (Segú, 2020). Subsequently, in 2013, the TLV was modified with a higher tax rate and stricter threshold. The tax rate increased to 12.5% of the rental value for the first year, and 25% for the second year. Moreover, from 2013 onward, properties left vacant for over one year were subject to the TLV instead of the two-year period. Similar to the outcome in Vancouver, the municipalities that were subject to the TLV experienced an improvement in the vacancy rate. Between 1995 and 2005, the vacancy rate of the municipalities with the TLV policy decreased from 6.5% to 5%, 2 Eight exceptional cases are 1) death of the registered owner, 2) undergoing redevelopment or major renovations, 3) owner in care (owner in hospital or care facility), 4) strata rental restriction (rentals are prohibited or maximum allowable number of rental units are already reached), 5) occupancy for fulltime employment (employer requires physical presence in Vancouver), 6) transfer of property, 7) court order, and 8) limited use residential property, which implies that the use of properties are limited to vehicle parking or the land parcel has limited use for residential purposes.
while the other municipalities without TLV underwent only a reduction of 0.5% in the vacancy rate (Segú, 2020). The state of Victoria introduced a similar tax policy, called Vacant Residential Land Tax or VRTL, with a stricter definition of vacant units. Contrary to the definition of vacant units in France, the state of Victoria classified residential units left vacant for six months as vacant units subject to the vacancy tax. Fitzgerald (2020) reports that 69,004 properties (4.1% of total properties) are estimated to be vacant based on water usage data, and this figure increased by 13.3% between 2017 and 2019 in the Greater Melbourne area. Thus, the issue of vacant properties in Victoria is not a negligible phenomenon. With the elevating concern of the local government, this new policy targeting vacant properties became effective from January 2018 onward, and the initial tax rate was set to be 1% per annum on the capital improved value of properties that have been left vacant for over 6 months in 16 councils3 located in the inner and middle Melbourne area. Figure 2 presents the location of the councils in the greater Melbourne council map. [Figure 2 here] Similar to the case of Vancouver, the VRTL in Melbourne also allowed 15 exemptions, including properties that are used as primary residences, retirement villages, municipal and public land, etc. Like other countries, which adopted the vacancy tax policy, devising measures to sort out vacant properties is not an easy task from the perspective of local authorities. In case of the state of Victoria, it was reported that the local authority would rely on the self-reporting system and utilize the utility usage data to match with the self-reporting system and vacant 3 16 councils subject to the vacant residential land tax are Banyule, Bayside, Boroondara, Darebin, Glen Eira, Hobsons Bay, Manningham, Maribyrning, Melbourne, Monash, Moonee Valley, Merri-bek, Port Phillip, Stonnington, Whitehorse, and Yarra.
properties4 . It was the goal of the local authority to increase the tax revenue by approximately AU$20 million per annum and utilize the fund to enhance housing affordability in Victoria5 . Contrary to the expectations of the potential tax revenue, Fitzgerald (2020) reported that the revenue raised per year remained below AU$ 7 million between 2017 and 2020. Based on the estimated number of vacant properties, the actual number of self-reported vacant properties for vacancy tax accounted for merely 2.6% in the 2019–2020 tax year. This figure indicates that the measures of tax enforcement need to be improved in order to be more effective. Literature Review Taxation has a close relationship with the pricing of properties. Dusansky et al. (1981) indicate that the housing price is negatively related to an increase in the tax rate. As they explain in their paper, the following discounting formula is used in extant literature: = ∑ ( − ) (1 + ) =1 where R is gross rental income; T is a property tax; and r stands for the rate of discount. In the case of vacancy tax, any change in T can be avoided if landlords actively rent out their properties. To rent out properties quickly, R is likely to decrease. On the other hand, lessors might be willing to shift the potential risk of an increase in tax to renters and increase the rental rate (Autor et al., 2014; Heinberg and Oates, 1970; Kee and Moan, 1976; Orr, 1972, 1968; Xiao and Zhou, 2023). Thus, it is not clear how the changes in the tax system would affect the 4 For more information, please see https://www.afr.com/property/victorias-vacant-housing-tax-1-percent-seen-as-thin-edge-of-the-wedge-20170307-gus821 5 After the implementation of the VRTL in Melbourne, it was reported that the number of vacant dwellings in a few cities increased, which indicates that the VRTL failed to a certain extent. However, the purpose of this paper is to identify the price change in the effective area after the implementation, and the issue of affordability change or change in rental supply level is not the major concern. For more information regarding the increase in vacant dwellings, please refer to the news article at < https://www.proquest.com/newspapers/vacant-dwellings-melbourne-up13pc/docview/2465712184/se-2?accountid=15179>
housing price. Moreover, as Palmon et al. (2017) indicate in their paper, empirical papers reveal that the impact of tax on capitalization does not have a consensus. They also argue that when tax revenue is reinvested, which cannot be directly observed, the price of houses is spuriously correlated with the tax. Based on the capitalization model introduced in Palmon et al. (2017), the value of a house is determined by the following equation: = () + where is the value of the jth property; () is rental value from the hedonic function; in the net user cost of housing; and is the annual property taxation. The issue of spurious correlation occurs when the hedonic function of is composed of observed and unobserved components. Moreover, when the unobserved component is correlated with the annual property tax, the coefficients would be biased. This is likely to happen when the tax revenue is re invested in the same tax area. In this paper, we assume that the correlation between the hedonic variables and the tax effect would be minimized, as we classify the control and treatment groups by the tax policy and empirically show the changes in the rental rates and housing prices. Further, residential properties are left vacant for various reasons. These reasons could include oversupply of houses (Molloy, 2016), market friction (Wheaton, 1990), tax rate (Arsen, 1992), or restricted opportunities of redevelopment (Sadayuki et al., 2020). Moreover, vacant properties are believed to have negative impacts on the neighboring properties (Accordino and Johnson, 2000; Hartley, 2014; Mikelbank, 2008; Sadayuki et al., 2020; Whitaker and Fitzpatrick IV, 2013). Whether the owner of the vacant property is willing to leave properties vacant for future sales or the owner is undergoing a case of foreclosure, the outcome is similar—poor management of the property, such as long grass in the yard and a deteriorating façade. Whitaker and Fitzpatrick IV (2013) find that vacant or delinquent properties located within 500 ft drag down the prices in the area by 1%–2%. Mikelbank (2016) and Hartley (2014)
