The impact of housing macroprudential policy on firm innovation: empirical evidence from China

Fueled by the remarkable upsurge in housing prices that have been witnessed across global over the past few decades, the housing macroprudential policy (HMP) plays an increasingly important role in alleviating the financial risks rooted in the housing bubbles. Despite considerable research on HMP’s effects on domestic and international financial markets, micro-level empirical evidence is scarce, and prior work has largely neglected spillover effects on local firms’ behavior—especially innovation, a primary driver of economic growth. This paper asks whether and how regulation of housing markets through HMP affects firms’ innovation activities, using a city–firm panel from China during a period of dynamically changing HMP. China provides an appropriate setting due to its prolonged housing boom and rapid expansion in R&D investment, alongside frequent local regulatory interventions. Stylized facts show weak growth in innovation before 2014 amid strong housing price growth, and a sharp rise in R&D after 2015 as housing prices fluctuated, suggesting potential crowd-out effects of high housing prices on innovation. The study tests whether tightened HMP can promote firm innovation, the channels through which it operates, and the heterogeneity across cities and industries, yielding implications for economic growth and policy design.

Institutional background of HMP in China. In response to housing booms and financial risks, many economies have implemented macroprudential policies targeting real estate, including dynamic loan-loss provisions, loan-to-value (LTV), and debt-to-income (DTI) ratios to mitigate systemic risk. China employs these tools to curb excessive leverage and speculation and to limit non-owner-occupied demand. LTV ratios vary with the number of existing properties owned, allowing higher leverage for first-time buyers. Beyond consumer-side tools, China also targets real estate firms’ leverage and short-term debt coverage (from 2020) and instituted real estate loan concentration limits at end-2020 to cap banks’ exposure. HMP and innovation. HMPs directly affect housing markets by lowering prices and curbing bubbles. Firms interact with real estate via mortgage-based financing and direct investment. Prior studies show real estate investment by non-financials can crowd out fixed investment and regional innovation, though real estate can also serve as collateral to ease financing constraints, potentially supporting innovation. This literature largely overlooks HMP’s effect on firm innovation. If innovation and housing investment are substitutes, tighter HMP that lowers real estate returns can redirect firms’ portfolios toward long-term innovative investment. Conversely, if firms are heavily invested in real estate, HMP-induced price fluctuations may weaken balance sheets and available funds, potentially crowding out innovation. Hypothesis 1 (H1): HMP will have a positive effect on firm innovation. HMP, housing dependency, and innovation. Two forces link housing and innovation: a crowd-out effect (high real estate returns draw resources away from innovation) and a collateral effect (real estate as valuable collateral enhances financing capacity). The net effect of HMP depends on cities’ reliance on housing. In housing-dependent cities, HMP’s influence on innovation may be limited; local fiscal pressures from lower prices and tax revenues may also reduce government support for innovation. In less housing-dependent cities, HMP may more strongly reallocate resources toward innovation. Hypothesis 2 (H2): The positive effect of HMP on firm innovation will be more pronounced in cities with lower housing dependency. HMP, financial development, and innovation. Innovation’s uncertainty and risk make financing capacity critical. Local financial development may condition HMP’s effect: in well-developed markets, better funding access dampens HMP’s impact; in less developed markets, tighter HMP may reduce shadow banking leakage, improve bank stability, and redirect credit toward firms, facilitating innovation. Hypothesis 3 (H3): The positive effect of HMP on firm innovation will be more pronounced in regions with less developed financial systems.

Data. The sample spans 2010–2019, avoiding the 2008 global financial crisis and COVID-19 onset. Firm-level R&D data come from WIND and CSMAR; city-level HMP implementation is compiled from People’s Bank of China disclosures and local government announcements, merged to firms by registered city. Exclusions: financial and real estate industries, ST/*ST/PT warning or delisted stocks, and firms with substantial missing data (e.g., ≥3 consecutive years). Continuous firm variables are winsorized at 1%. The final sample comprises 13,717 firm-year observations for 1,947 listed firms across 54 cities in 28 provinces. Variables. HMP is measured using quarterly changes in LTV and DTI tools at the city level. A categorical quarterly indicator is defined as 1 (tightening), 0 (no related policy), or −1 (loosening), then aggregated annually by summation to HMP_T ranging from −4 to +4. Additional dummies capture asymmetric effects: HMP_tight and HMP_loose. Robustness uses an alternative continuous proxy via average LTV ratios and a down payment–based measure (HMP_HL = 100% − average household LTV; higher implies tighter HMP). Firm innovation is measured primarily by R&D intensity (R&D expenditure / operating income), with log R&D as an alternative. Controls include firm size (log assets), ownership concentration (Top10), profit growth, firm age (log), digital transformation (Dig), employment cost (log), leverage (Lev), cash holdings (Cash), and city log GDP. Empirical specification. Baseline OLS regressions estimate the effect of city-level HMP on next-year firm innovation: RD_{i,t+1} = c + α HMP_{k,t} + γ Controls_{i,t} + μ_industry + τ_year + ε_{i,t}. A positive α indicates that tighter HMP increases subsequent R&D intensity. Descriptive statistics show mean R&D/income of 4.755% (SD 4.712%, min 0.02%, max 25.23%), and mean HMP_T of 0.204, indicating generally active and tightening policy stance. Mechanism analysis. Mediation regressions assess leverage and cash holdings as channels: RD_{i,t+1} = c + α HMP_{k,t} + β Mediator_{i,t} + γ Controls + μ + τ + ε. If α becomes insignificant or diminishes after including the mediator, this supports a channel effect. Heterogeneity analyses. Subsamples and interactions assess heterogeneity by industry, city housing dependency (ratio of housing investment to GDP; High HD indicator), and city financial development (loan balance/GDP; Fin indicator). Robustness. Endogeneity is addressed via IV-2SLS using residential land area (RLA) as an instrument for HMP_T (with under-, weak-, and over-identification tests), Heckman selection correction, and dynamic OLS including lagged R&D. Additional robustness includes alternative innovation and HMP measures and additional fixed effects (city FE; city, industry, and year FE; industry×year FE). Standard errors are clustered at the firm level.

