The Impact of Housing Support Expenditure on Urban Residents’ Consumption—Evidence from China

Abstract

Consumption plays an important role in economic growth and sustainable economic development. The Chinese government emphasizes the theme of promoting high-quality development. This aim has led to the implementation of strategies to expand domestic demand through a deep structural reform in the supply chain, enhance the endogenous power and reliability of the large domestic cycle, improve the quality and level of the international cycle, and accelerate the construction of a modern economic system. Based on the clarification of the consumption effect of housing support expenditure, this paper uses a combination of theoretical and empirical analysis to clarify the relationship between housing support expenditure and urban residents’ consumption. This is carried out to seek a breakthrough point for raising the total level of urban residents’ consumption expenditure in China from the level of housing support. This study revealed that government housing support has a certain degree of influence on the total level of urban residents’ consumption expenditure and that the government’s support of housing support is conducive to improving the total level of urban residents’ consumption expenditure. Also, if the housing price is not controlled and adjusted, the effect of government housing support on urban residents’ consumption expenditure will be greatly reduced, which will affect the lives of urban residents in the long run. Finally, suggestions and countermeasures for improving the overall level of urban residents’ consumption and the sustainable development of the social economy in China are provided.

1. Introduction

The restructuring of the world economy and the profound changes in the connotations and conditions of China’s strategic opportunity period have posed a huge challenge to China’s export-oriented and investment-led economic growth model. Changes in the economic development model and its implementation are inevitable. Shifting the economic growth trend from comparative advantage to competitive advantage and from being investment- and export-led to a new open economic model fueled by domestic demand is a major policy direction for China’s current and future development. Also, accelerating the establishment of a long-term mechanism for expanding consumer demand and releasing the consumption potential of the populace have become a realistic choice for China. International research shows that economic growth in large countries is mainly supported by domestic demand, with consumer demand having the largest share [1]. In contrast, China’s domestic demand has always accounted for a lower share of its total demand compared to other large economies over the same period [2]. As a result, the expansion of domestic demand to a high level to promote the strategic adjustment of the economic structure is at the heart of the Chinese government. This has shifted the economic growth momentum from being investment- and export-driven to being domestic-demand-driven. Regarding this agenda, much emphasis is laid on the consumer demand-driven aspect, which is a systematic project involving many areas. One of the key variables affecting residents’ consumption ability is housing [3]. China’s housing implementation or supply policy has gone through a process from complete welfare housing allocation to a combination of market and government provision. Three main phases can be traced in China’s history. The first is the period before 1978 known as the full welfare housing policy. The state and enterprises construct and provide housing at nearly free distribution to urban workers. This policy became a burden to the government and enterprises, and the distribution system also caused a series of problems [4]. The second happened from 1978 to 2007, known as residential commercialization. The central government delegated some decision-making power to local governments to boost their participation. The Asian crisis in 1997/98 impacted the Chinese economy, causing insufficient domestic consumption [4]. The policy to expand household consumption, stimulate domestic demand, and ultimately maintain stable economic growth became important at that time. This disrupted the physical housing distribution system to establish a market-oriented urban housing supply system called affordable housing. This housing policy increased housing investment in urban areas from 27.085 billion yuan in 1998 to 82.093 billion yuan in 2007 [5]. The final phase is the comprehensive structural adjustment that started in 2007 and is still in place to date. Before this, problems such as rapid increases in urban commodity housing prices emerged. This was partially due to the mismatch between rapidly rising housing prices and people’s income, and the imperfect management system for affordable housing, giving rise to the current reform. Under the guidance of the central government, local governments in China have standardized their housing support policies, adjusted the real estate market, and vigorously developed low-cost housing. The vigorous promotion of low-rent housing has played a huge role in solving the housing problem for low-income people in the urban registered residence population. Also, incorporating public rental housing, which was introduced in 2010, into China’s urban housing support system strengthened it. Again, the participation of the private sector in construction has been a breakthrough for China’s housing support system. This has helped to solve the funding problem for 36 million affordable housing units and significantly increased expenditure from 72.597 billion yuan in 2009 to 710.608 billion yuan in 2020 [5]. The investments and construction of affordable houses from 1997 to 2010 were worth 906.397 billion Yuan and 15,921.5 billion yuan [5]. Also, from the year 2009 to the year 2020, the Chinese central government and local governments contributed 489.694 billion Yuan and 5,546,989 billion Yuan, respectively, towards housing support expenditure, constituting 19.58% and 38.79% of the total expenditure, respectively [5].

Ensuring the basic housing needs of residents is the responsibility of the Chinese government at all levels. Therefore, measures have been put in place to ensure an effective and efficient financial structure for these projects. The expenditure on basic housing support for residents belongs to the category of public financial expenditure [4], meaning that its implementation is based on public financial expenditures. The housing subsidies for employees of administrative institutions are mostly arranged and resolved by the public financial budget. The housing provident fund of in-service employees in administrative institutions is mainly borne by public finance. Lands are provided through allocation and are exempted from land transfer income [4]. Construction administrative fees and operations are reduced by half. The cost of infrastructure for housing projects outside the community is borne by the government, with significant support from public finance. Tax incentives and preferential treatment are implemented for the rental income of low-rent housing [1].

The beneficiaries of housing support expenditure, also known as the demand side, are those who are eligible to rent or purchase affordable housing. The system was created to help all people, including the poor, to obtain accommodation. For instance, government agencies, institutions, and enterprise units, as well as employees at various levels in China, have entered the coverage of the housing provident fund system. As a specialized housing support fund, it can be used by employees to improve their living standards through housing provident fund loans or withdrawals. The mechanism through which the housing expenditure leads to consumption has been described in Figure 1.

Taking inferences from the housing situation in China, the main objective of this paper is to examine and clarify the relationship and mechanism between housing support-related expenditures and urban residents’ consumption behavior in China. In terms of the specific approaches utilized to carry out this paper’s objectives, the factors affecting the total consumption expenditure of urban residents are first summarized and analyzed, and the main factors are incorporated into the model for empirical analysis. Secondly, in the regression analysis, this paper first establishes a static panel regression, and the regression results show that, at present, the increase in guaranteed housing construction and government financial housing guaranteed expenditure has a positive impact on the expansion of urban residents’ consumption expenditure. Thirdly, considering the residents’ consumption habits and the endogeneity of the model, this paper incorporates the lagged period of total consumption expenditure into the model to construct a dynamic panel model. The panel model is re-estimated by using the systematic GMM method. The estimation results of this paper show that the signs of the core explanatory variables and other variables do not change significantly, which also shows the robustness of the regression results of this paper. To enhance the reliability of the conclusions in this section, the robustness of the model is tested in various ways in this paper.

To achieve the objective of this study after it is taken through the various processes, the following theoretical, empirical, and practical contributions will be made to the body of literature and assist with future policy implementation. First, research on housing support-related expenditures and consumer behavior is limited in the Chinese context, and this study intends to close this gap. Second, by closing this gap with the new findings, a theoretical foundation for introducing relevant industrial policies in China and the use of economic leverage to promote the healthy development of the real estate industry and the stable growth of residents’ consumption will be laid. Though the Chinese housing industry keeps expanding each year, studies regarding this area are limited, and the theories and methodologies used for the subject are few. This calls for new studies to strengthen the existing ones in order to improve the housing industry and consumption in China, and this study contributes to that. Finally, the practical implications suggested in this study will provide a reference for the central and local government policy systems. Also, private investors and estate owners can benefit from this study through its analysis of the relationship between these relevant housing factors.