investigated the impact of vacant properties more precisely. While the extant studies were unable to clearly identify vacant units across vacant, abandoned, or foreclosed properties, Mikelbank (2016) successfully segregated the impact of each factor on the housing price and found that the impact of the foreclosed properties is much larger than that of vacant or abandoned properties. Hartley (2014) also investigated the impact of vacant/foreclosed properties on the housing price in the same area and finds that the additional supply of housing decreases prices by 1.2% and the dis-amenity caused by the foreclosed property is almost negligible. Regardless, extant studies have only focused on the impact of vacant/abandoned/foreclosed properties on the prices of houses in the vicinity or the administrative area of the subject property. Our study investigates the issue of vacancy in the opposite direction. When the vacancy rate is anticipated to be lower after levying the vacancy tax, probably due to higher supply of rental/investment properties on the market, our focus of interest lies on the change in housing prices. As extant papers have argued, when vacant properties have a negative impact on adjacent properties, an increase in housing supply will also drag down the prices. On the positive side, with a higher density of residents, the city will be more vibrant and tax revenue from the vacancy tax policy can be reinvested to enhance the quality of the built environment. A few extant papers directly study the impact of the vacancy tax in other countries, such as Canada and France. Hu (2018) and Wang and Zhang (2019) investigate the efficacy of vacant tax policy in the case of British Columbia, Canada. Hu (2018) finds that the vacancy tax in British Columbia has a positive impact on the supply of residential properties, which eases housing affordability and lowers the value of new residential projects. However, the argument is less convincing as variables linked to the construction cost, which is a key variable in analyzing the new construction of real estate properties, is not considered in the model specification. Moreover, the connecting channel is not elaborated upon to explain the positive
change in the housing supply when the burden of the housing cost increases due to the increase in the tax rate. Wang and Zhang (2019) reveal mixed results of the price change after the introduction of vacancy tax policy in British Columbia by utilizing the housing price index (HPI) covering a five-year period before and after the policy implementation. They find no significant changes in the overall housing price level when the HPI in Vancouver and Toronto are compared. However, we find that an appropriate specification of fixed effects is not included in the model specifications, which might be the reason for no significant difference in the HPI. A thread of studies analyzed the impact of the vacancy tax on the supply of residential properties in the rental market. Segú (2020) has empirically shown that the vacancy rate decreased by 13% after the introduction of vacancy tax in France. More recently, Han et al. (2023) argued that the introduction of the Empty Home Tax (EHT), which has been effective in Vancouver, Canada, since 2017, can increase the number of new listings and the number of transactions as a few speculative vacancies are released in the rental and sales market. However, they also show that the EHT can worsen housing affordability in the long run, as property developers are less likely to be incentivized to supply new units to the market due to the new tax burden on the investors. While a few studies on vacancy tax have investigated the impact of vacancy tax policies in a few different countries, a few papers have focused on the housing price after the implementation of vacancy tax with micro data along with the comprehensive impact on affordability. Thus, in this paper, we attempt to ascertain how the housing price market responds when two opposite forces are competing. Research Design In this study, we considered the impact of implementing the vacancy tax as the treatment performed on the housing market. Accordingly, we identified the areas that implement the
vacancy tax as the treatment areas and the remaining areas as the control areas. Further, the vacancy tax was implemented on January 1, 2018; thus, we identified the two years after the implementation date (2018 and 2019) as the treatment period and the two years before as the control period (2016 and 2017). Consequently, we developed the Spatial Difference-inDifferences (SDD) model to estimate the causal effect of implementing the vacancy tax on the housing value. The SDD model has been widely adopted to estimate the causal effect of events similar to implementing vacancy tax on housing value, and it has been proven effective in controlling endogeneity and variable-omitting biases in event studies (Segú, 2020; Koo and Liang, 2022). Our baseline model is developed as given below. ln (,) = + × ( × ) + × + × + × ′ , + + + ,, Eq. (1) where (,,) is the sale or rental price that takes the logarithm for property i at time t. The dummy variable indicates the treatment area, which is the area subject to the vacancy tax. The dummy variable indicates the treatment period—that is, the period from January 1, 2018, to December 31, 2019. The coefficient of the interaction term of and measures the causal effect of implementing the tax as a DID coefficient. Further, we included ′ , , which stands for hedonic variables—such as the number of bedrooms, number of bathrooms, number of parking spots, the logarithm of the floor area, and distance to the Melbourne CBD (Central Business District)—and dummy variables that indicate the residential property type. Additionally, we included the time-fixed effect (year and month) and suburb fixed effect , as a suburb is the smallest housing submarket in the Australian context. Moreover, we cluster the standard errors by suburb year to ensure that the residuals are independent and identically distributed.
Further, we restrict our main test sample area to the area adjacent to the boundary of the taxed area, specifically within one kilometer of the boundary, to minimize the impacts of confounding factors during the sample period (Koo and Liang, 2022). We also adjust the boundary distance to 0.5 km and 1.5 km when specifying the treatment area in robustness tests. In empirical tests, we also attempt different model settings by excluding all the control variables and including additional dummy variables that indicate if the property has features like a study room, sunroom, swimming pool, fireplace, etc., to address the possible confounding issue of change in housing preferences during the sample period. Moreover, we develop a dynamic DID model to validate the common trend assumption. Specifically, we split the entire sample period by quarter, using sixteen dummy variables, eight for the control/pre-trend period from 2016 to 2017 and eight for the treatment period from 2018 to 2019. Then, we interact these quarterly variables with the dummy variable , which indicates the treatment period, to construct dynamic variables. We set the quarter before implementing the tax (last quarter of 2017) as the base period and develop the dynamic DID model as shown below: ln (,) = + ∑ × (, × ) 7 =1 + ∑ × (, × ) 8 =1 + ∑ × (,) 7 =1 + ∑ × (,) 8 =1 + × ( ) + × ′ , + + + , Eq. (2) where , indicates the quarters before implementing the tax, , indicates the quarters after implementation, and is the pre-trend DID coefficient used to validate the common trend assumption. The remaining model settings are the same as those in model Eq. (1).
