The impact of exogenous demand shock on the housing market: Evidence from the home purchase restriction policy in the People's Republic of China

Cao, Xiaping; Huang, Bihong; Lai, Rose Neng Working Paper The impact of exogenous demand shock on the housing market: Evidence from the home purchase restriction policy in the People's Republic of China ADBI Working Paper, No. 824 Provided in Cooperation with: Asian Development Bank Institute (ADBI), Tokyo Suggested Citation: Cao, Xiaping; Huang, Bihong; Lai, Rose Neng (2018) : The impact of exogenous demand shock on the housing market: Evidence from the home purchase restriction policy in the People's Republic of China, ADBI Working Paper, No. 824, Asian Development Bank Institute (ADBI), Tokyo This Version is available at: https://hdl.handle.net/10419/190245 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc-nd/3.0/igo/ ADBI Working Paper Series THE IMPACT OF EXOGENOUS DEMAND SHOCK ON THE HOUSING MARKET: EVIDENCE FROM THE HOME PURCHASE RESTRICTION POLICY IN THE PEOPLE’S REPUBLIC OF CHINA Xiaping Cao, Bihong Huang, and Rose Neng Lai No. 824 March 2018 Asian Development Bank Institute The Working Paper series is a continuation of the formerly named Discussion Paper series; the numbering of the papers continued without interruption or change. ADBI’s working papers reflect initial ideas on a topic and are posted online for discussion. Some working papers may develop into other forms of publication. The Asian Development Bank recognizes “China” as the People’s Republic of China. Suggested citation: Cao, X., B. Huang, and R. N. Lai. 2018. The Impact of Exogenous Demand Shock on the Housing Market: Evidence from the Home Purchase Restriction Policy in the People’s Republic of China. ADBI Working Paper 824. Tokyo: Asian Development Bank Institute. Available: https://www.adb.org/publications/impact-exogenous-demand-shock-housing- market-evidence-prc Please contact the authors for information about this paper. Email: [email protected] Xiaping Cao is associate professor of finance at Lingnan College, Sun Yat-sen University; Bihong Huang is a research fellow at the Asian Development Bank Institute; and Rose Neng Lai is a professor of finance at the University of Macau. The views expressed in this paper are the views of the author and do not necessarily reflect the views or policies of ADBI, ADB, its Board of Directors, or the governments they represent. ADBI does not guarantee the accuracy of the data included in this paper and accepts no responsibility for any consequences of their use. Terminology used may not necessarily be consistent with ADB official terms. Working papers are subject to formal revision and correction before they are finalized and considered published. Asian Development Bank Institute Kasumigaseki Building, 8th Floor 3-2-5 Kasumigaseki, Chiyoda-ku Tokyo 100-6008, Japan Tel: +81-3-3593-5500 Fax: +81-3-3593-5571 URL: www.adbi.org E-mail: [email protected] © 2018 Asian Development Bank Institute ADBI Working Paper 824 Cao, Huang, and Lai Abstract In order to deal with the rampant increase in housing prices, the Government of the People’s Republic of China implemented the home purchase restriction (HPR) policy to curb speculation and prevent housing bubbles. This policy triggered an exogenous demand shock to the housing market. Employing a two-step difference-in-differences approach, we find significantly negative policy effects on property transaction volume but a small impact on housing prices. Cities relying heavily on land sales for fiscal revenue experience a considerably higher increase in property investments after implementation of the HPR policy. Keywords: home purchase restriction policy, demand shock, housing bubble, land financing JEL Classification: G12, G18, H83 ADBI Working Paper 824 Cao, Huang, and Lai Contents 1. INTRODUCTION ......................................................................................................... 1 2. POLICY BACKGROUND AND LITERATURE REVIEW ............................................. 3 3. DATA AND EMPIRICAL STRATEGY .......................................................................... 6 3.1 Data ................................................................................................................. 6 3.2 Summary Statistics .......................................................................................... 8 3.3 Method of Estimation ....................................................................................... 8 4. EMPIRICAL RESULTS.............................................................................................. 10 4.1 Effects of the HPR Policy on Housing Prices ................................................ 10 4.2 Effects of the HPR Policy on Housing Sales ................................................. 12 4.3 Effects of the HPR Policy on Housing Construction and Investment ............ 13 4.4 Robustness Check ........................................................................................ 15 4.5 Sources of Cross-Sectional Variations .......................................................... 17 5. CONCLUSION .......................................................................................................... 21 REFERENCES ..................................................................................................................... 