Liquidity constraints and speculation: Evidence from the housing-resale restriction policy in Qingdao, China

This article seeks to explore the efficacy of the Housing-Resale Restriction (HRR) policy in curtailing housing speculation, complementing existing literature that primarily examines the impact of taxation. The HRR policy directly reduces speculative opportunities through heightened liquidity constraints. Employing the methodology of regression discontinuity in time (RDiT), this study analyzes the impact of liquidity constraints on housing price growth rates based on the monthly community-level data from Qingdao, China, where the HRR policy was the most stringent in terms of scope and holding time requirement. The results show that the HRR policy reduced housing price growth rates by 1.3881 to 2.1001% points on average. Moreover, the disincentive impact was less pronounced in affluent, recently developed communities and those near subway stations. These findings provide valuable insights for policymakers to regulate the real estate market, mitigate excessive speculation, and prevent market bubbles.


I. Introduction
Speculation is a pervasive force in real estate markets around the world.Stabilizing real estate market and improving housing affordability are crucial for macroeconomic stability and the well-being of individuals.Unlike genuine investment, speculation often stems from the potential for financial gain through short-term property resales (Lan, Moreira, and Zhao 2019).Globally, various policy instruments have been implemented based on unique socio-economic contexts to raise the holding costs or thresholds, thereby mitigating speculation.
A multitude of nations and territories have enacted regulatory measures in the real estate markets to prevent real estate asset bubbles, particularly as a response to the 2007 subprime mortgage crisis (Huang and Yang 2021).For instance, during the period of escalating housing prices from 2006 to 2008, Germany implemented substantial taxation on the resale of housing units (Muller, Almy, and Engelschalk 2010).In a similar vein, the Singaporean authorities adjusted their property market regulations by increasing stamp duties for both buyers and sellers from 2009 to 2013 (Deng, Gyourko, and Li 2019).The Hong Kong SAR has been proactive in implementing counter-cyclical measures, notably through the enhanced transaction taxes (stamp duties) and the reinforcement of loan-to-value ratios since 2009 (Wong et al. 2021).Estonia curtailed market speculation through the introduction of a land tax in 1993 (Cocconcelli and Medda 2013).Furthermore, certain jurisdictions have specifically targeted foreign investments to mitigate housing market speculation.Toronto imposed a 15% non-resident speculation tax on foreign buyers in April 2017 (Du, Yin, and Zhang 2022).In New Zealand, the Overseas Investment Amendment Act was passed in 2018, precluding foreigners from purchasing resale housing (Lu, Zhang, and Hong 2021).Similarly, Poland regulated foreign speculation with the enactment of specific law (Marks-Bielska 2013).While these interventions predominantly employ fiscal instruments and establish legal thresholds to curb speculation, this study delves into the impact of directly constraining property liquidity on speculation in the real estate market.
The rapid economic ascent and distinct housing market dynamics in China present a unique context for examining the effectiveness of antispeculation policies.Since the market-oriented reforms of the 1980s, China has witnessed significant growth in the owner-occupied housing sector.The growing demand for homeownership, limited individual investment options and intense capital investment intentions have fuelled a boom in the real estate market, resulting in declining housing affordability in urban China, especially in large cities (Chen, Hu, and Lin 2019;Chen and Wang 2020;Chen and Wen 2017;Fang et al. 2016;Wu 2015).To control excessive housing price growth, various cities in China have introduced diverse policy interventions to curb speculative purchases and prioritize the immediate living needs, including Housing-Purchase Restrictions (HPR)1 and augmented mortgage conditions.
While the HPR policy has been widely discussed, its effectiveness remains debated.Some scholars argued that the HPR policy significantly suppressed speculative demand (Li, Cheng, and Cheong 2017;Lu, Zhang, and Hong 2021;Zhang and Wang 2016), while others believed it had no significant effect on housing price suppression (Li et al. 2020;Zhang and Wang 2016).However, some studies have shown that the HPR policy has led to a decline in housing prices (Sun et al. 2017;Wu and Li 2018).Furthermore, practical loopholes, such as sham divorces or property registrations under children's names, have raised scepticism about the efficacy of the HPR policy in restraining speculation.
In order to further fortify anti-speculation efforts and enhance housing affordability, some cities have imposed stricter restrictions on housing liquidity, such as the Housing-Resale Restriction (HRR) policy, which mandates a defined holding period before resale.Compared to the HPR policy, the HRR policy recalibrates homebuyers' expectations and increases their holding costs.Despite its significance, research on the HRR policy remains limited.Yan and Hongbing (2018) found that the HRR policy reduced housing prices by suppressing speculative demand in the short run, using monthly data from 43 Chinese cities. Lan et al, (2019) found that the HRR policy significantly decreased housing prices and trading volume in the second-hand housing market based on a quasinatural experiment in Suzhou.However, due to the characteristics of the housing market, where prices typically rise rapidly and fall slowly, or simply experience a deceleration in growth rate, using housing prices as the dependent variable may not fully capture the policy's effects.Furthermore, the lack in examining the HRR policy's impacts across diverse housing submarkets, which vary in types, quality, and access to amenities, limits our understanding of its heterogeneous effects on housing markets.
In this study, using the case of second-hand housing market in Qingdao, we examined the impact of the HRR policy on housing price growth rates with detailed monthly community-level data.Given Qingdao's implementation of the most stringent restrictions in terms of scope and holding time requirement, it presents an ideal case study for this study.This article makes several notable contributions to the literature on housing regulatory policies.Firstly, existing studies have primarily focused on the effects of the HPR policy on the property market, with limited evidence on the effects of the HRR policy (Zhang et al. 2021;Zheng, Chen, and Yuan 2021).While the HPR policy affects housing prices and transaction volume by regulating the overall demand of homebuyers in a city, the HRR policy aims to curb speculative demand by increasing liquidity constraints, subsequently influencing housing prices and transaction volume.
Second, this study identifies the causal relationship between the HRR policy and the housing price growth rate.We argue that using the housing price growth rate helps avoid potential biases in policy evaluation.After the implementation the HRR policy, we observe a trend of slow decline in housing price, compared to the rapid rise before the policy (as shown in Figure 1).Therefore, we believe that the growth rate of housing prices is a more sensitive and accurate reflection of the policy impact on the housing market than the absolute housing price level.
Third, this study employs monthly data at the community level to assess the effect of the HRR policy.Unlike average housing prices at the citylevel, detailed information on community-level prices and property characteristics enables a more comprehensive examination of the heterogeneous effects of the HRR policy on properties across various types, submarkets and locations.
The remainder of this article is organized as follows.Section 2 outlines the policy background and theoretical framework; Section 3 details the study area, data, and identification strategy; Section 4 presents the empirical results; Section 5 and 6 discuss and conclude.