Sample Description Our main tests utilizing the above SDID models are performed on a database that includes housing transactional data in the state of Victoria, Australia. The housing transactional data is collected from the Australian Urban Research Infrastructure Network (AURIN). The data set includes the following information: sale price, sale date, property type, number of bedrooms, number of bathrooms, number of parking spaces, floor area, and geocode information (i.e., latitude and longitude). The database also contains additional hedonic features like a swimming pool, additional study room, fireplace, etc. Further, we collected the GIS files of the VIC LGA map from Data Victoria, an open data source operated by the state of VIC, to delineate the inner-Melbourne boundary subject to vacancy tax. The sample period is from January 1, 2016 (two years before implementing the vacancy tax) to December 31, 2019 (two years after the vacancy tax). Further, to minimize the possible impact of other confounding factors like changes in housing preferences, macroeconomic environment, and other housing policies, we follow the literature (Koo and Liang, 2022) to narrow down the sample area to within one kilometer of the boundary of inner Melbourne where the tax was implemented in the main tests. We present the statistical summary of key variables in Table 1 below. [Table 1 here] According to the statistical summary, the properties within the boundary of the tax area have higher prices and smaller sizes. Even though the difference between the control and treatment groups is not a concern in SDID design as long as the common trend assumption holds, we adopt Entropy Balancing to adjust the samples before running SDID in the following robustness tests to address the possible confounding factor that people’s housing preferences may change during the sample period. Further, Panel C reveals that the number of properties that have been sold and rented in the tax area increases relatively more than that in the non-tax
area after the tax is implemented. This is consistent with our expectation that implementing vacancy tax leads to higher supply and, thus, puts pressure on the housing prices. Main Test Results Baseline Model Results We present the baseline model results in Tables 2 and 3. In addition to one kilometer, we adopt two different cut-off distances to determine the sample area—500 meters and 1.5 km—to provide more robust test results. Columns (1), (4), and (9) report the results of models without including hedonic variables, and columns (3), (6), and (9) report the results, including additional hedonic variables. In addition, we include property-type fixed effect, year-month fixed effect, and suburb fixed effect and cluster the standard error by year-suburb. The minimum numbers of clustering in each model/column are 321 in the sales sample and 274 in rental sample, which is sufficient to hold the asymptotic assumption. [Table 2 here] According to Table 2, housing sold prices are found to be significantly positively correlated with size, number of bedrooms, number of bathrooms, and number of carparks, but negatively correlated with distance to CBD. These findings are all consistent with previous literature. After including the hedonic variables, our models explain at least 74.4% of the variation in housing prices, and this figure further increases to at least 75.8% after including additional hedonic variables. More importantly, the DID variable TaxPeriod * TaxArea coefficients remain statistically significant and stable at -2.1% to -2.6% when restricting the sample within a distance of 500 meters and 1 kilometer to the boundary and -3% to -3.4% within a distance of 1.5 km. These findings prove that implementing a vacancy tax reduces housing prices by 2.1%– 3.4%. Further, we run the same model to test the rental sample to validate our findings and
present the results in Table 3. We find that the implementing vacancy tax leads to a reduction in rental prices by approximately 2%–3%, and this magnitude is consistent with the sales sample. [Table 3 here] Common Trend Analysis Further, we adopt the dynamic DID model to validate the common trend analysis and testify if the negative impact of a vacancy tax on housing prices is only effective after implementing the tax. We perform the test using Eq. (2) on the transactional and rental samples, respectively. Then, we plot the estimated dynamic DID coefficients with an event study format in Figure 2. [Figure 2 here] According to Figure 2, all the dynamic DID coefficients (-6 to -1) are not statistically different from 0, and they are zero-centering with fluctuation, using both transactional and rental samples. More importantly, we perform an F-test on these pre-trend windows coefficients and validate that these coefficients are jointly statistically insignificant with P-values of 0.819 for the transactionalsample and 0.8054 for the rentalsample, respectively. Thus, the common trend assumption is validated. Further, the dynamic DID coefficients become systematically negative and statistically significant after the tax implementation. These test results reveal that the housing prices in the taxed and non-taxed areas did not take different development paths until the vacancy tax came into effect.
Additional Tests Falsification Tests Further, we perform falsification tests to validate if the significant estimates of the impact of the vacancy tax on housing prices are geographically unique to the taxed area. Therefore, we establish two placebo boundaries by pushing the original boundary of the taxed area forward and backward by two kilometers, respectively, to run two falsification tests. In these two falsification tests, we set the area within a distance of one kilometer of the placebo boundaries as the sample area, the side closer to the Melbourne CBD as the treatment area (FalsAreaIn/FalsAreaOut), and the other side as the control area. The identification of the treatment and control periods is the same as that done for Eq. (1) in the main test. Accordingly, we develop the falsification tests model as below: ln (,) = + × ( × ) + × + × + × ′ , + + + ,, Eq. (3) Table 4 presents the falsification test results, columns (1) and (3) present the results of the test on the placebo boundary that is two kilometers closer to the CBD, and columns (2) and (4) present the results of the test two kilometers further. The coefficients of falsified DID variables FalsAreaIn* TaxPeriod and FalsAreaOut*TaxPeriod are statistically insignificantly different from zero. Therefore, the housing prices on two sides of the placebo boundaries in the nearby area did not show significantly different trends after the vacancy tax came into effect. Therefore, the vacancy tax effects estimated in the previous main tests are geographically unique to the actual taxed area boundary. [Table 4 here]