22 ADBI Working Paper 824 Cao, Huang, and Lai 1 1. INTRODUCTION Considerable evidence indicates that the collapse of debt-laden housing bubbles is the main cause of many financial crises such as the recent Great Recession (Reinhart and Rogoff 2009). Jordà, Schularick, and Taylor (2015a) conclude that housing bubbles fueled by credit booms are most dangerous and costly. These facts have drawn renewed attention to the necessity and effectiveness of public policy in controlling real estate bubbles. Almeida, Campello, and Liu (2006) and Mian and Sufi (2009, 2015) argue that monetary policies such as low interest rates and easy credit cause bubbles and bursts.1 Cheng, Raina, and Xiong (2014) suggest that financial intermediaries contribute to housing bubbles. Glaeser (2013) finds that housing bubbles often occur when government intervention is minimal. There is an emerging literature investigating the effects of government intervention on the property market.2 Whether government intervention can help avoid housing booms and busts or simply postpone them remains a lingering question. These discussions suggest it is necessary to study the effects of government intervention in the housing market. The People’s Republic of China (PRC) provides a compelling setting for exploring the role of government regulations in the housing market for several reasons. Despite its short history, the importance of the PRC’s housing market cannot be underestimated. As one of the main drivers of the PRC’s economic growth, the real estate sector accounts for one-sixth of its GDP, one quarter of total fixed asset investment, 14% of total urban employment, and approximately 20% of bank loans (International Monetary Fund 2014). Unlike most countries, the PRC’s local governments are the ultimate owners of land and play the dominant role of controlling and managing housing and credit supply. They rely heavily on real estate-related income—land sales in particular—as a source of fiscal revenue. As the second largest economy and the largest trading nation in the world, a sharp slowdown in the PRC’s property sector could have a domino effect on the world economy, especially in the emerging markets. Ahuja and Myrovda (2012) predict that a 10% reduction in the PRC’s real estate investments would shave about 1% off the PRC’s real GDP within the first year and cause global output to decline by roughly 0.5% from the baseline. The PRC’s government has heavily interfered in the housing market through various regulations aimed at maintaining a stable market and curbing speculations or preventing bubbles. Evidently, the PRC’s real estate market has witnessed price upsurges in the past decade, although there were a few setbacks. MacDonald, Mussita, and Sobczak (2012) show that property prices in the PRC increased at a compound annual growth rate (CAGR) of about 16% between 2005 and 2011, more than the 13% recorded in the United States housing market between 2000 and 2005. The drastic price surge has caused extensive concern about a possible housing bubble in the PRC. Studies by Wu, Gyourko, and Deng (2012) and Ren, Xiong, and Yuan (2012) find no conclusive evidence of a housing bubble in the PRC but raise great concern about the over-valuation of housing prices. More recently, Fang, Gu, Xiong, and Zhou (hereafter FGXZ 2015) show that the rampant run-up in housing prices and speculation present significant challenges to the PRC’s economy and regulators. Chen, 1 In addition to Mian and Sufi (2009, 2015), there is a rapidly growing literature examining the links between monetary policy, mortgage borrowing, and housing price appreciation, including Jordà, Schularick, and Taylor (2015b); Del Negro and Otrok (2007); Goodhart and Hofmann (2008); Glaeser and Sinai (2013); Jarocinski and Smets (2008); Williams (2011); and Bernanke (2010). 2 See Almeida, Campello, and Liu (2006); Crowe et al. (2013); International Monetary Fund (2011); Igan and Kang (2011); Kannan, Rabanal, and Scott (2012); and Wong et al. (2011) for studies on government regulations on the real estate sector. ADBI Working Paper 824 Cao, Huang, and Lai 2 Liu, Xiong, and Zhou (2017) report on the real effect of investments crowded out by the housing bubble in the PRC. Several papers (Du and Zhang 2015; Jia et al. 2014; Sun et al. 2015) have examined the impacts of the HPR policy on the housing prices and sales in the markets of Beijing, Shanghai, and Guangzhou. None of them provide a systematic study of the impact of the policy. Faced with rampant price surges and speculation, the PRC’s government has adopted various policy tools, including the increase of the minimum down payment ratio; a cap on the loan-to-value ratio; higher mortgage rates for the second house; taxes on capital gains; credit rationing for real estate developers;3 and so on. When the effectiveness of these measures diminished, the PRC’s government resorted to the heavy-handed home purchase restriction (HPR) policy to curtail speculation. This policy was implemented first in Beijing in May 2010 and was later adopted by 45 other major cities. The PRC’s HPR policy is similar to the measures adopted by the Australia, Hong Kong, and Singapore governments4 in the sense that it directly reduces housing demand by disqualifying certain buyers. Under the PRC’s HPR policy, only investors who have local household registration (hukou) or those with work records in their cities for certain consecutive years are qualified to purchase new homes. Unlike other nationwide cooling measures, such as control of the mortgage rate, the HPR policy is decentralized in that cities can decide whether or not to adopt this policy on their own.5 The variation in the timing of the HPR policy adoption