2. Literature Review

Income is a crucial factor that affects consumption among both urban and rural residents. There are some classical theories of the factors which influence consumption, such as Keynes’s absolute income hypothesis, Duesenberry’s relative income hypothesis, Friedman’s permanent income hypothesis, etc. In Keynes’s absolute income hypothesis theory, consumption is a function of absolute income that depends mainly on consumers’ current disposable income, and this theory refers to the primitive, myopic consumer [6]. Duesenberry’s relative income hypothesis considers that consumers determine their current consumption level based on their past consumption habits and the consumption of others around them. This theory is the posterior description of climbing consumer behavior, and it highlights the consumption habits of consumers in considering consumption [7]. In Friedman’s permanent income hypothesis, budget constraint is also intertemporal, which leads to an additional externality setting: there is no liquidity constraint, i.e., consumers must have an unconstrained borrowing capacity to cover two income shortfalls when their current consumption exceeds their current income. In addition, consumers take into account their expected future income level, i.e., the uncertainty setting [8]. The life cycle hypothesis of Modigliani et al. [9] and the permanent income hypothesis of Friedman consider that consumers are rational and rationalize their lifetime income and smoothen their consumption according to the utility maximization principle. Leland’s model of precautionary motivated saving defines precautionary saving as additional savings caused by uncertain future income [10]. In addition to some factors of consumers’ income, subsequent scholars have pointed out that some objective factors such as economic cycles [11,12], environmental factors [13,14], consumers’ psychology [15,16], etc., can also have an impact on their consumption.

The impact of housing on resident consumption is a hot issue that has attracted the attention of scholars in recent years. A large amount of the literature has studied housing and its residents’ consumption [17,18]. For example, the relationship between housing consumption and residential consumption may be affected by liquidity [19]. The referenced study further argued that liquidity constraints lead to the failure of the life-cycle model of consumption, and credit constraints affect residential consumption expenditure, making consumption more sensitive to contemporaneous income. With income improvement, the increasing house prices and rents pressurize housing affordability for residents [20], and may put a squeeze on consumption, while housing allowance policies can reduce the pressure on residents to pay for housing [21]. The increased urban living cost brought about by high housing prices is an important factor that discourages rural residents from migrating to cities [22,23]. Thus, it is clear that government support through some policies in appropriate cases can improve the affordability of housing for residents and reduce their housing pressure, which in turn serves the purpose of expanding consumption. Expanding housing support by providing subsidized housing or providing rental subsidies for low- and middle-income people can have an important impact on residents’ consumption demand and consumption structure [24,25]. Public housing (PH) programs or low-income and affordable housing are a large part of housing policies in both developed and less-developed countries [26,27]. In contrast to these, the housing support supply is more recognized by scholars for reducing the housing burden of residents, improving the quality of life of low- and middle-income people, and increasing daily consumption [28]. While the government intervenes in the commercial housing market, it should also strengthen the management of public housing [29] to facilitate better living conditions for low- and middle-income residents.

The two main modes of housing support provision, housing subsidies in kind and housing cash subsidies, have also become a hot topic of debate among scholars [30,31], and the government provision of public housing may have an impact on the price of commercial housing [32]. Public rental housing reduces the burden of housing costs for various household types [33], and the improved stability, independence, and support provided to low-income households [34], as well as the provision of government housing, contribute significantly to an increase in household income [35]. Cash subsidies for housing, another form of housing support, have a greater impact on residents [36]. One example of this is the housing lottery system in Mumbai, India, where winners can choose to occupy or rent their homes; after three–five years, winners have better quality housing and report higher incomes compared to non-winners [37]. Another instance is the United States’ Housing Choice Voucher program which helps to reduce the rental burden of low-income households [38]. Government housing support policies have a positive impact on the health of occupants [39]. Regarding the issue of housing support and residents’ consumption, Tiwari and Hasegawa [40] pointed out that the consumption effect of public housing policies in Tokyo, Japan, was significant, with public housing participants increasing their consumption of housing goods by an average of 33% and consumption of non-housing goods by an average of 25% compared to before they participated in public housing programs. Providing social or public housing will impact the income of low- and middle-income people significantly. Additionally, providing social or public housing will have a significant impact on the income distribution of low- and middle-income people, who will spend less on housing and more on other household goods, as well as more on services such as entertainment [41].

It is a consensus among scholars that the impact of housing support on residents’ consumption has been recognized, and that the government’s provision of in-kind housing supply or cash housing subsidies can reduce residents’ housing payment pressure and thus expand consumption. However, scholars are currently more in the stage of theoretical discussion than data verification. Moreover, the actual effects of physical housing supply and cash housing subsidies on residents’ consumption are less discussed. This paper intends to fill in the above-mentioned gaps through a study based on relevant Chinese data.

3. Theoretical Analysis and Hypothesis Formulation

3.1. Analysis from the Perspective of Guaranteed Housing Supply

3.1.1. Analysis of the Impact of Affordable Housing on the Total Consumption of Urban Residents

Affordable housing is a kind of housing support supply method. The construction of affordable housing has become part of the national housing support program, mainly under the leadership of local governments of China [42]. It is carried out by property developers or units that raise funds to build houses. Its sales target low-income families in cities, and the sales prices are significantly lower than those of commercial houses [4]. Affordable housing embodies the nature of social support and has the characteristics of economy and applicability. The economy aspect is mainly reflected in the fact that the sales prices are significantly lower than those of commercial housing in the same period, which is more suitable for the housing and economic needs of low and middle-income families [42]. The applicability aspect is mainly reflected in the fact that affordable housing has the same construction standard and usage function as commercial housing, which can meet the usage needs of the residents [43]. Affordable housing is a basic housing system implemented in China, and the designated objectives and implementation of the system reflect the characteristics of housing subsidies. As such, it influences the allocation of resources in many ways. It is worth mentioning that, in terms of sales price, affordable housing generally has a maximum price, which is lower than the market average and is figuratively referred to as the “ceiling price” [4]. This has an important impact on consumers, producers, and the general welfare of society.

As shown in Figure 2, when there is no government intervention in the real estate market, the real estate supply curve (S) intersects with the demand curve (D) at point E. At this point, the real estate market reaches equilibrium (Pe). Pe is the equilibrium price, and the average measure is Qe. For the supply of affordable housing, the government sets a maximum price limit. Assuming that the price of affordable housing set by the government at a given time is Pm (That is, Pm < Pe), the demand for real estate will be Qd. However, because the price limit is lower than the market price, real estate developers are not willing to supply at Qs, while the consumer demand for affordable housing is much higher than the supply (that is, Qd > Qs). The maximum price of affordable housing makes the transaction price fall while reducing the volume of transactions in the market, leading to excess demand in the market (thus; Qd-Qs = deficit). The changes in market supply and demand and the market price will change the welfare status of residents at the equilibrium price level. As shown in Figure 1, compared to equilibrium, the consumer surplus increases by area A but decreases by area B. The amount of change in the consumer surplus is surface A-B. This indicates that consumers who bought affordable housing increased their surplus, but that consumers who could not buy affordable housing suffered a loss. Similarly, this paper obtains the changes in producer surplus. The surplus of the producer who carries out production incurs a loss at area A, while the producer who withdraws loses all the surplus, and the change in producer surplus is surface (A + C). Collectively, the market surplus produces a change in two parts: one part is the transfer of surplus, i.e., the surplus of area A is transferred from producers to consumers; the other part is the reduction in surplus, i.e., area B and area C. That is, the maximum price policy for affordable housing leads to a net welfare loss of area (B + C). As for consumers, area A is generally larger than B. Therefore, the overall maximum price of affordable housing is beneficial to consumers.