Robustness Tests We further perform two robustness tests to enhance the credibility of our estimates. First, even if differences between the control and treatment groups are not a concern in the DID design if a common trend assumption held, we still adopt Entropy Balancing to reduce the heterogeneity between the control and treatment groups to enhance the control of confounding factors in the form of a possible change in housing preferences during the sample period. We followed the recommendations in the literature (Hainmueller, 2012) by adopting a maximum entropy reweighting scheme that reweighted the covariates of the treatment and control groups to satisfy a balancing condition. Entropy balancing can adjust for inequalities between treatment and control groups with respect to the first, second, and, possibly, other moments of covariate distribution. It also obviates the continual balance checking and iterative searching required in propensity score matching, which has been criticized in recent literature (King and Nielsen, 2019; Shipman et al., 2017). Table 5 presents the balance of covariates between the treatment and control groups after the entropy balancing adjustments is achieved. Columns (1) and (2) present the test results on the entropy-balanced transactional and rental samples, thereby revealing that implementing the tax leads to a reduction of 2.4%–2.5% in housing prices, which is consistent with the main results. Most importantly, these results prove that the possible change in housing preferences in terms of property characteristics caused by COVID does not affect our main test findings. [Table 6 here] In addition to including the suburb fixed effect in previous main tests, we adopted a multilevel regression model to segregate suburb-level variations from individual property-level variations in the second robustness test. Specifically, we included three more suburb-level attribute control variables: population density, average weekly household income, and the median age
of suburban residents. The test results are presented in columns (3) and (4) of Table 6 and are consistent with the results of the previous main test. Regression Discontinuity Tests Finally, we adopt the spatial Regression Discontinuity (RD) model to test the relationship between the geographical distance to the taxed area boundary and the housing price in the periods before and after implementing the tax. Specifically, we construct variable , which represents the geographic distance between the property and the boundary of taxed area; it is positive within the boundary and negative outside the boundary. Then, we develop the spatial RD model as given below: ln (,) = + × ( × ,) + × , + × + × ′ , + + + ,, Eq. (4) Then, we adopt rdrobust command in Stata to estimate the RD model with sample periods before and after implementing tax separately and present the results in Table 7. The results reveal that RD in the boundary of the taxed area is statistically insignificant before implementing the vacancy tax. This estimate becomes greater in magnitude and statistically significant in the sample period after implementing the tax. We further adopt the parametric approach to estimate and plot the polynomial line in Figure 3, which reveals a cut-off at the boundary after implementing the tax that the housing prices in the taxed area are lower than the non-taxed area. These findings are consistent with the main test results that implementing a vacancy tax leads to a reduction in housing prices. [Table 7 here] [Figure 3 here]
Vacancy Rate Tests To validate our main findings, we conduct additional tests to investigate how implementing the vacancy tax affects the vacancy rate. We collected vacancy rate data on a suburb-month level from CoreLogic via SIRCA. Then, we aggregated transactional and rental data on the suburbmonth level and merge it with the vacancy rate database as well as 2021 census data on suburb level accessed via the Australian Bureau of Statistics. The final database contains vacancy rate, other housing market information such as mean of sold/rental price, socioeconomic information such as population density, and geographic information for each suburb in VIC state in each month from January 2016 to December 2020. Further, we identify the suburbs that are located within the taxed area and exclude the suburbs that are beyond the boundary of the taxed area, and ultimately develop the following DID model to investigate the impact of vacancy tax on vacancy rate: , = + × ( × ) + × + × + × ′, + + + , Eq. (5) where the , is the vacancy rate of suburb j in year-month t, and ′, are a vector of suburb level control variables that represent housing market movements and socioeconomic features of the suburb j in year-month t. Further, we follow the research of Segú (2020) to match the suburbs in the taxed and untaxed areas by adopting a Propensity Score Matching approach to control for the heterogeneity of the address of suburbs before running the DID model. We estimate the propensity score using a Probit regression model that predicts the possibility of locating within the taxed boundary by the suburbs’ attributes, including the mean of sold/rental price, mean of size of properties sold/rent, mean of bedrooms, carpark, and bathroom of properties sold/rent, proportions of houses and units sold/rented, population density, and average household income. We present
the results of the Probit model in Appendix IA.1. We then match properties on two sides of the boundary with a caliper distance of 25% of the standard deviation of the propensity score, and the balancing condition is satisfied in the propensity matched. Thereafter, we rerun the DID model on the matched samples and present the results in Table 8, which reveals that implementing vacancy tax leads to a statistically significant reduction in vacancy rate by approximately 0.2%. [Table 8 here] Conclusion Maintaining a moderate level of housing supply is one of the foremost goals of governments in relation to maintaining housing affordability. However, due to the issues of speculation and investors with multiple house ownership, not all the units developed are supplied in the market. This is more evident in the case of the rental market. It has been known that quite a few investors and owners utilize extra units as summer houses or investment assets without putting the properties on the rental market. This type of phenomena makes housing affordability even worse, as numerous units are left vacant without an efficient use of space assets. In this paper, we investigated the impact of vacancy tax policy in Victoria, Australia. By examining the policy implementation of the vacancy tax scheme, which came into effect in January 2018, as a treatment event, we compared the change in housing price before and after the policy implementation and across different areas depending on the applicability of the policy, as the new tax policy was applicable to only a few regions in the greater Melbourne area. We found that the housing prices in the pertinent area of the tax policy experienced a decline ranging from 2.1%–3.4%. An additional analysis on the vacancy rate confirmed that the policy affects the rental market positively by lowering the vacancy rate approximately by 2%. The results of the Regression Discontinuity analysis were consistent with the spatial DID analysis. The gap
between the sales price and rental rate between the areas affected by the vacancy tax and the areas that are not affected becomes significant after the effective date of the new policy. Finally, we find that the vacancy rate reduces by 0.2% after the vacancy tax policy was implemented in the study area. This study is the first to comprehensively investigate the impact of vacancy tax policy on various indicators of the housing market: housing price, rental rate, and vacancy rate. All the indicators reveal a positive impact of the policy on the housing market. While the practical methods to distinguish between vacant units and occupied units are under debate and require further improvement, the political implication of this paper is clear. Implementing the vacancy tax scheme helps lower the vacancy rate with more affordable price (rent) and supply more rental units in the rental market.