across cities provides a rare opportunity to study exogenous demand shock and its impact on the real estate sector. Using detailed city-level quarterly panel data for the years 2008–2013, we systematically assess the effects of government intervention on the PRC’s housing market. Our data cover various indicators for the real estate sector, including housing price (or index); sales of new homes; investment and construction by the developers; and land sales. We not only systematically analyze the HPR policy’s effect on the housing market but also capture heterogeneous market responses to the policy across cities. This study thus enables a deep understanding of the misalignment of interests when the housing market becomes “too big to fail” for the local economy. Considering that the adoption of the HPR policy is not random, we employ the two-step difference-in-differences (DID) approach developed by Donald and Lang (2007) and Greenstone and Hanna (2014) as our main empirical strategy to draw the causal inference of the policy effect on the housing market. Further, we perform a structural break test as the robustness check on the validity of DID design. We discover a temporary decrease in housing prices and a sharp plunge in the transaction volume of new homes following HPR policy implementation. This evidence is consistent with the policy motivation of curbing speculative demand in the property market. The policy, however, does not address the problem of excessive supply as the increase in property investment and construction continue after the implementation of this policy. 3 People’s Bank of China issued its No. 359 regulation in 2007 to strengthen the management of commercial real estate credit loans to real estate developers. 4 For example, the Singapore and Hong Kong, China governments have implemented several demandmanaging measures to restrict property purchases by foreigners, including a higher down payment ratio, a higher rate of buyer’s stamp duty on property transactions, etc. The Australian government has strengthened its restrictions on property buying by foreigners since 2010. Under those restrictions, temporary residents are allowed to buy established homes with approval from the foreign-investment regulator but have to sell when their temporary visas expire. 5 The central government only provides guidelines that the policy be implemented in the first-tier cities and can be extended on a need basis to the second- and even third-tier cities, rather than being mandated for all cities. ADBI Working Paper 824 Cao, Huang, and Lai 3 More importantly, we investigate the economic mechanisms that explain why the HPR policy is effective in dampening housing prices and transactions but ineffective in reducing construction. This finding echoes the crowd-out effects of the housing boom on manufacturing investment documented by Chen, Liu, Xiong, and Zhou (2017). Real estate booms are usually characterized by rapid increases in prices, but the accompanying construction boom and substantial increase in the number of vacant homes are particularly prominent issues in the PRC and are worthy of further investigation. Our results suggest that the effectiveness of the HPR policy is limited due to strong demand by the PRC’s residents for housing as a major investment vehicle and local authorities’ misaligned incentives and circumvention. In doing so, we relate the effectiveness of the policy with government incentives, especially the reliance on land financing for fiscal revenues. The empirical findings show that cities with heavy reliance on land financing experience no salient impact when they adopt the HPR policy. These findings suggest the moral hazard problems of local politicians in choosing regulatory measures for the “too-big-to-fail” sectors (Choudhry and Landuyt 2011). Our findings are similar in spirit to the recent literature on asset bubbles and speculation. It is well known that in an economy with heterogeneous agents, optimistic investors will bid up asset prices (Miller 1977; Harrison and Kreps 1978; Hirshleifer 2001; Xiong 2013). The HPR policy does not specifically target optimistic investors or speculators and hence will not dampen investor sentiment. Furthermore, the HPR policy can ultimately be considered as an alternative to raising the transaction cost for speculators who can strategically circumvent the HPR policy at some cost. However, Shiller (2000) shows that high transaction cost does not deter asset bubbles in the real estate sector. Thus, it is important to draw analytic inferences on the effect of the HPR policy and provide policy recommendations since the PRC’s government continued to strengthen home purchase restriction policies in many cities in 2016 and 2017 due to the failure of other standard policies in preventing housing bubbles and curbing speculation. The rest of this paper proceeds as follows: Section 2 summarizes the evolution of government policies toward the residential property market and reviews the relevant literature; Section 3 presents the data source and summary statistics and outlines the empirical strategy; Section 4 reports on the main empirical results; Section 5 compares the effects of the HPR policy across cities and Section 6 concludes the paper. 