3.1.2. Analysis of the Impact of Low-Rent Housing on the Total Consumption of Urban Residents

Low-rent housing is a welfare system provided by the government to families who meet the support criteria after the setting of certain standards [44]. Generally, only families in the urban low- and middle-income groups who have housing difficulties can enjoy this benefit. China issued the Measures for the Administration of Low-Rent Housing for Urban Low-Income Families in 2004 [4]. These measures dictated that China’s low-rent housing system is based on the payment of rent to the insured group, supplemented by the distribution of housing in kind and partial rent reductions. The impact of low-cost housing, which is available in the form of both in-kind allocation and rent subsidies, on residents’ consumption can be analyzed in two ways:

The first is the impact of the in-kind rent allocation method on urban low- and middle-income households. As shown in Figure 3, the income of low- and middle-income households is mainly spent on two commodities, one is housing consumption, i.e., the X-axis, and the other is the non-housing consumption of general goods, i.e., the Y-axis. The consumption budget line of low- and middle-income households is AB and the undifferentiated curve is U₁, which is tangential to point O₁. At this point, O₁ is the equilibrium point of consumption, and the optimal combination of goods maximizes consumers’ utility. At the equilibrium point O₁, consumers’ consumption of housing X and other general goods Y are x₁ and y₁, respectively. After receiving in-kind rent from the government, the consumption budget line moves to the right along the X-axis to line AC, which is tangential to the undifferentiated curve U₂ at the equilibrium point O₂, and the equilibrium consumption quantities are x₂ and y₂, respectively. The equilibrium point O₂ is higher than the equilibrium point O₁, and the consumption of housing x₂ and the total utility level of the household are at O₂. The level of efficiency of the household increases. The increase in the consumption of general goods Y is uncertain, and it depends on whether the income effect of in-kind rent allocation on the consumption of general goods is greater than the substitution effect of housing consumption on the consumption of general goods. In general, after strictly controlling the criteria of in-kind rent allocation, even if the in-kind rent allocation received by low- and middle-income households increases, their consumption expenditure on housing will not increase. The substitution effect of housing consumption on other general goods will not exist, and the consumption of other general goods by this group of households will increase due to the existence of an income effect after receiving in-kind rent allocation.

In fact, for the low- and middle-income earners, the in-kind allocation is to increase their housing consumption while keeping the original consumption basket of this group of consumers unchanged. At this point, as long as consumers can ensure that their housing needs are met, they can enjoy the increase in income effect brought about by the increase in real income. This brings about an overall increase in utility, as shown in Figure 2, when the consumer’s utility is moved from the line U₁ to the U₂ line, thus generating an expansion in consumption.

The second is the effect of the form of rent subsidy on low- and middle-income households. After receiving the rent subsidy for low- and middle-income households, this group will have their welfare level and consumption level enhanced to some extent due to the income effect. As shown in Figure 4, before receiving the rent subsidy, the consumption budget line of low- and middle-income households is AC and the utility curve is U₁. At this time, the consumption budget line AC and the utility curve U₁ intersect at the equilibrium point O₁, and the equilibrium consumption quantities of housing consumption and general goods consumption of low- and middle-income households at the equilibrium point O₁ are x₁ and y₂. After households receive rent subsidies, the consumption budget line of low- and middle-income households shifts up to the right. The new equilibrium point is O₂, and the equilibrium housing consumption and general goods consumption under the new equilibrium condition are x₂ and y₂. The equilibrium point O₂ of the low- and middle-income group after receiving the subsidy is higher than the equilibrium point O₁ before receiving the subsidy. In contrast to in-kind rent allocation, rental subsidies for housing are direct monetary subsidies that, first, increase the consumption budget of low- and middle-income households, which can be shown as a rightward shift of the consumption budget line in Figure 3, thus causing an increase in the utility of low- and middle-income people and eventually generating an expansion of consumption.

3.2. Analysis from a Public Expenditure Perspective

Housing support spending is an important component of public spending by the Chinese government [1]. This indicates that housing support expenditure has a significant consumption effect by itself. The consumption effect of government public expenditure means that the government, through expanding expenditure, especially livelihood public expenditure, will directly and indirectly trigger and expand consumption demand, and the market will thus accommodate a larger supply of consumption materials [43,45]. The increase in the supply of consumption materials will mean a further expansion of market capacity, which will have a positive effect on the coordination of market supply and demand. This will stimulate the growth of consumer goods production and form a vicious cycle and promote economic growth. Government public expenditure can be divided into different types according to the policy instruments [4] used. In the case of housing support, it is an important part of the social support system and is more often subsumed under government transfer expenditure. The increase in governmental transfer expenditure is generally considered to promote the consumption by the population and to have an induced effect on the consumption by the population. In particular, expenditures in the social support system play an important role in enhancing residents’ consumption confidence. From the above analysis of the factors influencing consumption, it is clear that the expansion of consumption depends not only on consumers’ current income but also on their expectations of the future. These expectations can be divided into two aspects: the expectation of income growth and the expectation of uncertainty. The latter refers mainly to social support. If a society has established a more complete social support system, individuals have a safety net against unexpected risks. People do not have to worry too much about future pensions, medical care, unemployment, and other problems, and increase current savings so that current consumption expenditures will increase. At the same time, for housing support spending, there are some connotations of government investment spending. For urban low-income households, housing support also changes their consumption functions and expected constraints, expanding their consumption expenditures and improving their living standards and quality of life. Therefore, the increase in government housing support expenditure can have a favorable impact on the increase in urban residents’ total consumption.

Synthesizing the above analysis, this paper determines that an increase in housing support expenditure will have a favorable impact on urban residents’ consumption, both at the level of guaranteed housing supply and at the level of increased public expenditure. Therefore, the hypothesis below is formulated.

Hypothesis 1. 

Government spending on housing support is conducive to an increase in the total consumption of urban residents, and the more is spent on housing support, the more is the consumption spending of urban residents.

4. Empirical Analysis

4.1. Data Sources and Variable Descriptions

From the above, it is clear that residents’ consumption behavior is the result of the combined effect of many factors. From the macroscopic point of view, there are mainly price level, industrial structure, interest rate level, social support level, public policies, etc.; from the microscopic point of view, residents’ income and wealth, residents’ consumption habits, household demographic structure, expectations, etc. are the main factors that influence the consumption behavior and consumption level of urban residents in China. However, many of the above factors cannot be fully quantified and analyzed due to the constraint of reality; therefore, in the specific analysis, according to the research results of scholars, some unquantifiable factors are dealt with in this paper.