References Autor, D.H., Palmer, C.J., Pathak, P.A., 2014. Housing market spillovers: Evidence from the end of rent control in Cambridge, Massachusetts. J. Polit. Econ. 122, 661–717. https://doi.org/10.1086/675536 Accordino, J., Johnson, G.T., 2000. Addressing the Vacant and Abandoned Property Problem. J. Urban Aff. 22, 301–315. Arsen, D., 1992. Property Tax Assessment Rates and Residential Abandonment: Policy for New York City. Am. J. Econ. Sociol. 51, 361–377. https://doi.org/10.1111/j.1536- 7150.1992.tb03487.x Autor, D.H., Palmer, C.J., Pathak, P.A., 2014. Housing market spillovers: Evidence from the end of rent control in Cambridge, Massachusetts. J. Polit. Econ. 122, 661–717. https://doi.org/10.1086/675536 Branas, C.C., South, E., Kondo, M.C., Hohl, B.C., Bourgois, P., Wiebe, D.J., MacDonald, J.M., 2018. Citywide cluster randomized trial to restore blighted vacant land and its effects on violence, crime, and fear. Proc. Natl. Acad. Sci. U. S. A. 115, 2946–2951. https://doi.org/10.1073/pnas.1718503115 Chen, X., Rafail, P., 2020. Do Housing Vacancies Induce More Crime? A Spatiotemporal Regression Analysis. Crime Delinq. 66, 1579–1605. https://doi.org/10.1177/0011128719854347 Dusansky, R., Ingber, M., Karatjas, N., 1981. The impact of property taxation on housing values and rents. J. Urban Econ. 10, 240–255. https://doi.org/10.1016/0094- 1190(81)90017-6 Fitzgerald, K., 2020. Speculative Vacancies 10 A Persistent Puzzle, The study of Melbourne’s vacant land and housing. Garvin, E., Branas, C., Keddem, S., Sellman, J., Cannuscio, C., 2013. More than just an eyesore: Local insights and solutions on vacant land and urban health. J. Urban Heal. 90, 412–426. https://doi.org/10.1007/s11524-012-9782-7 Han, L., Stacey, D., Chen, H., 2023. Frictional and Speculative Vacancies : The Effects of an Empty Homes Tax. Hartley, D., 2014. The effect of foreclosures on nearby housing prices: Supply or disamenity? Reg. Sci. Urban Econ. 49, 108–117. https://doi.org/10.1016/j.regsciurbeco.2014.09.001 Heinberg, J.D., Oates, W.E., 1970. The Incidence of Differential Property Taxes on Urban Housing: A Comment and Some Further Evidence. Natl. Tax J. 23, 92–98. https://doi.org/10.1086/ntj41861956 Housing Vancouver, 2022. Empty Homes Tax Annual Report, City of Vancouver, Vancouver, British Colombia. Housing Vancouver, 2021. Empty Homes Tax Annual Report, City of Vancouver, Vancouver, British Colombia. Housing Vancouver, 2020. Empty Homes Tax Annual Report, City of Vancouver, Vancouver, British Colombia.
Hu, J., 2018. The Effects of British Columbia’s Vacancy Tax and Foreign-buyer Tax Act on the Supply of New Residential Housing in Vancouver. University of Ottawa. Kee, J.E., Moan, T.A., 1976. The Property Tax and Tenant Equality. Harv. Law Rev. 89, 531–551. Koo, K.M., Liang, J., 2022. A view to die for ? Housing value , wildfire risk , and environmental amenities. J. Environ. Manage. 321, 115940. https://doi.org/10.1016/j.jenvman.2022.115940 Kvik, A., Rose, J., Curriero, F.C., Crifasi, C.K., Pollack, C.E., 2022. The association between vacant housing demolition and safety and health in Baltimore, MD. Prev. Med. (Baltim). 164, 107292. https://doi.org/10.1016/j.ypmed.2022.107292 Mikelbank, B.A., 2008. Spatial Analysis of the Impact of Vacant, Abandoned and Foreclosed Properties. Molloy, R., 2016. Long-term vacant housing in the United States. Reg. Sci. Urban Econ. 59, 118–129. https://doi.org/10.1016/j.regsciurbeco.2016.06.002 Orr, L.L., 1972. The Incidence of Differential Property Taxes on Urban Housing: Reply. Natl. Tax J. 25, 217–220. Orr, L.L., 1968. The Incidence of Differential Property Taxes on Urban Housing. Natl. Tax J. 21, 253–262. https://doi.org/10.1086/ntj41791794 Palmon, O., Smith, B.A., Journal, S., October, N., Palmon, O., Smith, B.A., 2017. New Evidence on Property Tax Capitalization Published by : The University of Chicago Press Stable URL : http://www.jstor.org/stable/10.1086/250041 Confirmations and Contradictions New Evidence on Property Tax Capitalization. 1998 106, 1099–1111. Sadayuki, T., Kanayama, Y., Arimura, T.H., 2020. The externality of vacant houses: The case of toshima municipality, Tokyo, Japan. Rev. Reg. Stud. 50, 260–281. https://doi.org/10.52324/001c.13522 Schachterle, S.E., Bishai, D., Shields, W., Stepnitz, R., Gielen, A.C., 2012. Proximity to vacant buildings is associated with increased fire risk in Baltimore, Maryland, homes. Inj. Prev. 18, 98 LP – 102. https://doi.org/10.1136/injuryprev-2011-040022 Segú, M., 2020. The impact of taxing vacancy on housing markets: Evidence from France. J. Public Econ. 185, 104079. https://doi.org/10.1016/j.jpubeco.2019.104079 Walter, R.J., Acolin, A., Tillyer, M.S., 2024. Association between property investments and crime on commercial and residential streets: Implications for maximizing public safety benefits. SSM - Popul. Heal. 25, 101537. https://doi.org/10.1016/j.ssmph.2023.101537 Wang, J., Zhang, Y., 2019. Vacancy Tax and Housing Price - Difference in Difference Model in Real Estate Market. Simon Fraser University. Wheaton, W.C., 1990. Vacancy, Search, and Prices in a Housing Market Matching Model. J. Polit. Econ. 98, 1270–1292. Whitaker, S., Fitzpatrick IV, T.J., 2013. Deconstructing distressed-property spillovers: The effects of vacant, tax-delinquent, and foreclosed properties in housing submarkets. J. Hous. Econ. 22, 79–91. https://doi.org/10.1016/j.jhe.2013.04.001
Xiao, C., Zhou, B., 2023. Property taxes and rental housing: Evidence from China. Real Estate Econ. 51, 931–958. https://doi.org/10.1111/1540-6229.12420 Zhang, H., McCord, E.S., 2014. A spatial analysis of the impact of housing foreclosures on residential burglary. Appl. Geogr. 54, 27–34. https://doi.org/10.1016/j.apgeog.2014.07.007
Figure 1: Change of vacancy and tenancy in Vancouver, Canada Source: City of Vancouver Note: This figure shows the annual change in vacancy rate and tenancy rate in Vancouver, Canada. 24 25 26 27 28 29 30 0.5 0.7 0.9 1.1 1.3 1.5 2017 2018 2019 2020 2021 Vacancy Rate(Left) Tenancy Rate(Right) (%) (%)
Figure 2: Councils subject to the vacancy tax in the Greater Melbourne area Source: https://www.gadens.com/legal-insights/vacant-property-tax-in-victoria/ Note: This figure shows the location of the councils subject to the vacancy tax in the Greater Melbourne area of Australia. Orange-colored area indicated the councils that are subject to the new vacancy tax.