2. POLICY BACKGROUND AND LITERATURE REVIEW The PRC’s private housing market, barely existent twenty years ago, largely continued to boom throughout the last decade both in volume and price.6 It is important not only for the households but also for the overall economy. Due to the shortage of investment tools in the PRC, households in general tend to devote most of their wealth to housing, both for consumption and investment. With the highest home ownership rate of 88% in the world (Economist Intelligence Unit 2011), the value of the PRC’s urban residential property market is estimated to be RMB 115 trillion as of the end of 2012, far outstripping the RMB 23 trillion for the stock market and RMB 26 trillion for the bond market. Real estate has made up more than 60% of the PRC’s household assets since 2008, dwarfing the 48% in the United Kingdom, 32% in Japan, and 26% in the United States.7 Moreover, housing investment is an important pillar for economic 6 A review of China’s housing market can be found in Chen et al. (2011). 7 Standard Chartered Report, “China—Real Estate: Good News in Tough Times,” 4 July 2013. ADBI Working Paper 824 Cao, Huang, and Lai 4 growth, particularly in the PRC, owing to its significant share in overall economic activity (Chen and Zhu 2008). Investment in housing accounts for 25% of total fixed asset investments, contributing to roughly one-sixth of the PRC’s GDP growth (Barth et al. 2012). What is therefore potentially dangerous is a housing market meltdown that can cause a catastrophic crisis, given its size and critical role in the economy (Helbling and Terrones 2003; Goodhart and Hofmann 2008).8 Zhou (2005) and Glindro, Subhanij, Szeto, and Zhu (2005) attempt to explain the underlying factors that caused the housing prices fluctuations, while some focus on the price misalignment and the sustainability of the PRC’s housing boom (Wu, Gyourko, and Deng 2012; Ahuja et al. 2010; Barth et al. 2012; Economist Intelligence Unit 2011). Others look at the relationship between housing prices and land policy regulation (Cai, Henderson, and Zhang 2013; Du, Ma, and An 2010). The PRC’s government has interfered actively and significantly in the private housing market. Ahuja et al. (2010) find that during the past decade, any discrepancies in housing prices were corrected relatively quickly due to government intervention. FGXZ (2015) point out that through major interventions, the PRC’s government has played a more important role in affecting the housing market than its counterparts in the rest of world. Since the mid-1990s, the PRC’s government has made great efforts to promote housing finance and stimulate the growth of the real estate sector to support housing reform and fight against the adverse economic impacts of the 1997 Asian financial crisis. For example, between 1998 and 2002, the central government lowered the mortgage rate five times to encourage home purchases. By 2005, the PRC had become the largest residential mortgage market in Asia, with an outstanding balance exceeding RMB two trillion ($300 billion), an almost 89-fold increase compared to the 1997 balance (Deng and Liu 2009; Zhu 2006). Meanwhile, the government rolled out various policies favoring housing development, such as broadening the scope of development loans and allowing pre-sales. As a result, annual housing investments increased by about six times between 1997 and 2005 (Ye and Wu 2008). The PRC’s housing market has experienced a rapid boom since early 2004. In response, the government implemented a series of policy tools to curtail speculative activities. For example, the minimum down payment ratio was raised to 40% in September 2007 and the mortgage rate was set 10% higher than the benchmark rate. These measures worked well for a short period, partially aided by the global financial crisis that began in 2007. In an effort to avoid an economic slowdown caused by the global financial crisis, the PRC’s government reversed its housing policy in October 2008. This included a series of measures to support housing market growth. Among them, the minimum mortgage rates were adjusted downward to 70% of the benchmark rate and the down-payment ratio was lowered to 20%. Preferential policies were also introduced for first-time home buyers. Fueled by easy credit and lax monetary policy, the housing market regained momentum in mid-2009 and started a new round of price run-ups and a massive construction boom across the nation. In response to the continuing surge in housing prices, the government launched a campaign against the overheated property market in early 2010. Various tightening measures were put in place, such as raising the down payment ratio, prohibiting 8 Davis and Van Nieuwerburgh (2014) provide an excellent review of the literature that explores the interconnections of macroeconomics, finance, and housing. ADBI Working Paper 824 Cao, Huang, and Lai 11 The oscillating trends observed in Figure 2 suggest that the parallel trends assumption of the simple DID or mean shift model (i.e., equation [2A]) may be violated in many cases. This is particularly true for PRC’s housing market where both prices and sales exhibited strong growing trends before the policy’s enactment. Equations (2B) and (2C) that account for differential trends are hence more likely to produce valid estimates. Table 2 reports the policy effects estimated by the two-step DID approach. Column (1) lists the estimate of 𝜋1 from equation (2A). It tests how 𝜎𝜏 on average changes after the policy is mandated. Column (2) presents the estimate of 𝜋1 and 𝜋2 from fitting into equation (2B), where 𝜋1 tests for policy effectiveness by accounting for the trend (𝜏). Column (3) shows the estimation results of equation (2C) that allow for a mean shift and trend break after the policy is in force. We report the estimated