- Baseline effect: Tightening HMP significantly promotes firm innovation. In multivariate OLS, HMP_T has a positive coefficient (e.g., 0.087, t=1.95), implying roughly an 8.7% marginal increase in R&D intensity (R&D/Income) in the following year per unit tightening. Univariate estimates are also positive (0.238, t=4.29). - Asymmetry: Tightening raises next-year R&D, whereas loosening reduces it more substantially, consistent with fewer restrictions encouraging real estate investment and depressing innovation. - Mechanisms: - Leverage channel: Firm leverage significantly impedes innovation (Lev coefficient ≈ −0.055, t≈−13.13). Including leverage renders the HMP effect smaller and insignificant, and HMP is associated with lower leverage, indicating HMP promotes innovation by reducing leverage. - Cash holdings channel: Cash positively relates to innovation (Cash coefficient ≈ 0.054, t≈8.79). Including cash reduces the HMP coefficient magnitude and significance, suggesting HMP enhances innovation by encouraging higher cash buffers. - Industry heterogeneity: Positive HMP effects are more evident in asset-heavy sectors (e.g., Transportation & Delivery, Manufacturing), while negative or insignificant effects are observed in Information Technology & Services and Farm/Forest/Fishing; Health and Social Work shows no significant effect. - City-level heterogeneity: - Housing dependency: In low-dependency cities, HMP_T is strongly positive (e.g., 0.130, t≈3.10). In high-dependency cities, effects are not significant. The interaction High HD × HMP_T is negative (≈ −0.205, t≈−3.00), indicating attenuation where housing dependency is high. - Financial development: In less financially developed cities, HMP_T is positive and significant (≈ 0.094, t≈2.03), while in better-developed cities the effect is negligible. The interaction Fin × HMP_T is negative (≈ −0.100, t≈−1.94), implying weaker promotion in financially advanced areas. - Endogeneity checks: IV-2SLS with RLA instrument yields a positive HMP effect (≈ 0.155, SE 0.066). Heckman selection correction also shows a positive effect (≈ 0.115, SE 0.059). Dynamic OLS with lagged R&D shows strong persistence (lag coefficient ≈ 0.92) and a positive HMP effect (≈ 0.038–0.044). - Robustness: Results hold using alternative innovation measures (log R&D), alternative HMP measures (down payment ratio–based HMP_HL), and with additional fixed effects (city FE; city+industry+year FE; industry×year FE), where HMP_T remains positive (e.g., 0.015*, 0.032**, 0.092**).

The study addresses whether and how housing macroprudential policy spills over to firms’ innovative investment. The empirical evidence indicates that tightening HMP redirects resources away from real estate, mitigates leverage, and encourages cash buffers, together facilitating higher R&D spending. This supports the substitution (crowd-out) perspective: by lowering the relative attractiveness of housing investment and limiting excessive credit expansion, HMP encourages firms to prioritize longer-term, value-creating innovation. The asymmetric effects, where loosening dampens R&D more forcefully, reinforce the sensitivity of firms’ investment allocation to the housing-finance environment. The heterogeneity analyses clarify boundary conditions: the promotion effect is strongest in cities less reliant on housing and with less developed financial systems, where HMP more effectively reduces speculative flows and shadow banking leakage and improves credit allocation to the real economy. In contrast, in housing-dependent or financially advanced cities, fiscal incentives or abundant financing options lessen HMP’s marginal impact. Together, these findings show that macroprudential interventions in housing can foster firm-level innovation under particular local conditions, informing both urban and financial policy design.

The paper documents a positive spillover effect of tighter housing macroprudential policy on firm innovation in China from 2010–2019. Using a large city–firm panel and multiple identification strategies, the results are robust to alternative measures, fixed effects, and endogeneity corrections. Mechanism analysis implicates reduced leverage and greater cash holdings as key channels. The effects are heterogeneous: they are stronger in cities with lower housing dependency and weaker financial development. These insights expand the literature on macroprudential policy by highlighting firm-level innovation responses and offer policy implications on balancing financial stability with growth objectives. Future research could explore asymmetric HMP effects on broader firm performance and examine interactions between macroprudential, fiscal, and monetary policies in shaping innovation outcomes.

- Sample scope: The analysis covers 13,717 firm-year observations for 1,947 listed non-financial, non-real-estate firms across 54 Chinese cities with relatively better HMP disclosure. Results may not generalize to unlisted firms, excluded industries, or cities with limited disclosure. - Time period: 2010–2019 avoids the 2008 crisis and COVID-19 period; effects may differ in other macro environments or post-2020 regulatory changes. - Measurement of HMP: Main measures rely on observed LTV/DTI tightening/loosening events aggregated annually, with alternative proxies (average LTV, down payment ratios). Policy intensity and enforcement nuances may be imperfectly captured. - Innovation measure: R&D intensity (and log R&D) is used instead of patent-based outcomes; it captures investment but not innovation output quality or success. - Potential unobservables: Despite controls, fixed effects, IV, Heckman, and dynamic models, residual omitted variables or measurement errors at the city or firm level may remain.