Figure 3 Common Trend Analysis Note: These figures plot the estimated coefficients of dynamic DIDs coefficients in Eq.2 to validate the common trend assumption using the sold and rental samples respectively.
Figure 4 Regression Discontinuity Plotting Note: These figures plot the Spatial Regression Discontinuity tests with sample periods before and after implementing the vacancy tax.
Table 1 Statistical Summary of Sample Panel A Statistics of key variables-Sold Sample Within 1km Obs Variable Description Mean Price Transactional price in log 13.50 Area Size Size of Land in log 6.30 No. Beds Number of Bedrooms 3.13 No. Baths Number of Bathrooms 1.69 No. Parking Number of carparks 1.92 DistanceCBD Distance of CBD in log -1.74 Panel B Statistics of key variables-Rental Sample Within 1km Obs 1Variable Description Mean Price Transactional price in log 6.05 Area Size Size of Land in log 6.17 No. Beds Number of Bedrooms 2.83 No. Baths Number of Bathrooms 1.48 No. Parking Number of carparks 1.63 DistanceCBD Distance of CBD in log -1.77 Panel C Transactional volume before and after tax Sold Sample Pre Post DiffereInner 23,390 18,780 -4,61Outside 18,974 14,289 -4,68Difference 4,416 4,491 75 (0.3Note: This table reports the statistical summaries of the transactional Melbourne that implements the vacancy tax. The sample period is from 1
73,151 Inside Obs 40,823 Outside Obs 32,328 Outside-Inside SD Mean SD Mean SD Diff p-value 0.42 13.58 0.44 13.41 0.38 -0.16 <0.001 0.85 6.27 0.87 6.34 0.84 0.007 <0.001 1.04 3.09 0.93 3.18 1.16 0.007 <0.001 0.73 1.68 0.77 1.70 0.68 0.009 0.004 0.99 1.93 1.01 1.91 0.96 0.002 0.003 0.30 -0.968 0.244 -0.715 0.272 0.26 <0.001 159,275 Inside Obs 92,228 Outside Obs 67,047 Outside-Inside SD Mean SD Mean SD Diff p-value 0.28 6.07 0.29 6.01 0.27 -0.06 <0.001 0.91 6.12 0.85 6.25 0.98 0.13 <0.001 0.90 2.90 0.86 2.74 0.94 -0.16 <0.001 0.61 1.49 0.62 1.47 0.60 -0.22 0.004 1.01 1.67 1.11 1.59 0.86 -0.08 0.003 0.25 -1.83 0.24 -1.70 0.24 0.13 <0.001 Rental Sample ence Pre Post Difference 10 43,024 49,250 6,226 85 32,448 34,625 2,177 2%) 10,576 14,625 4,049(9.41%) and rental databases within a 1KM distance to the boundary of inner 1st January 2016 to 31st December 2019.
Table 2 Main Test Results-Sold Sample 500 Meters VARIABLES (1) (2) (3) (4) TaxPeriod* TaxArea -0.021** -0.024** -0.024** -0.021*(-1.99) (-2.35) (-2.38) (-2.41)TaxPeriod 0.301*** 0.266*** 0.266*** 0.272**(13.73) (12.86) (13.98) (16.35TaxArea 0.056*** 0.052*** 0.050*** 0.074**(3.02) (3.53) (3.30) (4.13)Area Size 0.013 0.014** (1.61) (2.05) No. Beds 0.120*** 0.111*** (24.46) (23.31) No. Baths 0.077*** 0.074*** (17.36) (15.88) No. Parking 0.037*** 0.033*** (16.21) (15.78) DistanceCBD -0.202** -0.211*** (-2.23) (-2.65) Constant 13.694*** 12.238*** 12.268*** 13.679*(354.28) (57.15) (63.37) (389.26Observations 36,010 29,701 29,701 73,151R-squared 0.623 0.748 0.761 0.633Additional Hedonic YES Property Type FE YES YES YES YES Suburb FE YES YES YES YES YearXMonth FE YES YES YES YES SE Cluster SuburbYear SuburbYear SuburbYear SuburbYNo of Cluster 321 321 321 466
1KM 1.5KM (5) (6) (7) (8) (9) ** -0.026*** -0.026*** -0.030*** -0.034*** -0.033*** ) (-3.11) (-3.19) (-3.90) (-4.94) (-4.79) ** 0.243*** 0.248*** 0.265*** 0.243*** 0.246*** ) (17.06) (17.73) (20.30) (20.76) (21.30) ** 0.064*** 0.063*** 0.075*** 0.065*** 0.063*** ) (3.83) (3.83) (4.28) (3.98) (3.95) 0.019*** 0.018*** 0.013** 0.013*** (2.79) (3.26) (2.31) (2.69) 0.072** 0.065** 0.085*** 0.077*** (2.32) (2.27) (3.12) (3.04) 0.104*** 0.098*** 0.103*** 0.098*** (5.50) (5.58) (6.36) (6.64) 0.046*** 0.041*** 0.044*** 0.039*** (10.50) (11.28) (11.37) (12.12) -0.402*** -0.411*** -0.413*** -0.420*** (-7.29) (-8.11) (-9.74) (-10.42) ** 11.835*** 11.851*** 13.696*** 11.831*** 11.853*** 6) (79.94) (87.15) (409.59) (98.72) (105.91) 1 60,313 60,313 110,969 91,658 91,658 0.744 0.758 0.643 0.757 0.770 YES YES YES YES YES YES YES YES YES YES YES YES YES YES YES YES YES Year SuburbYear SuburbYear SuburbYear SuburbYear SuburbYear 466 466 568 568 568
Note: This table presents the results of the SDID Model (Eq.1) to investigprices. Columns (1) to (3) report the results using a sample within 0.5 KColumns (4) to (6) report the results using samples within 1 KM of the bouincludes property type fixed effect, suburb-fixed effect, month-year fixed statistics in parentheses: *** p<0.01, ** p<0.05, * p<0.1.