Housing restriction policies in urban China
In 1998, China transitioned from a state-led inkind housing allocation system to a marketdriven housing supply model, elevating the real estate industry to a crucial pillar of the economy.This shift spurred a period of vigorous real estate market expansion, characterized by rapid surges in housing prices and volumes.The 2008 subprime mortgage crisis served as a significant turning point.In response, China initiated the 'Four Trillion Stimulus Plan' to stimulate the economy, resulting in GDP growth of 9.6% in 2008 and 9.2% in 2009, accompanied by a national average housing price growth rate of 23.3% in 2009.Given these inflationary trends, Beijing and other cities introduced the HPR policy in 2010 to temper the excessive growth of housing prices.Post-2014, the initial wave of purchase restrictions successfully moderated the rapid escalation of housing prices in most Chinese cities, except for super-star cities like Beijing, Shanghai, Guangzhou and Shenzhen.However, this period also witnessed a swift accumulation of unsold housing inventory, particularly in third-and fourth-tier cities, as many residents postponed home purchases amid the slowed price growth (Zhang, Hui, and Wen 2015;Yang et al. 2017).
In order to tackle the mounting housing inventory and foster the sustainable development of the real estate industry, China's Central Urban Work Conference proposed a property destocking strategy in December 2015.Following this, governments at various levels rolled out a series of destocking policies, leading to a quick boost in transaction volumes and prices.In December 2016, the Central Economic Work Conference asserted for the first time that 'housing is for living, not for speculation', highlighting the essential role of residential property for habitation rather than speculative investment.Consequently, many local governments have initiated market cooling measures since 2017, including restrictions on purchases, sales, mortgages, and prices.
Among these measures, the HRR policy stipulates that buyers must hold the housing for a certain number of years before they can resell it.By restricting housing liquidity and increasing holding costs, this policy offers a more targeted approach of curbing speculative housing purchases compared to the more widely implemented HPR policy.The implementation and enforcement of the HRR policy vary across cities in aspects such as initiation time, restricted area, targeted housing type and minimum holding period.
In March 2017, the HPR policy was introduced in Qingdao along with a two-year HRR policy.However, the persistent upward trend in housing prices remained uncontrolled, as depicted in Figure 1 2 The possible reason is that the two-year holding period requirement aligned with the requirement for value-added tax (VAT) exemption.Since resale within two years would incur a 5% VAT on the sale price of the housing, which becomes exempt after two years, speculators rarely resell within two years.In response, the holding period requirement was extended to five years in April 2018.This revised regulation specified that housing purchased post-18 April 2018, could only be resold after being holding for five years.As indicated in Figure 1, the enforcement of the fiveyear HRR policy had triggered a downward trend in housing prices in Qingdao.Among the cities that implemented the HRR policy, Qingdao had the most extensive coverage (both new and secondhand housing in the city) and the longest holding time requirement (5 years).

