Housing prices and household consumption: a threshold effect model analysis in central and western China

Over the past three decades, the world economy has experienced unprecedented growth and a global redistribution of wealth and population. In China’s rapid urbanization, households have accumulated wealth, with housing becoming the most valuable asset (about 64% of household assets in 2019). Meanwhile, housing prices have risen sharply, pressuring households to increase savings, accept less favorable locations, and endure longer commutes and worse housing conditions. From 2005 to 2020, national average housing prices rose from RMB 2885/m² to RMB 6626/m²; the housing price-to-income ratio reached 29.06 in 2020, and affordability indices declined, signaling growing unaffordability. The central and western regions of China—classified as underdeveloped—display a simultaneous pattern of low income, low household non-housing consumption rates, and high housing prices. Despite multiple policy efforts (monetary, macro-prudential, fiscal, land) to regulate housing markets and stimulate non-housing consumption, consumption remains sluggish. Existing literature suggests both positive wealth effects and negative crowding-out effects of housing price changes on consumption, and highlights a role for the housing credit channel. However, evidence is lacking on whether credit constraints inhibit wealth effects in underdeveloped Chinese regions. This study asks: how do housing prices, in combination with housing credit constraints, affect household non-housing consumption in central and western China? How do these effects differ between the two regions? The study contributes by focusing on underdeveloped regions (as opposed to developed nations or China’s eastern cities) and by introducing housing credit constraints as a threshold variable in a varying coefficient panel model to endogenously identify thresholds. The paper proceeds with literature review, theoretical framework, methods and data, model building, results, discussion, and conclusions with policy implications.

Prior work in developed countries generally finds that rising housing prices promote aggregate consumption via the Life Cycle and Permanent Income hypotheses: changes in housing wealth affect current consumption (wealth effect) and collateral values can ease liquidity constraints (collateral/credit channel). For China, findings are mixed: some identify a crowding-out effect where higher housing prices constrain consumption; others find positive effects. Housing affordability is critical; unaffordable prices are linked to reduced non-housing consumption, especially for low- and middle-income households, due to higher rent/mortgage burdens. Mechanisms are commonly grouped as direct wealth effects and collateral constraint effects. In mature credit markets, home equity withdrawals via second mortgages enable homeowners to convert housing wealth into consumption; in China, such withdrawals are rare, and social norms (housing in marriage, elderly care, high savings, preference for ownership due to access to better public education resources) shape responses. A home-purchase channel is noted: rising prices induce potential buyers (and sometimes their parents) to cut consumption to save for down payments and mortgages—the "housing slave effect." Credit and consumption: In mature credit markets, housing wealth can be leveraged to raise consumption; in less developed credit markets, equity is hard to access, forcing higher savings and constraining consumption. Mortgage debt and credit conditions feed back with housing markets: more debt can raise housing demand and prices; rising prices boost consumption via wealth effects and then expand household debt. Underdeveloped regions in China: Compared with the east, central and western regions lag economically, with worse affordability. Policies have spurred real estate, yet heavy housing burdens persist. Some studies suggest crowding-out effects dominate in less developed regions; heterogeneity in financial market development makes the net impact unclear. Existing research often treats credit constraints as exogenous; however, credit constraints crucially affect homebuying, mortgage payments, and the effectiveness of housing wealth, thereby influencing consumption. Additionally, China’s hukou and nearby enrollment policies link homeownership to access to better public primary/secondary education, affecting housing demand and consumption behavior. This study addresses gaps by examining housing prices combined with credit constraints in central and western urban China.