effect of the policy four quarters after the implementation as 𝜋1 + 4𝜋3. Table 2: Trend Break Estimates of the Policy Effect on Housing Prices (1) (2) (3) (4) (5) (6) Panel A. PINew.NBS Panel B. PISecond.NBS 𝜋1: l(Policy) 2.42*** 0.03 0.06 1.17 –0.80 –0.78 (0.77) (1.41) (0.68) (0.73) (1.39) (0.53) 𝜋2: Time Trend 0.28* 0.85*** 0.23 0.81*** (0.14) (0.11) (0.14) (0.08) 𝜋3: l(Policy) × time trend –0.97*** –1.00*** (0.14) (0.11) 4-quarter effect = 𝜋1 + 4𝜋3 –3.81*** –4.76*** p-value 0.00 0.00 Observations 17 17 17 17 17 17 Panel C. Price.Cityhouse Panel D. PI.FGXZ 𝜋1: l(Policy) 528.96** 146.36 152.17 –0.02 0.06 0.06** (201.41) (399.92) (228.44) (0.03) (0.05) (0.02) 𝜋2: Time Trend 45.05 196.41*** –0.01 0.01*** (40.78) (36.18) (0.01) (0.00) 𝜋3: l(Policy) × time trend –258.60*** –0.04*** (47.28) (0.00) 4-quarter effect = 𝜋1 + 4𝜋3 –882.20*** –0.10*** p-value 0.01 0.00 Observations 17 17 17 17 17 17 Notes: This table presents the regression results estimated from the equations (2A), (2B), and (2C) for the impact of the HPR policy on housing prices and rental prices. Robust standard errors are in parentheses. The “4-quarter effect” reports the effect of the policy four quarters after implementation with the p-value testing the significance of linear combinations shown below the four-quarter estimates. *** Significant at the 1% level. ** Significant at the 5% level. * Significant at the 10% level. The regression results presented in Table 2 confirm the graphical analysis in the previous subsection that the HPR policy dampened the rampant housing price surge. The results estimated from the most comprehensive second-stage specification (equation [2C]) indicate that four quarters after the policy was in force, NBS property price indexes—PINew.NBS and PISecond.NBS—declined by 3.81 and 4.76 points, respectively, or 2.8% and 3.5% of the sample mean. However, the fall in the prices released by City House is phenomenal, dropping by RMB 882 or 11.43% of the sample means four quarters after the policy was enforced. The price index estimated by FGXZ ADBI Working Paper 824 Cao, Huang, and Lai 12 (2015) fell by 0.1 points, or 7% of the mean value of that in the adopting cities a quarter before policy implementation. 4.2 Effects of the HPR Policy on Housing Sales This subsection examines the policy effect on housing sales. Figure 3 presents event study analysis of the HPR policy impact on housing sales. It indicates that the policy has remarkable impacts on new residential property transactions. Compared with the quarter preceding policy implementation, the floor space sold, the number of flats sold, and the sales amount precipitously dropped in the fourth quarter of policy adoption. Figure 3: Event Study of the HPR Policy on Housing Sales Notes: The figures provide a graphic analysis of the effect of the HPR policy on housing sales by depicting the estimated 𝜎𝜏s from equation (1) against the event time 𝜏. The quarter of the policy implementation, 𝜏= 0, is demarcated by a vertical dashed line in all figures. All property market measurements are normalized to equal zero at 𝜏=−1 and demarcated by the horizontal dashed line. Table 3 provides the regression results for new housing sales. The results derived from equation (2C) imply that four quarters after the policy adoption, the number of units sold, the floor space sold, and the sales amount plummeted by 7,510 units, 783.3 thousand square meters and RMB 12 billion, respectively, at the magnitudes of 55%, 56.3%, and 102% compared to the whole sample mean. This phenomenal fall in the sales volume hints that the policy enforcement is effective in dampening speculation by nonresidents or policy-sensitive buyers. The plunge in both prices and transaction volume is consistent with the findings reported by Sun et al. (2013) in their Beijing sample. ADBI Working Paper 824 Cao, Huang, and Lai 13 Table 3: Trend Break Estimates of the Policy Effect on Housing Sales (1) (2) (3) (4) (5) (6) Panel A. SaleUnit Panel B. SaleFloor 𝜋1: l(Policy) 877.99 –4,329.06** –4,329.75*** 20.04 –442.75** –449.40*** (1,111.68) (1,644.34) (1,333.83) (107.97) (170.65) (137.68) 𝜋2: Time Trend 613.15*** 1,081.28*** 54.48*** 104.60*** (167.64) (212.06) (17.37) (22.17) 𝜋3: l(Policy) × time trend –795.07** –83.49** (276.36) (28.61) 4-quarter effect = 𝜋1 + 4𝜋3 –7,510.0*** –783.3*** p-value [0.00] [0.00] Observations 17 17 17 17 17 17 Panel C. SaleAmount 𝜋1: l(Policy) 972.43 –4,247.24 –4,244.03** (1,696.00) (3,116.36) (1,944.24) 𝜋2: Time Trend 614.65* 1,750.73*** (317.72) (309.01) 𝜋3: l(Policy) × time trend –1,930.47*** (402.80) 4-quarter effect = 𝜋1 + 4𝜋3 –11,965*** p-value [0.00] Observations 17 17 17 Notes: This table presents the regression results estimated from the equations (2A), (2B), and (2C) for the impact of the HPR policy on the sales of new homes Robust standard errors are in parentheses. The “4-quarter effect” reports the effect of the policy four quarters after implementation with the p-value testing the significance of linear combinations shown below the four-quarter estimates. *** Significant at the 1% level. ** Significant at the 5% level. * Significant at the 10% level. 