Theoretical framework
While the existing literature usually employs theoretical models to examine the impact of changes in liquidity constraints related to macroeconomic policies (e.g.monetary policy, loan policy) on real estate prices (Chen and Leung 2008;Iacoviello 2005;Kiyotaki and Moore 1997;Kwan, Leung, and Dong 2015), this study concentrates on the impact of HRR policy that directly increased restrictions on housing asset liquidity.The HRR policy mandates a specified holding period postpurchase before homebuyers can resell, directly increasing the holding costs for homebuyers for speculators.To analyse the policy effect, we develop a utility function of homebuyers, as shown in Equation ( 1).This function illustrates how rising liquidity constraints (specifically, the minimum holding duration before resale) may impact homebuyers' utility.
where: U i represents the homebuyer's utility; H i denotes the housing price and housing-related equity, such as access to amenities; C signifies the cost of maintaining housing ownership, which is a nonlinear function of loans (l i ), property taxes (f i ) , maintenance costs (m i ), and holding period (T).l i is a dummy variable that assumes a value of 1 when housing mortgages are utilized, and 0 otherwise.t represents the proportion of fees and taxes paid during the housing purchase.eH i encapsulates the homebuyer's expectations of housing prices when holding the property H i , with a positive value indicating a bullish forecast, and a negative value suggesting a bearish one.
The homebuyer's budget constraint function is formulated as follows.
where: Y i represents the homebuyer's budget, encompassing the house's purchase price (including taxes), holding costs (e.g.mortgage cost, maintenance cost), and other miscellaneous expenses.p l is the mortgage costs, and p r is the interest costs of own funds (including interest or other investment benefits).P i denotes the price of other composite goods X i .Adhering to the principle of utility maximization, a Lagrangian function is constructed based on Equations ( 1) and (2).Then we obtain the quantities of properties and other goods (H * , X * ) that maximize the homebuyers' utility.
The HRR policy, which necessitates homebuyers to retain their properties for a specified period (e.g.five years in Qingdao) before they can resell it, affects homebuyers' utility in two ways.Firstly, as illustrated in Equation ( 1), the reduction in housing asset liquidity correlates with an increase in mortgage costs (p l ), interest costs (p r ), property taxes (f i ), and maintenance costs (m i ) over time (T).The rise in holding costs adversely affects homebuyers' utility.Secondly, by extending the holding period to five years, the HRR policy bolsters the government's efforts to rein in housing prices, thereby instilling a negative expectation (eH i ) about future housing prices, which further diminishes homebuyers' utility.The confluence of escalating holding costs and declining house pricing expectations renders the transaction less lucrative in the short term, thus curbing speculative demand and tempering housing price growth.