Study area and data: The sample covers 18 provincial-level units in central (Shanxi, Anhui, Jiangxi, Henan, Hubei, Hunan) and western (Inner Mongolia, Guangxi, Chongqing, Sichuan, Guizhou, Yunnan, Tibet, Shaanxi, Gansu, Qinghai, Ningxia, Xinjiang) China. In 2020, central and western urban populations were about 215.18 million and 219.25 million, with lower disposable incomes than the national average. Both regions showed low non-housing consumption rates and relatively high housing prices. Data are mainly from China’s Statistical Yearbook, Real Estate Statistical Yearbook, and local yearbooks (2000–2020). The panel spans 18 provinces, 2005–2020. CPI2004=100 is used to deflate nominal prices; some variables are log-transformed. Variables: Explained variable: Household non-housing consumption per urban resident (CON), excluding housing expenditures. Explanatory variable: Housing price fluctuation (HPF) defined as (HP_t – HP_{t-1})/HP_{t-1}. Threshold variable: Housing credit constraints (HCC), proxied by a loan-to-income type measure (ratio of mortgage loans to GDP as reported), where a higher ratio implies looser constraints. Controls: Urban resident disposable income (URDINC), child dependency ratio (CDR, population under 15 relative to 15–64), and elderly dependency ratio (EDR, 65+ relative to 15–64). Theoretical framework: A life-cycle/permanent income model with housing as both consumption and collateral. The household maximizes expected lifetime utility U(CON_t,H_t) subject to budget and borrowing constraints, with borrowing limited by θ_t P_t H_t. Using a log utility form U = ln CON_t + j ln H_t, the Euler equations show that non-housing consumption depends on current/future variables, housing prices, and the tightness of credit constraints (captured by γ_t and θ_t). Hypotheses: (H1) Housing prices directly and indirectly affect the household allocation between non-housing and housing consumption; (H2) consumption decisions adjust in response to housing prices combined with credit constraints. Econometric model: Panel threshold regression (Hansen, 1999), allowing the effect of HPF on ln(CON) to differ across regimes defined by HCC relative to an estimated threshold γ. Single-threshold specification: ln CON_it = μ_i + β1 HPF_it I(HCC_it ≤ γ) + β2 HPF_it I(HCC_it > γ) + β3 Z_it + ε_it, with province fixed effects and controls Z_it = {ln URDINC, CDR, EDR}. Threshold estimation minimizes residual sum of squares over candidate γ; significance is tested via bootstrap F-tests and LR-based confidence intervals. Data tests and estimation: Panel unit root tests (LLC and Fisher-ADF) indicate stationarity at levels for all variables (significant at 10% or better). Threshold tests via bootstrap identify a significant single threshold in both regions (central and western). LR statistics are used to form 95% confidence intervals for γ. Separate regional models are estimated for central and western regions.

- Threshold existence and values: A single HCC threshold is significant in both regions. Estimated threshold values: central γ = 0.0108 (95% CI [0.0090, 0.0110]); western γ = 0.0015 (95% CI [0.0013, 0.0017]). - Effects of HPF on non-housing consumption by regime: - Central region: - HCC ≤ γ: β(HPF) = -0.047 (SE 0.0498), negative (not statistically significant at 10%). - HCC > γ: β(HPF) = 0.149*** (SE 0.0515), positive and significant. - Western region: - HCC ≤ γ: β(HPF) = -0.463*** (SE 0.0732), significantly negative. - HCC > γ: β(HPF) = -0.015 (SE 0.0305), negative but not significant; elasticity magnitude decreases versus first stage. - Controls (elasticities/coefficients): - ln(URDINC): 0.901*** (central, SE 0.0182); 0.923*** (western, SE 0.0147). - CDR: 0.614*** (central, SE 0.175); 0.251* (western, SE 0.131). - EDR: -0.852*** (central, SE 0.255); -1.216*** (western, SE 0.232). - Model fit and sample: R² = 0.991 (central), 0.982 (western). Observations: 96 (central), 192 (western). - Descriptive dynamics: 2005–2020, average HP rose from ~RMB 1708 to ~RMB 4757 (central) and ~RMB 1711 to ~RMB 4839 (western); HPF fluctuated with common cycles. URDINC grew faster than non-housing consumption; consumption rates (CON/URDINC) declined steadily, with 2020 dips influenced by COVID-19. - Interpretation: Credit constraints act as a regime-switching mechanism. When HCC is looser than the threshold (HCC > γ), rising prices stimulate non-housing consumption in the central region (wealth/liquidity effects dominate), while in the western region rising prices still tend to crowd out consumption (albeit less strongly). Overall, HCC and HPF jointly affect consumption, but they are not the primary drivers of persistently low non-housing consumption rates in these underdeveloped regions.