4.3 Effects of the HPR Policy on Housing Construction and Investment We now turn to investigating the HPR policy effect on investment and construction by real estate developers. Figure 4 shows the event study analysis and Table 4 presents the regression results. No sizable policy effect is observed for real estate investment. On the contrary, its growth momentum remained strong in our sample period. The insignificantly positive regression coefficient on the four quarters’ policy effect indicates that property developers did not change their investments despite the policy which was designed to cool the housing market. These findings are reinforced by the insignificant estimation results for the floor space started, under construction, and completed presented in Panels A, B, and C of Table 4. The ineffectiveness of the HPR policy in taming the massive property construction boom is consistent with the reality that the HPR policy is mainly designed to depress speculation from the demand side. A manifestation of this includes the emergence of several ghost cities with an abundance of empty houses. ADBI Working Paper 824 Cao, Huang, and Lai 14 Figure 4: Event Study of the HPR Policy on Housing Construction Notes: The figures provide a graphic analysis of the effect of the HPR policy on housing construction by depicting the estimated 𝜎𝜏s from equation (1) against the event time 𝜏. The quarter of the policy implementation, 𝜏= 0, is demarcated by a vertical dashed line in all figures. All property market measurements are normalized to equal zero at 𝜏=−1 and demarcated by the horizontal dashed line. Table 4: Trend Break Estimates of the Policy Effect on Housing Investment and Construction (1) (2) (3) (4) (5) (6) Panel A. FloorStarted Panel B. FloorUnderConstruction 𝜋1: l(Policy) 279.12** 196.87 196.02 4,519.11*** 803.13 802.69 (95.65) (195.77) (202.10) (602.06) (491.48) (509.59) 𝜋2: Time Trend 9.68 18.98 437.18*** 427.82*** (19.94) (32.40) (50.09) (80.54) 𝜋3: l(Policy) × time trend –15.59 16.03 (41.96) (105.37) 4-quarter effect = 𝜋1 + 4𝜋3 133.68 866.79 p-Value [0.62] [0.36] Observations 17 17 17 17 17 17 Panel C. FloorCompleted Panel D. Investment by Developers 𝜋1: l(Policy) 166.49** 120.50 121.58 4,189.16*** –428.02 –432.95 (63.87) (131.57) (126.17) (725.36) (478.88) (434.71) 𝜋2: Time Trend 5.41 28.14 543.20*** 438.31*** (13.41) (19.94) (48.80) (68.70) 𝜋3: l(Policy) × time trend –38.91 179.55* (26.09) (89.89) 4-quarter effect = 𝜋1 + 4𝜋3 –34.05 285.26 p-value [0.45] [0.25] Observations 17 17 17 17 17 17 continued on next page ADBI Working Paper 824 Cao, Huang, and Lai 15 Table 4 continued (1) (2) (3) (4) (5) (6) Panel E. Pieces of Land Sold 𝜋1: l(Policy) –0.24 –1.57 –1.57 (1.04) (2.12) (2.20) 𝜋2: Time Trend 0.16 0.21 (0.22) (0.35) 𝜋3: l(Policy) × time trend –0.10 (0.45) 4-quarter effect = 𝜋1 + 4𝜋3 –1.96 p-value [0.50] Observations 17 17 17 Notes: This table presents the regression results estimated from the equations (2A), (2B), and (2C) for the impact of the HPR policy on the construction of residential property, land price, and land sales revenue. Robust standard errors are in parentheses. The “4-quarter effect” reports the effect of the policy four quarters after implementation with the p-value testing the significance of linear combinations shown below the four-quarter estimates. *** Significant at the 1% level. ** Significant at the 5% level. * Significant at the 10% level. Construction of residential properties is a long process and may respond to policy implementations with long lags in construction. For example, floor area completed may be determined by housing investment decisions made two to three years ago. Even floor area started may lag as land purchases and removal of old residents and structures take time. To avoid this measurement issue, we test the policy impact on land transactions because they provide more timely reflections of developers’ willingness to construct. Panel E indicates that the policy has no measurable impacts on the pieces of land sold, implying that the purchase of land by developers did not change after the policy was implemented. 4.4 Robustness Check We perform several robustness checks to verify the validity of the two-step DID estimation results. Considering that some variables like housing prices, sales, and construction might grow exponentially, we replace them with logarithm value and re-estimate all models. All the results are qualitatively similar.17 In addition, we conduct the structural break test and a different sample test. Structural Break Test—We first employ the structural break test developed by Greenstone and Hanna (2014) to check the robustness of the two-step DID design. The basic idea is to assess whether there is a structural break in the policy parameters (i.e., 𝜋1 and 𝜋3) estimated from the second-stage specification of equation (2C) around the time of policy implementation. The test first identifies the time at which the largest change in parameters (represented by the largest change in the F-statistics) occurs and then generates p-values to judge whether the changes in those parameters are different from zero. A significant break around the time of policy implementation, i.e., 𝜏= 0, or some quarters after 𝜏= 0 would prove the existence of a policy effect from the DID results. In contrast, failure to find a break, or finding of a break significantly before the time of policy adoption, would imply the ineffectiveness of the policy. 