Study area
Qingdao, nestled in the southeast of Shandong Province, is an important coastal city and an international port city (Zhang and Rasiah 2013).By the end of 2021, Qingdao has a permanent population of 10.26 million (ranking 16th among more than 300 Chinese cities), and its Gross Domestic Product (GDP) amounted to CNY 1,413.6 billion in 2021, of which the tertiary industry contributed 60.8%. 3 Owning to its scenic surroundings and coastline, Qingdao, often dubbed the 'Oriental Switzerland', is renowned for its historical and natural resources.This has attracted a large influx of speculators and investors to its real estate market, leading to a significant surge in property prices in recent years.The average housing price has escalated from CNY 2,297 to CNY 14,440 per square metre over the past two decades, reflecting an average annual growth rate of 10%. 4

Data sources
This article leverages the housing market data of Qingdao sourced from Anjuke (qd.zu.anjuke.com),one of the premier housing information platforms in China where continuous monthly housing price data at the community level is available (Chen et al. 2016).The housing price data collected spans from July 2017 to January 2019.Considering policy lags and other temporal factors, May 2018 is designated as the cut-off point (given that the policy implementation date is 18 April 2018), and data from 9 months prior to and following the policy introduction are included.The growth rates of housing prices at the community level were further calculated.Due to the significant problem of missing variable data for suburban communities, we narrowed our study area to six central districts of Qingdao, namely Shinan District, Shibei District, Huangdao District, Laoshan District, Licang District and Chengyang District.This selection allowed us to construct a balanced panel sample.In addition, the proximity of each community to local amenities (e.g.schools, subway stations, and landscapes) was gauged using Amap (www.amap.com).Table 1 presents the descriptive statistics of key variables.
Figure 3  Additionally, to circumvent overestimating the policy effects, robustness tests were performed using samples from all communities across Qingdao and the aggregated housing prices at the district/county level, respectively.Moreover, we collected data of transaction prices and volumes from Lianjia (www.lianjia.com), another leading housing search and information platform in China.It is noteworthy that not every housing unit is traded every month, thus the transaction prices at the unit level cannot constitute balanced panel data.

Identification strategy
The regression discontinuity (RD) estimation is widely employed to evaluate policy effects.The RD method utilizes samples before and after a cutoff point, treating the former as the control group and the latter as the treatment group, to determine the local average treatment effect (LATE) at the cut-off point.Cross-sectional data is often used in RD to mitigate the effects of other interventions and minimize potential biases.For instance, Thistlethwaite and Campbell (1960) employed the RD to assess the impact of bonuses on students' academic performance using grades as the cut-off point.Chen et al. (2019b) used housing area as the cut-off point to estimate the value of hukou (household registration).However, when time is used as the cut-off point, panel data should be employed instead of time series data for estimation (Hausman and Rapson 2017).Anderson (2014) used the regression discontinuity in time (RDiT) to evaluate the impact of transportation strikes.Based on Anderson (2014), this study constructs the RDiT using the monthly community-level panel data of Qingdao to test the LATE of the HRR policy.The linear and linear interaction models are constructed as follows: where: R ijt represents the outcome variable, namely the housing price growth rate of community i in county j in month t (the percent sign is omitted).D it is the policy processing variable, and its coefficient represents the LATE of the HRR policy, also referred to as the RD treatment effect of the policy.As mentioned above, considering the lag of the policy effect, May 2018 is taken as the cut-off point.Months are recorded starting from August 2017, with August 2017 being represented as 1 and subsequent months incrementing by one.The cut-off point in May 2018 is designated as 10.The running variable T it is calculated as Month-10.Before conducting the RDiT, linear and quadratic fittings were performed to examine whether the growth rate of housing prices jumped before and after the implementation of the HRR policy.The results depicted in Figure 4 demonstrate a significant downward jump in the housing price growth rate at the cut-off point (May 2018).Post the HRR policy, the growth rate of housing prices in Qingdao initially decelerated and then transitioned to a negative trajectory.

Baseline results
The results of the baseline model for parameter estimation are reported in Table 2.A bandwidth (h) of 9 was chosen, encompassing samples from 9 months before and after the HRR policy.Standard errors were clustered at the community level.The parameter estimates for the linear model are displayed in Columns (1) and (2).Without covariates, the coefficient on D is −1.3881, indicating a 1.3881% point reduction in housing price growth rate due to the HRR policy.Controlling for covariates, the result in Column (2) indicates a reduction in the housing price growth rate by 1.5697% points, consistent with the result without covariates.This consistency suggests a robust effect of the HRR policy in dampening housing price growth.Columns (3) and ( 4) report the estimates for the linear interaction models.The RD treatment effects of the HRR policy are observed to be −2.0553 and −2.1001, respectively, further strengthening the conclusion that the HRR policy effectively inhibited the growth of housing prices.These results indicate that the HRR policy reduced speculation and thus decreased housing price growth by imposing the liquidity constraints on housing resale.