The findings confirm that housing credit constraints are a threshold-type mechanism altering how housing price fluctuations affect household non-housing consumption. Below the threshold (tighter credit), HPF has a negative effect on consumption in both regions (crowding-out dominates). Above the threshold (looser credit), the central region sees HPF positively affecting consumption (increased liquidity/wealth effect), whereas the western region still experiences negative effects, though with reduced magnitude. Regional heterogeneity likely reflects differences in homeownership and market structure. CHFS data indicate lower homeownership in the western region than the central region, attenuating wealth gains from price appreciation and intensifying rent burdens for non-owners. Rising prices elevate living costs, induce higher saving for down payments, and reduce current non-housing consumption—consistent with the "housing slave effect." Despite policies that relaxed HCC over time—improving the liquidity of housing wealth—the study observes low and declining non-housing consumption rates amid low incomes and rising prices. This pattern is linked to unaffordable housing (high PIR across provinces) combined with social-institutional factors, notably the linkage between homeownership and access to better public primary/secondary education under nearby enrollment and hukou-related rules. These factors incentivize households, especially with school-aged children, to prioritize home purchase over non-housing consumption, even as credit becomes looser. Thus, while easing credit can enhance the wealth channel, it may also fuel further price increases and worsen affordability, limiting net gains in consumption, especially in the western region where crowding-out persists. The results suggest that credit constraints and price fluctuations together shape consumption but are not the dominant cause of low non-housing consumption rates; rather, affordability pressures and education-linked ownership incentives are central.

This study uses a panel threshold model on 18 provinces in central and western China (2005–2020) to quantify how housing price fluctuations and housing credit constraints jointly affect household non-housing consumption. Main conclusions: (1) Housing credit constraints exhibit a significant single-threshold role in mediating HPF’s impact on consumption. (2) In the western region, HPF’s elasticity on non-housing consumption remains negative across regimes, with reduced magnitude when HCC exceeds the threshold. (3) In the central region, HPF’s effect switches from negative to positive once HCC exceeds the threshold. (4) Relaxing HCC enhances housing wealth liquidity, increasing consumption in the central region but not overturning crowding-out in the western region. (5) Persistently low non-housing consumption rates are better explained by unaffordable housing combined with homeownership’s linkage to access to better public education resources than by credit constraints alone. Policy implications: Policymakers should not rely on real estate booms or further credit loosening to drive consumption. Instead, prioritize stabilizing housing prices, raising incomes, and reducing income uncertainty. Address structural drivers of homeownership demand by balancing public education resources across districts to lower ownership-for-access pressures. Relaxing HCC without stabilizing prices and addressing education-linked incentives may spur prices and exacerbate unaffordability while delivering limited consumption gains. Future research: The authors plan to explore more deeply the relationship between public education and housing prices to better understand its influence on household economic behavior.

The study relies on provincial-level panel data for 18 central and western provinces from 2005–2020, which may mask intra-provincial heterogeneity and micro-level household behaviors. Due to non-availability of macro-level housing wealth data, housing price fluctuation (HPF) is used as a proxy for changes in housing wealth, and housing credit constraints (HCC) are proxied by a loan-to-income style ratio (mortgage loans to GDP), which may not capture all dimensions of credit access (e.g., home equity withdrawals are rare in China). The threshold framework identifies a single threshold per region; more complex nonlinearities or multiple thresholds are not modeled. Data used are subject to availability restrictions and are not publicly available without permission, which limits replication. The study does not directly model the education–housing linkage; its role is discussed and left for future research.