17 Due to space constraints, we do not report the results here. They are available upon request. ADBI Working Paper 824 Cao, Huang, and Lai 16 We follow Greenstone and Hanna (2014) to use the Quandt Likelihood Ratio (Quandt 1960) statistic to select the maximum value of the F-statistics to test the existence of a break at an unknown date. Figure 5 and Table 5 report our estimation results. As shown in Figure 5, the structural breaks of the NBS price index—PINew.NBS and PISecond.NBS—occur before policy implementation while the QLR statistic identifies the significant breaks three quarters preceding the event. This finding implies the ineffectiveness of the policy in curbing the surge in housing prices. However, the test on the price index calculated by FGXZ (2015) is very significant because it shows the precipitous drop in housing prices three quarters after policy enforcement, i.e., 𝜏= 3. The evidence might indicate that the hedonic housing price index developed by FGXZ (2015) has a better quality than the NBS property price index. Figure 5: F-statistics from the QLR Test for the HPR Policy Notes: The figures show the structural break tests using the Quandt Likelihood Ratio (QLR) statistic. The horizontal axis is the event time τ. The vertical axis is the F-statistics for the QLR tests. With respect to the transactions, Figure 5 evidently chooses the occurrence of the biggest F-statistics around 𝜏= 0. Moreover, Table 5 reveals that the null hypothesis of no break at 𝜏=−1 can be significantly rejected for the sales amount. These findings further prove that the policy causes a sharp decline in the property transaction volume. ADBI Working Paper 824 Cao, Huang, and Lai 17 The structural break test results for real estate investment, floor space started, under construction, and completed, as well as the pieces of land sold, are broadly supportive of the findings of the previous two subsections. The null hypothesis of zero effect cannot be rejected, confirming that the policy does not stop the construction boom, nor does it help to address the potential oversupply of housing. Table 5: Structural Break Analysis Quarter of Maximum F-statistics QLR Test Statistic PINew.NBS –3 76.007*** PISecond.NBS –3 68.713*** Price.Cityhouse –3 27.751*** Price.CREIS –3 9.294 PI.FGXZ. 3 52.614*** SaleUnit 0 9.405 SaleFloor 0 9.4297 SaleAmount –1 15.456*** Investment 4 4.462 FloorStarted 4 5.387 FloorUnderConstruction 1 2.908 FloorComplete 5 2.469 Notes: This table presents the results of structural break tests using the QLR test statistic and the corresponding quarter of the break in the data estimated from the specification of equation (2C). *** Significant at the 1% level. ** Significant at the 5% level. * Significant at the 10% level. Robustness Test with Alternative Model and Different Samples—NBS reports the property price index for a relatively small sample of 70 cities. Using the housing price data from CREIS, City House, and FGXZ (2015), we expand the sample to include 139 cities (45 of them adopting the HPR policy). The robustness results reported in Appendix 2 are consistent with those for the sample of 70 cities. Several points need to be noted: First, the decrease in the housing prices or price index following policy implementation is much larger than that of the 70-city sample; second, the transaction measured by the sales amount, sales unit, and floor space sold plummeted significantly after the policy was in force, although it was smaller in magnitude than that of the 70-city sample; and third, the HPR policy had no measurable impacts on housing construction and investment. 4.5 Sources of Cross-Sectional Variations The empirical evidence presented in the previous sections implies that the HPR policy successfully dampens home purchases but does not curb the construction boom and excessive supply. This subsection explores the factors that may explain why demand and supply respond differently to the policy. The qualitative and quantitative evidence suggest the discrepancy reflects real estate developers’ positive expectations of housing demand and local authorities’ over-reliance on the real estate sector. ADBI Working Paper 824 Cao, Huang, and Lai 18 Due to the shortage of investment channels, the property market is the main outlet for PRC households saving for retirement or children’s marriages (Wei and Zhang 2011). Real estate has made up more than 60% of Chinese household assets since 2008, dwarfing the 48% in the United Kingdom, 32% in Japan, and 26% in the United States. In the PRC, the demand for housing is especially strong in cities where income rises rapidly. We hypothesize that if price does not decline after policy implementation in the cities where demand is expected to be strong, then the developers will continue to build, even if the implementation of the HPR policy is a big and surprising shock to the market. We follow Greenstone and Hanna (2014) to assess this hypothesis. The main idea is to divide the sample cities into those above and below the median values of their GDP growth rate which reflect the demand for housing, estimate separate 𝜎𝜏s for these cities with equation (1), stack the two sets of 𝜎𝜏s obtained from the estimation of equation (2C), and then test whether the policy effects after implementation are the same for the two sets of policy-adopting cities. Considering that real estate developers’ response to the HPR policy may lag behind demand, we test the effect across two types of cities for four (π1+ 4π3) and eight quarters (π1+ 8π3) after the policy was enforced. Panels 1–4 of Table 7 show that the estimates of all price indicators are positive, indicating that housing prices not only did not decline but actually increased in cities whose GDP growth rates were above the median values. Hence, we expect real estate developers in these cities to continue construction. Consistent with our expectation, Panels 5 and 6 in Table 6 suggest that the investment and floor space increased after policy implementation in cities where long-run demand for housing is expected