Introduction

In the context of current global economic turbulence and domestic economic transformation, the impact of economic policy uncertainty (EPU) on the consumption market is particularly significant. The Chinese government places high priority on expanding domestic demand, especially the growth of the rural consumption market, to promote high-quality economic development. On July 31, 2023, the National Development and Reform Commission released measures aimed at restoring and expanding consumption, focusing on enhancing consumption capacity, improving consumption conditions, and tapping into market potential.

Because of the narrowing urban‒rural gap in China, the consumption behavior of rural residents has gradually become a focal point of research. The consumption of rural residents is not only directly related to their quality of life and welfare levels but also plays a crucial role in the stable growth and sustainable development of the national economy. Currently, China faces multiple challenges, including economic transformation, consumption upgrades, and structural reforms on the supply side. In this context, the impact of EPU on rural residents’ consumption behavior deserves particular attention. This uncertainty may lead to fluctuations in the consumption market, increase residents’ perceptions of risk, and raise concerns about the future economy, which in turn affects their consumption decisions. Given this, this study proposes a core research question: What characteristics does the consumption behavior of rural residents in China exhibit against the backdrop of EPU? How does EPU influence the consumption structure of rural residents, especially in terms of basic consumption versus consumption for enjoyment and development? Additionally, this study explores whether social security levels play a mediating role, as well as the nature and extent of this mediating effect. Furthermore, it investigates whether a threshold effect exists in the relationship between EPU and rural residents’ consumption expenditures at different social security levels. Therefore, an in-depth exploration of how EPU shapes the consumption behavior of rural residents is essential for revealing the motivations behind consumer decision-making to optimize economic policies and stimulate consumption potential in rural areas.

Literature review and research hypotheses

Literature review

Measurement of economic policy uncertainty

EPU refers to the difficulty that individuals and firms face in predicting the next steps of policymakers regarding macroeconomic policies and understanding the impacts of these policy changes (Gulen and Ion 2016). The current literature generally employs two approaches to measure EPU: discrete political events and continuous indices. These methods provide researchers with a more comprehensive understanding of EPU and its effects on economic behavior.

The discrete political events approach focuses on using specific political events, such as government transitions, as proxies for EPU. Government transitions often lead to shifts in local government policies, thereby increasing policy uncertainty (Chen and Miao 2010). Owing to differences in governance philosophies and the personal experiences of new leaders, substantial changes in policy direction frequently occur following leadership transitions. As a result, political election uncertainty has been found to cause greater economic volatility in election years than in non-election years (Julio and Yook 2012; Boutchkova et al. 2012). Macroeconomic risks, such as those reflected in Federal Reserve announcements, are also highly sensitive to policy uncertainty levels (Savor and Wilson 2013). Similarly, the replacement of key officials has been used to capture policy uncertainty, with empirical results indicating that changes in policy resulting from leadership transitions can significantly impact resource allocation (Jia et al. 2013; Chen et al. 2017). Local officials, in particular, may adjust their fiscal policies to mitigate the adverse effects of policy uncertainty by increasing transfer payments, thereby reducing income inequality and improving consumption disparities (Zhou et al. 2020). Changes in leadership may also lead to regional policy discontinuities, exacerbating uncertainty and restraining income growth for rural populations (Xiao et al. 2022).

However, the discrete political events approach lacks continuity and timeliness, making fluctuations in policy uncertainty difficult to fully capture over time. To address this limitation, researchers have developed continuous EPU indices. For example, the standard deviations of output and inflation have been used as proxies for uncertainty, with studies showing that output uncertainty negatively affects investments (Driver and Moreton 1991). The conditional variances of real GDP, industrial production, consumer price index (CPI) inflation, and S&P 500 index returns have also been used to measure macroeconomic uncertainty, with results showing that greater uncertainty hinders the ability of managers to accurately predict firm-specific information (Baum et al. 2006). One study used stock market volatility in the UK as a measure of uncertainty and demonstrated that rising policy uncertainty increases the value of real options, leading firms to be more cautious when making investment decisions (Bloom et al. 2007). Macroeconomic cycles have also been used to gauge uncertainty, with findings showing that increased uncertainty heightens credit risk for corporate bonds (Zhou et al. 2011). The standard deviation of forecast errors has served as another measure of uncertainty, with research indicating that rising uncertainty dampens firms’ incentives, hindering economic growth and development (Bachmann et al. 2013). Baker et al. (2016) tracked keywords such as “economy,” “policy,” and “uncertainty” in the South China Morning Post and used the data to develop an EPU index for China. This index, valued for its continuity and timeliness, has gained considerable traction in the academic community. Additionally, some studies have integrated stock market volatility, macroeconomic variables, and EPU indices to construct a more comprehensive framework for measuring macroeconomic uncertainty (Jin and Zhang 2019).