Robustness tests
We conducted robustness tests by varying the bandwidth, estimation method, and sample.Firstly, the flexible RD parameter estimation model allows for an adjustable bandwidth (h).We tested the robustness of the baseline models by choosing two bandwidths, h = 3 and h = 5, respectively, and performing regressions based on Equations ( 3) and (4).The regression results for the linear model in Column (1) and the linear interaction model in Column (3) of Table 3 show that when the bandwidth (h) is set to 3, the RD treatment effects are −2.3923 and −4.1018 respectively.When the bandwidth (h) is 5, the RD treatment effects in Columns ( 2) and ( 4) are −1.6737 and −2.6593, respectively.This consistency across models confirms a steady inhibitory effect of the HRR policy on housing price growth.Secondly, non-parametric RD methods were employed to verify the robustness of the baseline model results.Compared to parametric estimation, non-parametric estimation does not require specifying the model form.In Table 4, the nonparametric RD estimation in Column (1) uses the method proposed by Imbens and Lemieux (2008), while Columns (2) and (3) follow the method proposed by Calonico et al. (2014) and Calonico et al. (2017).The results confirm that the HRR policy significantly dampened the growth of housing prices.
Thirdly, to mitigate potential overestimation of the policy's effect, we expanded the sample to cover the entire city of Qingdao, as the HRR policy applied to all ten districts/counties.In Table 5, Columns (1) and (2) present results using data from all communities in Qingdao, while Columns (3) and (4) show results using aggregated housing price statistics at the district/county level.
Covariates were not included in these models due to significant missing characteristic information in the full sample and the absence of monthly characteristics at the district/county level.To reduce bias arising from the functional form, a nonparametric RD estimation method was employed in these robustness tests.The results in Table 5 corroborate the baseline findings, showing a significant impact of the HRR policy on reducing housing price growth rates.These tests suggest that the baseline regression results do not overestimate the policy effects.In addition, to provide a more intuitive understanding of the policy impact across all communities and at the district/county level, we plotted linear and quadratic fittings of housing price growth rates, as shown in Appendix Figures A1 and A2, respectively.Fourthly, we conducted a placebo test using unrestricted housing types and taking the nonpolicy implementation time as the cut-off point.Restriction policies vary across different housing types (Somerville, Wang, and Yang 2020).In the case of Qingdao, the HRR policy specifically applied to commodity housing and excluded other housing types, such as serviced apartments and commercial properties.As shown in Columns (1) and (2) of Table 6, the HRR policy did not significantly affect the price growth rate of unrestricted building types.This finding confirms that the impact of the HRR policy was limited to restricted housing types, passing the placebo test.
To rule out the effect of seasonal trends, we performed placebo tests using January 2018 and September 2018 as alternative cut-off points, respectively.To ensure independence from the function form, we employed a non-parametric estimation.As reported in Columns (3) and (4), the results show no significant impact on housing price growth rates, further confirming that the baseline regression results are not influenced by seasonal fluctuations.

Heterogeneity analysis
The characteristics of a community, such as its quality, age, and location, significantly affect market demand and housing price growth.In this subsection, we examined the heterogeneous effects of the HRR policy based on differences in price level, housing age, and access to amenities.Firstly, we included the interaction term, D×HP, in the model.The significantly positive coefficient on the interaction term in Column (1) of Table 7 indicates that more expensive communities were less affected by the policy, implying that the HRR policy had less impact on high-quality properties.
The age of a property is another crucial factor affecting housing demand, as it often correlates with quality and remaining tenure.5 Newer properties are likely to be better furnished and equipped.To examine whether the HRR policy's impact varies with age, the interaction term, D×AGE, was added into the model.The significantly negative coefficient of the interaction term in Column (2) implies that the policy had a stronger dampening effect on older properties, suggesting that older  (1) Significance at the 10%, 5%, and 1% levels is indicated by *, **, and ***, respectively.(2) Standard errors are presented in parentheses.
(3) The results are derived through non-parametric estimation using a triangular kernel.
properties might have difficulty maintaining their value and were more affected by the HRR policy.
Location and traffic accessibility are key determinants in the housing market.We categorized communities based on their proximities to subway stations within a 1-kilometre radius and introduced the interaction term, D×SUBWAY, into the model.The result in Column (3) shows that the growth rate of housing prices near the subway was less inhibited after the policy implementation.This is because housing near the subway attracts more rigid demand, while the HRR policy primarily targets speculative demand.This finding suggests that housing in desirable locations maintained its value despite restrictive policies.