4.1.1. Data Sources

In this chapter, to verify the study’s hypothesis and inference from previous works on the impact of housing support expenditure on the total consumption of urban residents, relevant data are selected for empirical analysis. First, regarding the selection of the sample cross-section, Shanghai abolished the Affordable Housing Development Center in 2002 and then had subsidized housing until 2009. Its data are incomplete and discontinuous, so this paper excludes Shanghai from the sample. Furthermore, the Tibet Autonomous Region, Hong Kong, Taiwan, and Macau have more missing data, so it they also excluded. Finally, this paper selects 29 provinces (municipalities and autonomous regions) in China, except Shanghai, Tibet, Hong Kong, Macao, and Taiwan, as the sample space; secondly, for the detailed data statistics of housing support expenditure in China, the corresponding data are only available in 2010.

Prior to that, affordable housing was almost the only form of housing support in most areas of China during 1998–2010 due to its very important role in China’s housing support system. In addition, the Chinese government adopted various approaches to solve the housing problems of low- and middle-income households. After 2010, the Chinese government’s approach to housing support has gradually transitioned from mainly guaranteed housing supply to mainly financial subsidies, i.e., from “compensating for bricks and mortar” to “compensating for people”. The statistics for affordable housing are more detailed in the relevant database in China. Therefore, this paper has carried out the following treatment. Firstly, this paper selects the data on affordable housing from 1999 to 2009 for analysis. Considering the timeliness of the data, and at the same time, to facilitate the comparison of the two types of subsidies, this paper also selects the data of government housing expenditure from 2010 to 2020 for regression. The details are as follows:

This paper selects panel data from 29 provinces (municipalities) and autonomous regions in mainland China from 1999 to 2020, and the data of each variable are obtained from the website of the National Bureau of Statistics, China Statistical Yearbook, China Financial Yearbook, China Population and Employment Statistical Yearbook, Wande Information (Wind Information) database, EPS Global Statistical Platform, the Sixth Population Census, etc. To eliminate the influence of price factors in each year, the actual values of the selected indicators are obtained by taking 1998 as the base period and the data of each indicator (except the proportional data) are taken as the natural logarithm to eliminate the influence of heteroskedasticity.

4.1.2. Variable Description

Detailed variable descriptions are as follows:

(1)

Dependent variable: In this section, the main investigation is the impact of government housing support expenditure on the overall consumption level of urban residents. Therefore, the overall consumption level of urban residents must be the main dependent variable. It is also necessary to choose the variables that are more representative of the overall consumption level of urban residents as its proxy variables. In previous studies, the variables that measure the overall consumption level of residents are mainly total per capita household consumption expenditure, resident consumption rate, resident consumption-to-GDP ratio, the average propensity to consume, and marginal propensity to consume. In the analysis of this paper, consumption per capita household expenditure (consumption) is used as the explanatory variable to visualize the interaction between the variables.

(2)

Investigation variables: The main investigation variable of this paper is government housing support expenditure, which is divided into two time periods; including the housing support expenditure data from 1999 to 2009 and the data related to affordable housing, as well as the data related to government housing support financial expenditure in each region from 2010 to 2020. As can be seen, this paper applies the meaning of government housing expenditure in a broad sense, i.e., unless otherwise specified in the following, the amount of affordable housing investment, as the dependent variable, and government housing expenditure, as the dependent variable, are collectively referred to as housing expenditure in this paper. The use of these two kinds of data for examination helps this paper to analyze the problem more carefully. The corresponding indicator is housing support expenditure per urban resident (hse—housing support expenditure), which is obtained by dividing the total housing support expenditure of each province and city by the number of the urban population in that province and city.

(3)

Control variables: The control variables in this paper are residents’ income, urban household dependency ratio, the proportion of the secondary industry, the proportion of the tertiary industry, and the housing price. According to the consumption theory, income is the main factor which affects residents’ consumption; therefore, urban per capita disposable income (income) is selected as the indicator of residents’ income; the ratio of the minor population (up to 14 years old) and the old population (65 years old and above) to the working age population (15–64 years old) in urban households is selected as the indicator of urban household dependency ratio (bring); the ratio of the secondary industry (industry) is the ratio of secondary industry GDP to regional GDP; the ratio of the tertiary industry (service) is the ratio of tertiary industry GDP to regional GDP. With the fluctuation in housing prices in recent years, the price of commercial housing has also become an important variable which affects urban residents’ consumption; therefore, the average prices of commercial housing (chp—commercial housing prices) in each province and city are selected as an indicator of housing price in this paper. The results of specific statistical tests for each variable are shown in the following tables (

Table 1. Definitions and statistical properties of the main variables (1999–2009).

	Variable Name	Variable Meaning	Unit	Average Value	Standard Deviation	Minimum Value	Maximum Value	Number of Observations
Dependent variable	lnconsume	Real value of per capita cash consumption expenditure of urban households	Yuan/person	8.780271	0.3177255	8.148623	9.646826	319
Examining variables	lnhse	Actual value of affordable housing investment per capita (1999–2009)	Yuan/person	4.698569	0.8399499	1.595801	6.923169	319
Control variables	lnincome	Real value of per capita disposable income of urban residents	Yuan/person	9.054412	0.3485144	8.372239	10.0485	319
	lnchp	Actual value of average sales price of commercial housing units	Yuan/per square meter	7.618805	0.4365458	6.709304	9.393124	319
	service	Tertiary industry share	%	0.3849921	0.0629665	0.286	0.755	319
	industry	Percentage of secondary industry	%	0.4636041	0.0750656	0.1976	0.615	319
	bring	Urban family dependency ratio	%	0.3435917	0.043563	0.2415781	0.4792317	319

and

Table 2. Definitions and statistical properties of the main variables (2010–2020).

	Variable Name	Variable Meaning	Unit	Average Value	Standard Deviation	Minimum Value	Maximum Value	Number of Observations
Dependent variable	lnconsume	Real value of per capita cash consumption expenditure of urban households	Yuan/person	9.864469	0.303365	9.216152	10.74415	319
Examining the variables	lnhse	Real value of financial expenditure on housing support per urban resident	Yuan/person	6.483025	0.652691	4.113154	8.30304	319
Control variables	lnincome	Real value of per capita disposable income of urban residents	Yuan/person	10.2391	0.3288864	9.521038	11.23323	319
	lnchp	Actual value of average sales price of commercial housing units	Yuan/per square meter	8.752894	0.4518343	8.008033	10.53649	319
	service	Tertiary industry share	%	0.476531	0.0875437	0.3246	0.839	319
	industry	Percentage of secondary industry	%	0.4209928	0.0825777	0.158	0.62	319
	bring	Urban family dependency ratio	%	0.3242129	0.0524292	0.2013459	0.4635434	319

Data sources: China Statistical Yearbook, China Financial Yearbook, China Population and Employment Statistical Yearbook, Wande Information (Wind Information) database, EPS data platform, and data from the Sixth Population Census in relevant years.

).

4.2. Sample Data Stationarity Test

To ensure the reliability of the test results, the ADF-Fisher method [46], LLC method [47], and IPS method [48] were selected to test the smoothness of each variable in this paper. In this section, unit root tests are performed on the IV and DV.

Table 3. Results of unit root test for each variable.