The impact of economic policy uncertainty on consumption

The impact of EPU on household consumption has long been a focal point of academic research. Most studies have concentrated on the macroeconomic level, whereas studies specifically addressing rural household consumption are relatively scarce. Scholars hold differing views on how EPU influences consumption. Growth option theory and the Hartman‒Abel effect offer explanations for its potential positive effects, whereas the precautionary savings effect, real option effect, and financial friction effect elucidate its potential negative impacts (Born and Pfeifer 2014).

From a positive perspective, some studies indicate that an increase in income uncertainty may lead to higher consumption expenditures among rural residents, as this uncertainty can sometimes result in unexpected income (Chen 2014). Research employing VAR models has revealed that long-term inflation expectations may rise due to fluctuations in uncertainty (Leduc et al. 2007). Additionally, empirical studies in China suggest that EPU has a significant positive effect on business activities, indicating an incentive effect (Gu et al. 2018). The results of three-dimensional impulse response tests demonstrate that EPU positively promotes rural consumption (Zhang and Liu 2019). Empirical tests via GMM estimation reveal that EPU leads to a short-term upgrade in the consumption structure through a guiding effect (Zhang and Liu 2020). When uncertainty declines, proactive fiscal policies can significantly alleviate consumption volatility (Hu and Wang 2020). A TVP-SV-VAR model study revealed that EPU has a significant positive time-varying effect on urban household consumption expenditures and that, as consumers adapt to and understand policies better, the adverse impacts of shocks decrease (Nan et al. 2022). Furthermore, some research suggests that long-term interest rate uncertainty can hinder consumption, whereas its short-term impact on household consumption is relatively weak, indicating that uncertainty’s influence on daily household consumption has a notable time-varying characteristic (Njindan 2019).

On the other hand, some studies have highlighted the negative effects of EPU on consumption. Certain research findings suggest that under liquidity constraints and uncertainty, the level of household consumption may exhibit a declining trend over certain periods (Wan et al. 2001). During periods of economic transformation, forward-looking consumers might reduce current consumption to smooth future consumption (Hang et al. 2013). The sensitivity of rural consumers to policy implementation tends to lag, indicating that uncertainty exerts a certain suppressive effect on short-term consumption levels (Huang and Guo 2015); however, as their perception and understanding of policies improve, this negative impact gradually diminishes (Nan et al. 2022). An analysis using the SVAR model indicates that EPU negatively affects consumption by influencing corporate decision-making (Tian et al. 2016). Liu et al. (2020) reported that when the impacts of EPU on macroeconomic variables were studied, shocks to consumption and investment were the most severe, with significant negative effects. When uncertainty levels decline, the implementation of proactive fiscal policies can effectively mitigate consumption volatility (Hu and Wang 2020). Moreover, numerous scholars (Li and Yu 2021; Fu and Lü 2022) have demonstrated that EPU suppresses consumption by affecting precautionary savings. Building on this, Biljanovska et al. (2021) examined the spillover effects of EPU on economic activities in other countries through panel vector autoregression and discovered that approximately two-thirds of the spillover effects on real output, private consumption, and private investment originate from abroad. In addition to exerting a suppressive effect on private consumption, He (2023) reported that EPU significantly negatively impacts household consumption budgets, with families experiencing greater daily consumption fluctuations being more sensitive to changes in EPU (Nam et al. 2021). Furthermore, forward-looking consumers tend to reduce current consumption to smooth future consumption when faced with uncertainty (Hang and Yan 2013), although whether this mechanism applies to rural residents remains to be thoroughly investigated. Morikawa (2019) found through empirical testing that the uncertainty surrounding Japan’s tax policies led to an increase in the propensity of households to save, which contributed to stagnation in Japanese consumption growth. Bahmani-Oskooee and Maki (2020) expanded beyond a single-country perspective and found that policy uncertainty has a short-term negative impact on consumption in the G7 countries, with varying effects across nations. Murray (2014) noted that fiscal policy uncertainty negatively affected the United States’ real GDP, consumption, investment, and unemployment rates, resulting in a decrease in the real GDP growth rate by ~2%. Similarly, Saygin and Evren (2018) showed that EPU adversely affected economic growth, consumption, and investment in Turkey. These studies underscore the complex and multidimensional nature of the impact of EPU on consumption; thus, in-depth analysis from various perspectives and levels is needed.

The impact of social security levels on rural household consumption

The literature on the impact of social security levels on rural household consumption presents a range of perspectives, which can be categorized into three main viewpoints: positive effects, ambiguous effects, and negative effects.