Further analysis
Drawing on Leung and Teo (2011), we studied the spatial spillover effect of the HRR policy.Specifically, we investigated whether speculators redirected their investments to neighbouring cities after the implementation of the HRR policy in Qingdao.If speculators migrated to other cities, it could potentially result in a significant increase in housing prices in those cities.By analysing several cities near Qingdao, namely Weifang, Yantai and Rizhao, using nonparametric RD estimation at the district/county level, the results in Table 8 indicate that the HRR policy implemented in Qingdao did not significantly impact the neighbouring cities.The choice of non-parametric RD estimation was driven by its ability to yield more accurate results without the need for controlling numerous covariates as required in parametric RD benchmark estimation.This methodological decision was further supported by the lack of monthly variables at the county/district and city levels.The linear and quadratic fittings in Appendix Figure A3 also show no significant jump in housing price growth rate at the cutoff point.These results underscore the localized impact of the HRR policy.
As mentioned earlier, the changes in housing price levels before and after policy implementation may be less pronounced than the changes in the housing price growth rates, but analysing the changes in price levels can still provide supporting evidence for evaluating the policy.In addition, to address the concern that using  the listing price to proxy the transaction price may lead to some estimation bias, we collected historical micro transaction data from Lianjia. 6he results in Columns ( 1) and (2) of Table 9 indicate an average decrease in transaction prices ranging from 4.88% to 6.54% due to the HRR policy.This is further supported by linear and quadratic fittings of the logarithm of transaction prices shown in Appendix Figure A4 These results robustly indicate that the HRR policy significantly lowered transaction prices, aligning with our earlier findings of reduced housing price growth rates.
In addition, we acknowledge the discrepancy between listing and transaction prices, so we used the listing price from the same source to reestimate and compare whether there is a bias.The results in Columns (3) and (4) as well as fittings in Appendix Figure A5 based on unit-level listing prices also confirm the substantial dampening effect of the HRR policy.Although the listing price decline is more noticeable, the significance of the transaction price decline remains.This suggests consistent conclusions can be drawn using both listing and transaction prices, justifies using listing prices in the baseline model for a larger sample size, and reinforces the conclusion that the HRR policy had a marked effect on moderating housing price dynamics in Qingdao.
Furthermore, the HRR policy, by limiting the liquidity of housing, could concurrently lead to a decrease in transaction volumes.As displayed in Table 10, transaction volumes have indeed shown a decline at the community, district/county, and city levels following the implementation of the policy.This evidence further validates the efficacy of the HRR policy in tempering the housing market from another perspective.

V. Discussion
The concept of speculation opportunity, as defined by Lan et al. (2019), refers to the potential for property buyers to profit from short-term resale.Recent empirical studies have extensively documented the extent to which transactions in the housing market are motivated by buying and selling for short-term gains, and how these activities correlate with the housing price cycle (Leung and Ng 2019).Leung and Tse (2017) introduced the concept of arbitraging middlemen, or 'flippers' −  investors aiming to profit from buying low and selling high − into a standard housing market search model.Their findings emphasized that speculation is highly dependent on the funding costs of speculators.Consequently, they suggested that intervening in the funding market of speculators would be a targeted and effective policy approach.The empirical study in this article examines another policy instrument to curb speculation in the housing market.The implementation of the HRR policy in Qingdao took place during a period of skyrocketing housing prices, which approximately doubled between July 2016 and April 2018 (as depicted in Figure 1).In this context, the primary objective of the local government was to stabilize housing prices and prevent a real estate market bubble that could precipitate financial risks.Given that the predominant surge in housing prices during this period was fuelled by speculative demand which was less affected by other policy constraints, the HRR policy, as a direct liquidity constraint, proved to be an effective intervention at that time.Our research supports that the HRR policy significantly curbed excessive growth in housing prices, but the impact was diminished for higher-quality, newer, and prime location housing.The results indicate no significant spillover effects on neighbouring cities, suggesting the localized impact of the HRR policy.
Furthermore, it is crucial for governments to exercise caution when managing expectations within the private sector, as indicated by studies such as Chen et al. (2015) and Zhou (2016Zhou ( , 2018)).These studies underscored the potential for over-reaction or under-reaction to policies, emphasizing the importance of clear communication and a comprehensive understanding of policy goals and mechanisms.It is also worth noting that urgency liquidity needs (e.g.job relocation, large medical expenses) may render restrictions on housing liquidity unreasonable.While speculation can drive up housing prices, it also contributes to market liquidity, which is vital in an asset-based welfare nation like China for urgent liquidity needs (Leung and Zhang 2011;Ronald and Doling 2012).Therefore, as argued by Leung and Tse (2017), without a framework to properly assess optimal housing market liquidity, there is a risk that government interventions may either overreach or fall short in stabilizing housing prices.