1999–2019 Data
Test Methods/ Variables		ADF-Fisher Test (Chi-Square Statistic)	LLC Inspection (t-Star)	IPS Inspection W (t-Bar)	Whether to Include the Unit Root
lnconsume	Level Value	59.4797 (0.4215)	−2.27820 (0.0114)	0.821 (0.794)	Yes
	First order differential	208.7858 (0.0000)	−7.51615 (0.0000)	−4.194 (0.000)	No
lnhse	Level Value	125.3062 (0.0000)	1.53816 (0.9380)	1.384 (0.917)	Yes
	First order differential	196.4498 (0.0000)	−3.55169 (0.0002)	−3.599 (0.000)	No
lnchp	Level value	49.1878 (0.7884)	−2.77352 (0.0028)	0.595 (0.724)	Yes
	First order differential	181.3194 (0.0000)	−5.85586 (0.0000)	−3.748 (0.000)	No
lnincome	Level value	67.7208 (0.1794)	−7.08551 (0.0000)	−0.754 (0.225)	Yes
	First order differential	211.9550 (0.0000)	−7.17043 (0.0000)	−3.733 (0.000)	No
Data for 2010–2020
lnconsume	Level Value	228.1153 (0.0000)	−11.1735 (0.0000)	−5.9295 (0.0000)	Smooth and stable
	First order differential
lnhse	Level Value	370.8961 (0.0000)	−4.9686 (0.0000)	−6.6545 (0.0000)	Smooth and stable
	First order differential		−13.0158 (0.0000)
lnchp	Level Value	34.3678 (0.9943)	0.4834 (0.6856)	4.9931 (1.0000)	non-stationary
	First order differential	185.3725 (0.0000)		−4.8698 (0.0000)	Smooth and stable
lnincome	Level value	734.8359 (0.0000)	−9.7697 (0.0000)	−9.0816 (0.0000)	Smooth and stable
	First order differential

Note: The statistics of each method are outside the parentheses, and the corresponding p values are in parentheses; the results of each unit root test method with intercept term are reported here.

shows the results of the smoothness tests for the IV and DV:

Three unit root tests for the panel data were selected for this paper. Overall, the level values of some variables are not smooth, but all are smooth at their first-order difference values. It can be concluded that the differential values of this panel data are smooth.

4.3. Model Construction and Analysis of Empirical Results

4.3.1. Model Construction

In order to estimate the direction and magnitude of the impact of government housing support spending on residents’ consumption, the following benchmark panel model is constructed:

$$\ln {consume}_{it}={\beta }_0+{\beta }_1\ln {hse}_{it}+\gamma X_{it}+{\alpha }_i+{\alpha }_t+{\epsilon }_{it}$$

where the subscripts i and t denote the ith province (city) and year t, respectively. lnconsume_it represents the total per capita real consumption expenditure of urban households taken as a logarithm, lnhse_it is the main observed variable, and this variable uses the data of per capita investment in affordable housing in each province from 1999 to 2009 and, in 2010–2020, it uses the data of provincial financial spending.

For housing expenditure data, in this paper, the logarithm is taken for 1999–2009 or 2010–2020.

$X_{it}=({\ln income}_{it},{\ln chp}_{it},{service}_{it}{,industry}_{it},{bring}_{it}$) is a vector of other variables, mainly containing the following variables: $\ln income$ represents the logarithm of the real disposable income per capita of urban residents; $\ln chp$ represents the logarithm of the average sales price of commodity houses; ${industry}_{it}$ and ${service}_{it}$ represents the share of the secondary and tertiary industries in each province, respectively; ${bring}_{it}$ represents the urban household dependency ratio whose formula is (number of people under 14 and 14 years old + number of people over 65 and 65 years old/number of people aged 15–64); α_i represents the individual fixed effects, α_t represents year fixed effects, and ε_it represents random disturbance terms. Here, the coefficient β₁, which is the main concern of this paper, is significantly positive if β₁ indicates that the increase in government housing support expenditure has a positive effect on total residential consumption expenditure, and is negative if the opposite is true. Meanwhile, this paper also focuses on the coefficients of other variables.

4.3.2. Analysis of Empirical Results

In general, for static panel data, there are usually three estimation methods: mixed models (P-OLS), random effects models, and fixed effects models [49]. To determine the static benchmark model regression method in this section, the following selection is made in this paper. First, the F-test is conducted in this paper, and it is found that the value of the F-statistic is less than 0.05, indicating that the fixed-effects model is better than the mixed-effects model. Second, the Hausman test [50] is applied in this paper to select between the fixed-effects model and the random-effects model, and the results show that the fixed-effects model is the best. Therefore, the fixed-effects model is chosen as the baseline model estimation.

In the case of examining the impact of government housing guarantee expenditure on the total consumption expenditure of urban residents, the dependent variable is the actual value of the total per capita cash consumption expenditure of urban residents. As shown in

Table 4. Regression results of the impact of government housing support expenditure on total per capita consumption expenditure of urban residents.

Variables	1999–2009		2010–2020
	Model 1-Fe	Model 2-Fe	Model 3-Fe	Model 4-Fe
lnhse	0.027 * (0.017)	0.152 ** (0.063)	0.021 *** (0.007)	0.023 *** (0.008)
bring	0.246 (0.216)	0.526 ** (0.254)		−0.376 *** (0.096)
lnincome	0.847 *** (0.018)	0.857 *** (0.018)	0.859 *** (0.010)	0.869 *** (0.027)
lnchp	−0.012 (0.016)	0.049 (0.033)		−0.045 * (0.027)
lnhse*bring	−0.080 * (0.048)	−0.142 ** (0.057)
lnhse*lnchp		−0.013 ** (0.006)
service	0.230 * (0.136)	0.150 (0.141)		0.627 *** (0.219)
industry	0.284 ** (0.122)	0.200 (0.128)		0.247 (0.206)
Intercept term	0.900 *** (0.101)	0.9010 *** (0.101)	0.938 *** (0.083)	0.940 *** (0.204)
Number of samples	319	319	319	319
Within-R2	0.787	0.767	0.776	0.779

Note: ***, **, and * represent significance at 1%, 5%, and 10% levels, respectively, with standard errors in parentheses.

, to examine the impact of government housing guarantee expenditure on the total consumption expenditure of urban residents, this paper uses the regression analysis of affordable housing statistics from 1999 to 2009 and government housing guarantee expenditure statistics from 2010 to 2020, respectively; the regression results of these two parts of the data are also compared.

As shown in Table 4, Models 1 and 2 are the regression results of the data from 1999 to 2009, and the core explanatory variable is the amount of urban per capita investment in affordable housing. Models 3 and 4 are the regression results using data from 2010 to 2020, and the core explanatory variable is the amount of urban per capita expenditure on housing support. The regressions’ results reveal the following points:

First, from the overall perspective, models 1 to 4 show that the coefficients of disposable income and housing support are positive and mostly significant in both periods, which is in line with the theory and reality, i.e., income is the most important variable in determining consumption. In absolute terms, income is directly proportional to consumption. Second, from the core explanatory variables, the coefficients of housing support expenditure are positive. This indicates that, in general, the construction of subsidized housing and the increase in housing support expenditure have a significant positive impact on the increase in urban residents’ consumption expenditure, i.e., it is conducive to the expansion of urban residents’ overall consumption support. From the results in Table 4, it can be seen that the expenditure data on affordable housing construction from 1999 to 2009 and the expenditure data on fiscal housing support from 2010 to 2020 both present the same results, that is, the core explanatory variable, hse, has a positive and significant impact on the dependent variable, consumption. Using the 1999–2009 data regression results for Model 1-Fe and Model 2-Fe, it can be seen from Model 1-Fe that, for every 1% increase in government affordable housing expenditure, urban residents’ consumption expenditure increases by 0.027%, which is significant at the 10% level; from Model 2-Fe, it can be seen that, for every 1% increase in government affordable housing expenditure, urban residents’ consumption expenditure increases by 0.152%, which is significant at the 5% level. Using the data from 2010 to 2020, the regression results are shown in Model 3-Fe and Model 4-Fe. From Model 3-Fe, it can be seen that, for every 1% increase in government financial housing support expenditure, urban residents’ consumption expenditure increases by 0.021%, which is significant at the 1% level; from Model 4-Fe, it can be seen that, for every 1% increase in government fiscal housing support expenditure, urban residents’ consumption expenditure increases by 0.023%, which is significant at the 1% level. This indicates that the government’s expansion of support spending and the co-existence of multiple support methods have a greater impact on the increase in urban residents’ consumption expenditure than direct support housing construction.