gate the impact of implementing a vacancy tax on housing transactional KM of the boundary of inner Melbourne that implements the vacancy tax. undary, and Columns (7) to (9) report results within 1.5 km. Every column effect, and cluster the standard error at the suburb-year level. Robust t
Table 3 Main Test Results-Rental Sample 500 Meters VARIABLES (1) (2) (3) (4) TaxPeriod* TaxArea -0.029*** -0.025*** -0.024*** -0.024**(-2.99) (-3.25) (-3.48) (-2.92)TaxPeriod 0.108*** 0.089*** 0.088*** 0.102**(7.68) (8.64) (9.11) (9.89)TaxArea 0.009 0.004 0.006 0.013(0.63) (0.41) (0.60) (0.79)Area Size -0.012*** -0.014*** (-2.81) (-3.77) No. Beds 0.093*** 0.089*** (6.91) (7.14) No. Baths 0.144*** 0.128*** (25.44) (15.62) No. Parking 0.016** 0.014** (2.56) (2.57) DistanceCBD -0.019 0.008 (-0.23) (0.10) Constant 5.996*** 5.519*** 5.577*** 5.945**(167.12) (30.05) (30.65) (95.45Observations 76,165 63,519 63,519 159,27R-squared 0.451 0.669 0.692 0.461Additional Hedonic YES Property Type FE YES YES YES YES Suburb FE YES YES YES YES YearXMonth FE YES YES YES YES SE Cluster SuburbYear SuburbYear SuburbYear SuburbYNo of Cluster 274 272 272 322
1KM 1.5KM (5) (6) (7) (8) (9) ** -0.025*** -0.024*** -0.012* -0.018*** -0.019*** ) (-3.76) (-3.95) (-1.94) (-3.40) (-4.03) ** 0.084*** 0.087*** 0.103*** 0.083*** 0.086*** ) (10.05) (11.41) (12.93) (11.83) (13.80) -0.005 0.001 0.026** 0.008 0.016** ) (-0.57) (0.08) (2.19) (1.14) (2.20) -0.010*** -0.013*** -0.006 -0.009*** (-2.61) (-4.23) (-1.55) (-2.87) 0.102*** 0.099*** 0.106*** 0.102*** (10.16) (10.44) (13.91) (13.82) 0.147*** 0.134*** 0.147*** 0.134*** (34.06) (24.98) (39.99) (30.72) 0.019*** 0.017*** 0.021*** 0.019*** (3.72) (3.68) (4.89) (4.86) -0.037 -0.012 -0.089** -0.058 (-0.51) (-0.19) (-2.32) (-1.60) ** 5.483*** 5.535*** 5.888*** 5.305*** 5.368*** ) (35.05) (38.95) (127.44) (65.01) (69.30) 5 133,541 133,541 247,414 207,438 207,438 0.698 0.718 0.453 0.698 0.718 YES YES YES YES YES YES YES YES YES YES YES YES YES YES YES YES YES Year SuburbYear SuburbYear SuburbYear SuburbYear SuburbYear 319 319 377 375 375
Note: This table presents the results of the SDID Model (Eq.1) to investigColumns (1) to (3) report the results using a sample within 0.5 KM of the (4) to (6) report the results using samples within 1 KM of the boundary, anproperty type fixed effect, suburb-fixed effect, month-year fixed effect, anin parentheses: *** p<0.01, ** p<0.05, * p<0.1.
gate the impact of implementing a vacancy tax on housing rental prices. boundary of inner Melbourne that implements the vacancy tax. Columns nd Columns (7) to (9) report results within 1.5 km. Every column includes nd cluster the standard error at the suburb-year level. Robust t-statistics
Table 4 Falsification Boundary Tests Sold Sample Rental Sample VARIABLES (1) False Boundary One (2) False Boundary Two (3) False Boundary One (4) False Boundary Two FalsAreaIn* TaxPeriod 0.001 -0.002 (0.08) (-0.23) FalsAreaIn 0.009 0.024*** (0.68) (3.92) FalsAreaOut*TaxPeriod -0.005 -0.002 (-0.70) (-0.41) FalsAreaOut 0.019*** 0.022*** (2.78) (3.89) TaxPeriod 0.195*** 0.212*** 0.068*** 0.085*** (12.80) (15.72) (9.14) (10.02) No. Beds 0.161*** 0.114*** 0.115*** 0.103*** (19.65) (27.95) (14.18) (19.37) No. Baths 0.087*** 0.072*** 0.161*** 0.118*** (16.25) (16.22) (30.78) (29.94) No. Parking 0.060*** 0.033*** 0.030*** 0.022*** (21.30) (20.34) (8.15) (11.26) Constant 12.350*** 12.659*** -0.138*** 0.225*** (86.08) (106.24) (-2.76) (4.54) Observations 65,333 52,047 207,335 157,053 R-squared 0.775 0.733 0.744 0.644 Property Type FE YES YES YES YES Suburb FE YES YES YES YES YearXMonth FE YES YES YES YES SE Cluster SuburbYear SuburbYear SuburbYear SuburbYear Note: This table reports the falsification test results using Eq.4. Column (1) reports the results of moving the boundary 1km towards the Melbourne CBD and column (2) reports the results of moving the boundary 1km further. Robust t-statistics in parentheses: *** p<0.01, ** p<0.05, * p<0.1
Table 5 Balancing Condition Before and After Entropy Balancing Panel A: Sold sample Treated Control Before EB mean variance skewness mean variance skewness No. Beds 3.124 0.8483 0.6515 3.228 1.391 42.26 No. Baths 1.684 0.593 4.346 1.705 0.4559 0.9794 No. Parking 1.93 1.015 1.975 1.908 0.9254 2.057 Treated Control After EB mean variance skewness mean variance skewness No. Beds 3.124 0.8483 0.6515 3.124 0.8065 0.245 No. Baths 1.684 0.593 4.346 1.684 0.4453 0.9294 No. Parking 1.93 1.015 1.975 1.93 1.043 2.15 Panel B: Rental sample Before EB Treated Control Sale sample mean variance skewness mean variance skewness No. Beds 2.900 0.727 0.709 2.747 0.879 0.340 No. Baths 1.501 0.393 1.025 1.481 0.366 1.064 No. Parking 1.669 1.221 36.120 1.588 0.740 1.758 After EB Treated Control Rent sample mean variance skewness mean variance skewness No. Beds 2.900 0.727 0.709 2.899 0.922 0.486 No. Baths 1.501 0.393 1.025 1.501 0.376 1.019 No. Parking 1.669 1.221 36.120 1.669 0.824 1.875 Note: This table reports the balancing conditions of covariates before and after Entropy Balancing.