to be strong. Table 7: Differences in HPR Policy Effects across Cities on GDP Growth Difference in 4-quarter Effect Difference in 8-quarter Effect Panel 1. PINew.NBS 2.56 4.85* (0.10) (0.19) Panel 2. PISecond.NBS 0.75 1.55 (0.51) (0.64) Panel 3. PI.FGXZ 0.04** 0.12 (0.07) (0.41) Panel 4. Price.Cityhouse 920.87** 1824.1*** (0.05) (0.01) Panel 5. Investment 484.46 1527.4 (0.39) (0.68) Panel 6. FloorStarted 200.14 619.67 (0.66) (0.37) Notes: This table explores how the HPR policy’s effect four quarters after implementation varies in cities above (as opposed to below) the median measures of GDP growth. p-values are in parentheses. *** Significant at the 1% level. ** Significant at the 5% level. * Significant at the 10% level. ADBI Working Paper 824 Cao, Huang, and Lai 19 The heavy and growing reliance of Chinese local governments on the real estate sector for fiscal income and economic growth generates far-reaching impacts on the housing market. Due to the intergovernmental fiscal relationship established in 1994, local governments receive half of the nation’s fiscal revenue but are responsible for 80% of spending (The Economist 2012). Having heavy expenditure responsibilities, local governments depend substantially on off-budgetary sources such as profits from expropriating farmers’ land, revenue related to land sales and transactions, and so forth (Huang and Chen 2012). As shown in Figure 6, the ratio of land sales revenue to municipal government budgetary revenue18 increased from less than 1% in the early 1990s to around 80% in 2010. Among our 70 sample cities, the average ratio of land sales revenue to budgetary revenue for the years 2001–2011 shows large variations across cities, ranging from 11% to 117%. Cities having meager fiscal resources or tremendous needs for infrastructure investment exhibit higher degrees of reliance on land finance. Figure 6: Ratio of Land Sales Revenue to Budgetary Revenue Notes: The data for the years 1989–2009 is from Barth, Lea, and Li (2012), and the rest is calculated by the authors where the data of land sales revenue is obtained from the China Land & Resources Yearbook (2011–2013) and the data of budgetary revenue is from the CEIC. With heavy reliance on the real estate sector to finance their spending and investment in infrastructure, local governments face the dilemma of whether to correct housing bubbles or maintain land prices and economic growth via property investment. To understand local governments’ real incentive in the housing market, we first examine the change in land supply for residential property before and after the HPR policy was implemented because local governments are the ultimate owners of land in the PRC. If cooling down housing prices is the main objective of local governments, more land will be supplied to the market following policy implementation. However, Figure 7 indicates that the supply of residential land declined considerably in HPR-implementing cities after 2011. The share of residential land in total land supply was on average approximately 50% before 2010, but fell to 30% in 2014. The change in the annual growth rate of residential land supplied to the market is more phenomenal, declining sharply from 65% in 2009 to –11% in 2014.19 18 Budgetary revenue consists mainly of tax revenue and state-owned enterprise contributions. 19 We do not use DID estimation to test the change of land supply following the policy implementation because the quarterly data on land supply is currently unavailable. ADBI Working Paper 824 Cao, Huang, and Lai 20 Figure 7: Residential Land Supply in HPR-implementing Cities Note: The data is from CREIS. Despite political pressure from the central government to control the housing price surge, local governments may still support construction activity by requiring government-controlled banks to provide cheap and easy credit to the developers (Glaeser et al. 2017). Following Greenstone and Hanna (2014), we divide the sample cities into those above and below the median value of the ratio of land sales revenue to budgetary revenue and then estimate separate 𝜎𝜏s for these cities with equation (1), stacking the two sets of 𝜎𝜏s obtained from the estimation of equation (2C), and then test whether the policy effect is the same for the two sets of policy-adopting cities. Table 8 reports the estimation results. The positive estimate of the 8-quarter policy effect on real estate investment, floor space started, and floor space under construction indicates that construction activity continued to increase after policy implementation in cities where land sales revenue accounts for a large share of local fiscal revenue. This implies that the top-down effort in curbing housing prices via the HPR policy was not fully supported by the local authorities. Excess supply in the housing market is an unavoidable consequence of misaligned incentives. A manifestation of this includes the emergence of several ghost cities with an abundance of empty houses. Table 8: Differences in HPR Policy Effects across Cities for Land Finance Reliance Difference in 4-quarter Effect Difference in 8-quarter Effect Panel A Investment 1,947.0* 3,577.0** 0.09 0.04 Panel B FloorStarted –68.21 84.03 0.89 0.91 Panel C FloorUnderConstruction 810.54 577.48 0.47 0.73 Notes: This table explores how the HPR policy’s effect four quarters after implementation varies in cities above (as opposed to below) the median measures of land finance reliance. *** Significant at the 1% level. ** Significant at the 5% level. * Significant at the 10% level.