The above results can be analyzed and explained as follows: from the current situation, it can be stated that the government’s increase in housing support expenditure is conducive to promoting urban residents’ consumption expenditure. In the current context of high housing prices, urban residents are under more pressure to buy houses. Residents who do not have houses but who have the expectation to buy houses, or households who do not have houses and cannot resist risks such as accidents, will choose to cut down their expenditures to save, thus causing a decrease in consumption expenditures. If the government plays a significant role in this area, so that residents’ housing problems are solved or their living pressure is reduced, residents will increase their spending on current consumption instead of compressing their current cost of living for a rainy day.

The above analysis is a regression using a static panel. However, there may be correlations between the variables, specifically the explanatory variables and the explained variables. In reality, the current consumption of residents has a “ratchet effect,” i.e., the current consumption is influenced by the consumption of the previous period. In this paper, after adding a one-period lag of consumption, the explanatory variables are more correlated with the explanatory variables, which means that there is a certain endogeneity problem in the model, so it is necessary to establish a dynamic model for re-estimation. In the following analysis, the lagged period of consumption is included in the model as the independent variable, a dynamic panel model is constructed, and a new estimation method is applied to estimate the model and perform robustness tests. The new model form is as follows:

$$\ln {consume}_{it}={\beta }_0+{\beta }_1\ln {consume}_{it-1}+{\beta }_2\ln {hse}_{it}+\gamma X_{it}+{\alpha }_i+{\alpha }_t+{\epsilon }_{it}$$

There are estimation methods such as first-order difference GMM and systematic GMM which can be used when examining the relationship between variables in dynamic states, and the choice and analysis of the two methods are required for their specific application. However, when the endogeneity problem exists, the estimation results of the above two methods are not unbiased, i.e., the estimation results will have some bias. In this case, the more appropriate method is the instrumental variables method. In the above analysis, the model inevitably suffers from endogeneity when the lagged term of consumption is added. To obtain unbiased estimates, the endogeneity problem needs to be addressed. To address this type of problem, Arellano and Bond propose a first-order difference estimation method [51]. The problems of biased and non-consistent estimation results due to endogenous explanatory variables in the model are better solved by applying this method. Of course, the first-order difference estimation method has its shortcomings, one of which is the weak instrumental variable problem caused by the lack of instrumental variables, which was analyzed in detail by Blundell and Bond [52] in their 1998 research. Systematic moment estimation was first studied by Arellano and Bover [53], and was later supplemented and refined by Blundell and Bond [52]. The combination of horizontal and differential regression equations for estimation is the main method for the estimation of systematic moments. In the specific operation, the lag level is the instrumental variable for the first-order difference, which in turn serves as an instrumental variable for the level variable. Compared with the first-order difference GMM, the systematic GMM effectively improves the weak instrumental variable problem caused by insufficient instrumental variables in the latter estimation. It can use the difference and level values of endogenous explanatory variables as instrumental variables to overcome the endogeneity problem of explanatory variables without seeking other instrumental variables. Systematic GMM estimation has a better finite sample nature, which reduces the bias introduced in first-order difference GMM estimation to a greater extent and is more effective in overcoming the endogeneity problem of the model. Therefore, in this paper, the model is further estimated using the systematic GMM approach after adding the lagged terms of the explanatory variables. The estimation results are presented in

Table 5. Regression results of the impact of government housing support expenditure on total per capita consumption expenditure of urban residents (dynamic regression results).

	1999–2009		2010–2020
	Model 1-Gmm	Model 2-Gmm	Model 3-Gmm	Model 4-Gmm
L.lnconsume	0.648 *** (0.087)	0.596 *** (0.067)	0.848 *** (0.011)	0.636 ** (0.206)
lnincome	0.319 *** (0.076)	0.340 *** (0.056)		1.628 *** (0.212)
lnhse	0.006 * (0.004)	0.185 * (0.094)	0.023 * (0.014)	0.030 ** (0.015)
lnchp		−0.109 ** (0.050)		−0.16 ** (0.054)
bring		0.353 (0.392)
lnhsebring		−0.095 (0.083)
lnhselnsh		−0.019 ** (0.009)
service
industry
Number of samples	261	261	261	261
ar1p	0.000	0.000	0.000	0.000
ar2p	0.229	0.379	0.015	0.144
hansenp	0.255	0.286	0.250	0.225

Note: ***, **, and * represent significance at the 1%, 5%, and 10% levels, respectively, and robustness standard errors are in parentheses.

below.

The two main methods for discriminating the validity of the instrumental variable settings are the second-order serial correlation test AR(2) and the model overidentification constraint test. These two methods were proposed by Arellano and Bover [53] in 1995. The second-order serial correlation test is mainly used to discriminate whether there is serial correlation in the residual terms of the system moment estimates, in which the estimates for the target equation are valid only if there is no serial correlation in the second-order autoregression. The over-identification test, on the other hand, is mainly used to discriminate the overall validity of the instrumental variables in the system moment estimates, and there are generally two types of tests, the Sargan test and the Hansen test. The original hypothesis of both is that the instrumental variables are valid, and if rejected, it indicates that the instrumental variables are invalid, and vice versa. While the Sargen test is invalid under heteroskedasticity, the Hansen test is generally used under the GMM framework. In this paper, the Hansen test is applied. From the above results, this paper finds that the p-value of the models’ Hansen tests is not 0. Therefore, it can be seen that the instrumental variable chosen for the model is valid. The sign of the core explanatory variables and other variables in the regression results of the system GMM are generally consistent with the sign of the fixed effects regression in the static panel, which shows the robustness of the regression results. In addition, this paper also applies further robustness checks on the above results in various ways in the following sections. First, the paper analyzes the regression results under the dynamic panel, as follows.

As model 1 is validated with 1999–2009 data, this paper first analyzes the lagged period and current period income of residents’ consumption and housing support variables. The results show that the impact of both housing support-related variables on urban residents’ consumption is significantly positive at the 10% or 5% level; the impact of residents’ last period consumption on current period consumption is significantly positive, which indicates that urban residents’ consumption expenditure, in this paper, is more influenced by their consumption habits; the impact of income on residents is very significant, and the increase in income greatly promotes the increase in residents’ consumption expenditure. The same results are still presented when the data from 2010 to 2020 are used for verification.