VI. Conclusion
Speculation plays a significant role in the rapid escalation of housing prices.Speculators engage in the purchase and subsequent resale of properties to reap profits.This study aims to investigate whether restricting housing liquidity can effectively curb speculative purchases and mitigate housing price growth.Using monthly community-level balanced panel data, this article identifies the causal relationship between liquidity constraints and housing price growth rates based on the implementation of the HRR policy in Qingdao, China.The parametric and non-parametric RD estimations consistently demonstrate that the HRR policy reduced housing price growth rate by an average of 1.3881 to 2.1001% points, significantly curbing excessive growth in housing prices.We address a crucial gap in existing research by examining the effects of the HRR policy across various housing submarkets using micro-level data.Our analysis includes different aspects of housing such as quality, age, and access to amenities.The heterogeneity analysis reveals that housing in higher-quality, newer, and superior locations exhibited lower susceptibility to the HRR policy.Further analysis based on transaction prices and volumes further confirms that the HRR policy had a cooling effect on the housing market.However, the policy did not present a significant spillover effect on the housing market in neighbouring cities.
This study demonstrates that when housing prices experience a sharp increase, the implementation of HRR policy (increasing housing liquidity constraints) can effectively stabilize housing prices.The findings provide evidence for policymakers to develop strategies aimed at curbing speculative demand by increasing holding costs, influencing homebuyers' expectations, and regulating liquidity constraints.However, it is important to consider potential negative consequences of such policies, as stringent constraints may overlook urgent liquidity needs associated with unexpected large expenses.Determining the optimal level of governmental interventions and housing market liquidity is a critical area of future research.Moreover, the mechanism of this policy warrants further exploration, particularly in terms of accurately measuring speculative demand.This nuanced approach will be crucial in achieving a balanced and effective housing market policy.

Disclosure statement
No potential conflict of interest was reported by the author(s).
Figure 2 depicts residential investment and fixed asset investment in Qingdao from 2008 to 2017.The share of residential investment in total fixed assets and non-productive fixed assets has been declining since 2008, indicating a diversification in Qingdao's industrial development beyond real estate investment post the subprime mortgage crisis.
displays the sample distribution and housing prices transformed into a continuous grid using Kernel Density Estimation in ArcGIS.Part (a) displays the spatial distribution of the sample communities in central districts, clearly indicating that residential communities are clustered on the flat terrain.Part (b) exhibits the housing prices of the communities as of May 2018.Parts (c) and (d) show the growth rates of housing prices as of January 2018 and September 2018, respectively.A comparison of Parts (c) and (d) reveals that the growth rate of housing prices decelerated after implementing the HRR policy, with a widescale negative growth phenomenon, as indicated by the cooler colour in Part (d).

Figure 3 .
Figure 3.The spatial distribution of community samples.(a) Sampled communities in Qingdao; (b) Housing price in May 2018; (c) Growth rate of housing price in January 2018; (d) Growth rate of housing price in September 2018.

Figure 4 .
Figure 4.The jump in housing price growth rate at the cut-off point.The confidence intervals for the fitted plots are 99%.

Figure A5 .
Figure A5.The jump in housing listing price at the cut-off point.

Figure A4 .
Figure A4.The jump in housing transaction price at the cut-off point.

Table 1 .
Definition and summary statistics of key variables.

Table 3 .
Robustness tests: changing the bandwidth.

Table 8 .
The spatial spillover effect of the HRR policy.

Table 9 .
Impact of the HRR policy on housing price level.

Table 10 .
Impact of the HRR policy on transaction volume.