Table 6 Robustness Tests Entropy-balanced sample Multilevel VARIABLES (1) Sold Sample (2) Rent Sample (3) Sold Sample (4) Rent Sample TaxPeriod*TaxArea -0.024*** -0.025*** -0.025** -0.025*** (-7.25) (-15.73) (-2.00) (-3.13) TaxPeriod 0.235*** 0.080*** -0.028*** 0.043*** (21.17) (21.10) (-3.05) (5.88) TaxArea 0.061*** -0.004* 0.065** -0.004 (11.35) (-1.70) (2.52) (-0.24) No. Beds 0.132*** 0.100*** 0.075** 0.103*** (55.39) (57.28) (2.38) (5.77) No. Baths 0.076*** 0.146*** 0.104*** 0.147*** (17.53) (131.95) (5.14) (24.41) No. Parking 0.040*** 0.019*** 0.048*** 0.019*** (34.70) (6.57) (8.77) (3.68) DistanceCBD -0.389*** -0.038*** -0.341*** -0.051 (-18.82) (-3.29) (-3.94) (-0.54) Median_Age 0.028*** 0.020*** (4.40) (3.66) Population Density -0.017 -0.038 (-0.63) (-1.63) Median Income 0.349*** -0.049 (3.20) (-1.50) Constant 11.836*** 5.386*** 9.107*** 5.093*** (216.51) (190.52) (12.45) (19.42) Observations 61,948 153,951 60,313 153,082 R-squared 0.762 0.693 Property Type FE YES YES YES YES Suburb FE YES YES YES YES YearXMonth FE YES YES YES YES Number of Suburbs 118 79 SE Cluster Suburb Suburb Prob > chi2 0.000 0.000 var(_cons) 0.043 0.015 var(Residual) 0.045 0.023 Note: This table reports additional test results using an Entropy-Balanced sample and a multilevel regression model. T-statistics in parentheses *** p<0.01, ** p<0.05, * p<0.1.
Table 7 Spatial Regression Discontinuity Test Results Sold Sample Rental Sample VARIABLES Pre_Tax Post_Tax Pre_Tax Post_Tax RD_Estimate -0.004 -0.0225*** 0.0003 -0.036*** (-1.08) (-6.94) (0.024) (-10.178) Observations 32,521 27,792 74,513 83,862 Property Type FE YES YES YES YES Month FE YES YES YES YES Year FE YES YES YES YES SE Cluster SuburbYear SuburbYear SuburbYear SuburbYear Note: This table reports the Spatial Regression Discontinuity tests results using Eq.4 using the sold and rental samples in periods before and after implementing vacancy tax. Z-statistics in parentheses *** p<0.01, ** p<0.05, * p<0.1
Table 8.Impact of vacancy tax on vacancy rate VARIABLES Model(1) Model(2) Model(3) TaxAreaXTaxPeriod -0.002** -0.002** -0.002*** (-2.00) (-2.02) (-2.61) TaxArea 0.004*** 0.004*** (7.37) (7.19) TaxPeriod -0.008*** -0.008*** -0.007*** (-3.93) (-4.32) (-3.74) Constant 0.027*** 0.099*** 0.036*** (20.25) (11.24) (6.95) Observations 1,804 1,804 1,804 R-squared 0.193 0.271 0.200 YearXMonth FE YES YES YES Controls YES YES Number of Suburbs 122 Suburb FE YES Note: This table reports the DID tests results to investigate the impact of implementing vacancy tax on vacant rate using Eq.5 on the propensity matched suburbs sample with different model settings. Z-statistics in parentheses *** p<0.01, ** p<0.05, * p<0.1
Table Appendix IA.1 This table reports results from a probit estimation which is used to calculate the propensity score of locating within the taxed area based on the suburbs’ attributes for matching. The dependent variable is a dummy variable that equals one if the suburb is located within the taxed area. The explanatory variables are suburbs’ attributes. T-statistics are reported in parentheses. *p<0.10, **p<0.05, ***p<0.01. VARIABLES Population Density 2.517*** (26.71) Household Income -0.355 (-1.54) Mean of sold price 0.000*** (24.96) Mean of sold properties’ area size -0.000 (-1.14) Mean of bedrooms number sold -1.699*** (-8.84) Mean of bathrooms number sold -0.103 (-0.48) Mean of parking number sold -0.310*** (-2.63) Proportion of houses sold -1.910*** (-3.73) Proportion of units sold -0.333 (-0.59) Mean of rental price 0.003*** (2.77) Mean of rented properties’ area size 0.000 (1.33) Mean of bedrooms number rented -0.575** (-2.31) Mean of bathrooms number rented -0.759** (-2.55) Mean of parking number rented -0.834*** (-4.19) Proportion of houses rented 0.093 (0.20) Proportion of units rented -3.824*** (-7.22) Constant -12.634*** (-7.18) Prob > chi2 0.000 Pseudo R2 0.7458 Area under ROC 0.9817 Observations 15,991