Next, the paper gradually adds the interactions of house price with dependency ratio, housing support variables with dependency ratio, and housing support variables with commodity house price. The results show that the coefficients of housing support-related variables are still significantly positive for urban residents’ consumption expenditures. This indicates that the increase in guaranteed housing construction and housing support expenditures still has a facilitating effect on the expansion of urban residents’ consumption expenditures. Finally, this paper adds the industrial structure variable. The results show that the sign of each variable of the model does not change. Therefore, from the obtained results, it can be seen that, at the current stage of development in China, the construction of subsidized housing and the expansion of housing support expenditures have a positive impact on urban residents’ consumption, i.e., they are conducive to the expansion of urban residents’ consumption. From the results in Table 5, it can be seen that the expenditure data on affordable housing construction from 1999 to 2009 and the expenditure data on fiscal housing support from 2010 to 2020 both present the same results, that is, the core explanatory variable, hse, has a positive and significant impact on the dependent variable consumption. Using the 1999–2009 data regression results for Model 1—Gmm and Model 2—Gmm, it can be seen from Model 1—Gmm that, for every 1% increase in government affordable housing expenditure, urban residents’ consumption expenditure increases by 0.648%, which is significant at the 1% level; from Model 2—Gmm, it can be seen that, for every 1% increase in government affordable housing expenditure, urban residents’ consumption expenditure increases by 0.596%, which is significant at the 1% level. Using the data from 2010 to 2020, the regression results are shown in Model 3—Gmm and Model 4—Gmm. From Model 3—Gmm, it can be seen that, for every 1% increase in government financial housing support expenditure, urban residents’ consumption expenditure increases by 0.848%, which is significant at the 1% level; from Model 4—Gmm, it can be seen that, for every 1% increase in government fiscal housing support expenditure, urban residents’ consumption expenditure increases by 0.636%, which is significant at the 5% level. The underlying idea of this conclusion is also verified by most scholars. This paper also finds that both the household dependency ratio and house price will weaken the consumption expansion effect of housing support, i.e., the higher the dependency ratio is in households, the more the house price will have a negative impact on the consumption expansion effect of housing support.

In addition, regional differences are a phenomenon that will inevitably occur in the process of regional development. After the reform and opening up, with the speed of economic development, the flow of capital, labor, and other factors has also accelerated. However, the development among regions is not balanced, which mainly stems from the factor endowment, geographical location, and policy implementation among China’s regions. In general, a few large cities in China and several eastern regional provinces tend to be the focus regions for capital, raw material production, and new industries, and the economic gap between regions has gradually widened. Since the reform and opening up, China’s economy as a whole has developed relatively rapidly, but the economic development between regions is not balanced, and there is a structural break in the economic development between regions, while the difference in the level of economic development between regions has gradually widened, and the imbalance has also expanded and no effective economic cycle has been formed between regions. However, the Chinese government has been making efforts to narrow the development differences between regions and improve the overall economic development level, such as the implementation of the “Western Development,” “Northeast Revitalization,” and “Central Rise.” Despite these efforts, as we can see in this paper, there is still a tendency for the differences in economic development between regions to expand as time goes by. This trend is an objective reality and this paper should not ignore it when conducting any problem analysis, and should also take this factor into consideration in this paper’s research as much as possible.

This paper examines the regional differences in the impact of housing support-related expenditures on urban residents’ consumption in China from the perspective of geographic division, i.e., at the level of three major regional divisions in the East, Central and West regions (as in

Table 6. Regional affiliation of provinces and cities in mainland China (except Shanghai and Tibet).

Region	Include Provinces and Cities	Number
Eastern Region	Beijing, Zhejiang, Jiangsu, Fujian, Guangdong, Shandong, Liaoning, Tianjin, Hebei, Hainan	10
Central Region	Heilongjiang, Jilin, Hubei, Hunan, Shanxi, Henan, Anhui, Jiangxi	8
Western Region	Ningxia, Shaanxi, Inner Mongolia, Qinghai, Sichuan, Xinjiang, Chongqing, Yunnan, Guangxi, Gansu, Guizhou	11

).

In the regression, this paper still uses two stages of data, i.e., data related to China’s subsidized housing investment from 1999 to 2009 and data related to China’s financial housing support expenditure from 2010 to 2020, to illustrate the problem together. The regression results are shown in

Table 7. Sub-regional regression results.

1999–2009 (Fixed Effects Regression)				2010–2020 (Fixed Effects Regression)
	Eastern Region	Central Region	Western Region	Eastern Region	Central Region	Western Region
lnincome	0.340 *** (0.087)	0.324 *** (0.034)	0.311 *** (0.076)	0.535 *** (0.100)	0.836 *** (0.037)	0.836 *** (0.037)
lnphi	0.162 * (0.073)	0.448 (0.282)	−0.149 (0.235)	0.069 (0.17)	0.003 * (0.016)	0.057 *** (0.014)
lnchp	0.117 ** (0.037)	0.268 (0.167)	−0.075 (0.154)	0.763 (0.477)
bring	0.184 (0.467)	0.807 (0.681)	0.295 (0.480)
lnphi bring	−0.053 (0.114)	−0.193 (0.144)	−0.118 (0.095)
lnphilnchp	−0.018 ** (0.007)	−0.049 (0.035)	0.026 (0.030)
service	0.110 (0.112)	0.076 (0.057)	0.256 (0.190)	14.316 *** (1.640)	0.623 ** (0.278)	1.001 ** (0.400)
industry	0.085 * (0.045)	−0.153 (0.086)	0.083 (0.236)	12.505 *** (3.336)	0.447 ** (0.217)	1.534 *** (0.429)
Intercept term	−0.384 (0.393)	−1.712 (1.403)	1.227 (1.217)	−19.029 *** (6.165)	0.772 *** (0.280)	2.363 *** (0.381)
Number of samples	110	88	121	110	88	121
Within-R2	0.439	0.469	0.287	0.669	0.451	0.778

Note: ***, **, and * represent significant at 1%, 5%, and 10% levels, respectively, and the robustness criteria in parentheses are errors.

.

If divided by the three regions of East, Central, and West, the impacts of housing support-related expenditures on urban residents’ consumption expenditures differ significantly among the three regions. From the results presented in Table 7, it can be concluded that, (1) for the eastern region, the impact of construction of guaranteed housing on urban residents’ consumption is greater and more significant, while the impact of financial subsidies for housing support on consumption is not significant. Every 1% increase in housing construction expenditure brings a 0.162% increase in urban residents’ consumption expenditure; (2) for the central and western regions, the increase in financial housing support expenditure has a more significant impact on their consumption, and the increase in expenditure on guaranteed housing construction does not have a significant impact on the consumption of urban residents in the region. As shown in the fourth column of Table 7, for every 1% increase in fiscal housing expenditure in the Central region, urban residents’ consumption expenditure increases by 0.003%, which is significant at the 5% level, while the second column shows that this item is not significant. Similarly, in the seventh column, for every 1% increase in financial housing expenditure in the West region, urban consumption expenditure increases by 0.057%, which is significant at the 1% level. The third column shows that this item is not significant.

4.4. Model Robustness Tests

The model robustness test is designed to see if the results to be validated change when the parameter settings are changed. If the sign and significance of the variables change with the change in parameter settings, it means that the selected or set model is not robust. When the sign and significance of the variables do not change with the parameter settings, it means that the selected model is robust. Generally speaking, scholars often use three methods to test the robustness of their model. The first method is to start from the measurement method, and the results of OLS, Fix-Effect, GMM, and other regressions can be compared to see whether the results are still robust. The second method is to start from the data—the data can be classified according to different criteria—and test the variables’ sign and significance under different classifications. The third approach is to start from the variables, generally by replacing the variables with those related to the core variables and observing the sign and significance of the coefficients. In this paper, two of these methods are selected to test the robustness of the model, i.e., one is to use multiple econometric regressions to see whether there is any significant change in the results, and the other is to start from the data and adjust the data to see the regression results.

Robustness tests via multiple econometric regressions have been carried out already. For example, the static panel data are, first, regressed using a fixed-effects model (Fix-Effect), and then a dynamic model is built using a systematic GMM. Their outcomes revealed that the sign and significance of the core variables did not change significantly, and the other variables only fluctuated slightly. Based on the econometric approach, the results show that the estimations of this paper are robust. Second, this paper adjusts the data by excluding the Beijing, Tianjin, and Chongqing municipalities from the regressions. As shown in

Table 8. Regression after excluding the sample of municipalities.

	1999–2009		2010–2020
	Model 1-Gmm	Model 2-Gmm	Model 4-Gmm	Model 5-Gmm
L.lnconsume	0.621 *** (0.098)	0.557 *** (0.054)	0.8478 *** (0.012)	0.3276 * (0.177)
lnincome	0.344 *** (0.086)	0.366 *** (0.048)		1.3164 *** (0.184)
lnhse	0.005 * (0.005)	0.181 * (0.097)	0.0316 * (0.018) *	0.0325 * (0.019)
lnchp		0.105 ** (0.048)		−0.1419 ** (0.057)
bring		0.344 (0.389)
lnhse*bring		−0.102 (0.090)
lnhse*lnsh		−0.018 * (0.010)
service		0.231 ** (0.097)
industry		0.031 (0.061)
Observations	234	234	234	234
ar1p	0.000	0.000	0.002	0.009
ar2p	0.231	0.166	0.017	0.379
hansenp	0.346	0.335	0.381	0.373

Note: ***, **, and * represent significant at the 1%, 5%, and 10% levels, respectively; robustness standard errors are in parentheses; p-values for each test are reported below.

, the sign and significance of the main explanatory variables related to housing support expenditures in this chapter do not change significantly after the removal of the municipalities directly under the central government sample. This indicates that the findings in this chapter of the paper still hold after the robust regressions are applied to adjust the data.

5. Discussion

This paper empirically analyzes the impact of housing support on urban residents’ consumption in China using affordable housing data from 1999 to 2009 and housing support fiscal expenditure data from 2010 to 2020, respectively. The results of both static and systematic GMM regressions show that either an increase in the supply of affordable housing or an increase in fiscal spending on housing support by the government is beneficial to the expansion of urban residents’ consumption. The static regression results show that a 1% increase in the level of government housing support will boost the residential consumption expenditure by about 0.021–0.152%, and the dynamic regression results also reach the same conclusion that a 1% increase in the level of government housing support will boost the residential consumption expenditure by about 0.006–0.185%. The dynamic regression results show that financial subsidies for housing support have a better effect on the expansion of urban residents’ consumption than the construction of affordable housing. In addition, the results also indicate that the increase in commodity housing prices has a significant crowding-out effect on urban residents’ consumption.

The view of this paper is similar to that of most Chinese scholars [44,54]. However, the innovation of this paper is that, firstly, most scholars stay at the theoretical level and less-empirically analyze and verify the consumption effect of housing support expenditure. Second, it comprehensively examines the consumption effects of housing support in terms of both “making up for bricks,” i.e., affordable housing construction, and “making up for people,” i.e., financial expenditures on housing support. Thirdly, it examines the relevant effects at both the static and dynamic levels, which makes the study more scientific.

Notwithstanding, the main shortcomings of this study are as follows:

First, the factors influencing the consumption behavior of the residents need to be deepened. Although this paper summarizes the factors affecting urban residents’ consumption behavior and tries to reflect them in the model, it is still limited by the non-measurability of many factors and does not include all of them; however, these factors, such as the expectations and individual characteristics of residents, have a very important impact on residents’ consumption behavior. In future research, we will further investigate this issue.

Second, this study mainly explores the impact of the current expansion of housing support-related expenditures on urban residents’ consumption, but does not cover much regarding the consumption effects of long-term housing support-related expenditures. From the current perspective, housing support still has an expanding effect on the total consumption expenditure of urban residents. However, the supply of subsidized housing and financial housing support expenditure has an impact on the commercial housing market. It is not a matter of “the more the better or the bigger the better,” as there is a problem of moderate scale. Especially the financial housing expenditure, as part of the government’s public expenditure, should follow certain development rules and be reduced or maintained at its proper scale after the economic and social development reaches a certain level. The issue of the appropriateness of housing support-related expenditures involves more and broader contents, and the issue of the appropriate scale of housing support-related expenditures is also a direction for future efforts in this paper.

6. Conclusions and Policy Recommendation

6.1. Conclusions

This paper constructs a panel model based on the provincial level from the perspective of guaranteed housing supply and government financial expenditure. It uses the data of affordable housing from 1999 to 2009 and financial expenditure on housing guarantee from 2010 to 2020 to empirically test the impact of housing guarantee expenditure on urban residents’ consumption and the impact mechanism. It is found that, at this stage, government housing support has a certain degree of influence on the total level of urban residents’ consumption expenditure. Additionally, the government’s support for housing support is conducive to improving the total level of urban residents’ consumption expenditure. However, the current high housing price level has weakened the positive effect of government housing support on urban residents’ consumption expenditure to a certain extent. In the long run, if the housing price is not controlled and adjusted, the effect of government housing support on urban residents’ consumption expenditure will be greatly reduced, which will affect the lives of urban residents.

Different forms of housing support spending have different impact outcomes in different regions. For urban residents in the East region, the construction of guaranteed housing is more likely to release their consumption capacity, while for the Central and West regions, the impact of financial housing subsidies on their consumption is more pronounced.

6.2. Policy Recommendation

The research in this paper provides insights for governments at all levels to scientifically and effectively implement policies and better play the role of effective government support to the efficient market.

First, it is necessary to increase the effective supply of guaranteed housing and establish a sound and diversified housing support system. Housing support is a basic responsibility of the government. Expanding the effective supply of guaranteed housing and establishing a sound and diversified housing support system will have a direct driving effect on housing consumption and expanding domestic demand. Relying entirely on market mediation to solve the housing problem of residents will lead to a section of low- and middle-income residents having no housing, which will eventually affect their consumption. From this study’s analyses, it can be noted that improving China’s housing support system and establishing a diversified housing support system can promote urbanization and the gradual resolution of the urban–rural duality problem in order to expand domestic consumption demand.

Second, it is necessary to expand housing support expenditure, and to combine classified supply and stratified support. The combination of the market and support should be used to maximize the integration of resources and the detailed division of support targets, in order to fully stimulate the consumption potential of residents of all strata while safeguarding their livelihood.

Third, to improve the efficiency of the housing support supply, the relevant expenditure should be moderately tilted between regions. In regions with backward economic development, housing support coverage is limited, and residents are affected by uncertainties and have a low sense of support, which ultimately leads to the constraint of residents’ consumption in this region. However, the consumption potential of residents in less economically developed regions is huge, and their consumption desire is also stronger. Therefore, while developing the regional economy, it is also necessary to obtain a moderate tilt in policy, especially in financial support.