Some studies suggest that improvements in social security levels can lead to increased consumption by reducing the need for savings. This perspective posits that social security acts as a substitute for savings, thereby enhancing the consumption capacity of rural residents. By extending the analysis of the life-cycle model, it has been shown that social security reduces individual savings requirements, leading to higher consumption expenditures (Feldstein 1979). Empirical research conducted in Turkey indicates that increased social security spending has a robust positive effect on overall consumption (Aydede 2007). Similarly, in the long run, social security significantly promotes consumption by rural residents (Jiang et al. 2010). Further studies have demonstrated that expanding social security coverage can effectively reduce rural residents’ reliance on guaranteed income, thereby increasing consumption (Yin and Ji 2011). In economically underdeveloped regions, the consumption-promoting effect of social security spending is even more pronounced (Jiang et al. 2012). Additionally, pension insurance has been found to have a significant positive effect on consumption expenditures by urban residents (Jiang and Quan 2018). The new rural pension policy has also notably increased the level of basic living consumption by rural residents, with varying degrees of impact depending on the level of economic development (He and Li 2020). Research has further indicated that the new rural pension policy not only promotes upgrading of the farmer consumption structure but also that the precautionary savings mechanism is a crucial avenue through which this effect is realized (Wang et al. 2023).

Conversely, some studies argue that the impact of social security levels on consumption is not always clear-cut and may be influenced by various factors. While social security spending can sometimes promote consumption, its effects may become ambiguous because of intergenerational transfer effects and other factors (Barro and MacDonald 1979). Specifically, pensions from different sources may lead to varying changes in household consumption; for example, in the UK’s pension system, the impacts of state-funded portions differ from those of individual contributions (Blake 2004). Furthermore, while the coverage rate of pension insurance may suppress consumption growth, its replacement rate may positively influence consumption, highlighting that the effects of social security on consumption depend on the balance of multiple factors (Xue and Zhao 2019).

In contrast, some studies suggest that increases in social security levels may suppress household consumption through asset substitution effects and retirement effects. An analysis of data from China between 1980 and 2004 revealed that although social security spending exhibited an asset substitution effect on household consumption, this effect was less than the retirement effect, ultimately leading to a suppression of consumption (Yang and Wang 2007). Similar studies have indicated that social security spending crowds out some consumption expenditures by farmers (Xie 2011). Additionally, when pension income and life expectancy are uncertain, households tend to increase precautionary savings, consequently reducing current consumption (Van Santen 2019). Although pension insurance can alleviate consumption declines resulting from increased life expectancy, its overall effect remains limited (Cai 2015). Some studies indicate that the wealth substitution effect of China’s pension system on urban residents’ consumption is less than the retirement effect, thus reinforcing its suppressive effect on consumption (Yang et al. 2016).

Research hypotheses

According to real options theory, from the consumer's perspective, economic uncertainty can indirectly amplify the future value of goods. In other words, buying goods now may be more cost-effective than purchasing the same goods in the future, which may increase consumers’ inclination to make purchases. Consequently, EPU may stimulate consumer spending, particularly for durable and essential items. When EPU is high, consumers often stock up on durable and essential goods to mitigate the risks associated with future uncertainties. This behavior aims to preempt the potential negative impacts that may arise from unpredictable shifts in the supply and demand of these goods. However, rural consumers tend to have a lower awareness of the importance of bolstering their human capital to combat future uncertainties. During periods of significant EPU, they may overlook the value of augmenting their knowledge base and physical well-being to counter the effects of future risk. Instead, they often defer expenditures on investments in human capital, such as cultural, educational, recreational, and healthcare activities (Zhang and Liu 2020). In light of these considerations, this paper proposes Research Hypothesis 1.

Hypothesis 1: Economic policy uncertainty is expected to increase rural residents’ essential consumption while decreasing their discretionary consumption.

Social security income represents a vital income stream for rural residents, and the adequacy of this support significantly influences the relationship between EPU and rural consumption. According to Keynesian consumption function theory, the marginal propensity to consume is notably greater among low-income groups than among their high-income counterparts. Consequently, an increase in the disposable income of low-income individuals leads to a more substantial rise in consumption than the same increase would for those with higher incomes. Theoretically, social security expenditures should foster a rise in consumption levels. In line with the permanent income hypothesis, social security also bolsters residents’ expectations of having a stable, long-term source of income. Furthermore, it mitigates concerns about future uncertainties and lessens the need for precautionary savings designed to hedge against potential risks, thus encouraging greater consumer spending. Consequently, if social security provisions are inadequate, rural residents may be more vulnerable to the effects of economic uncertainty, making them more inclined to adopt conservative consumption habits. Conversely, robust social security programs can enhance economic stability and purchasing power among rural residents while diminishing the adverse effects of EPU on their consumption behavior. In light of these considerations, this paper proposes Research Hypothesis 2.

Hypothesis 2: The degree of social security is hypothesized to mediate the relationship between economic policy uncertainty and rural resident consumption patterns.

China has long demonstrated a pronounced imbalance in regional development, with economic disparities between regions creating a divergence in consumer expectations and patterns (Yuan and Xu 2021). In areas with varying economic conditions, the initial wealth levels of residents also vary substantially, affecting their capacity to manage risks and engage in consumption. Consequently, the influence of EPU on rural residents’ consumption expenditure is likely to exhibit significant regional variations. Furthermore, the overall environment and policy directives of different regions can shape residents’ understanding and interpretation of economic policies, influencing their consumption decisions and choices. Against this backdrop, this paper proposes Research Hypothesis 3.

Hypothesis 3: The influence of economic policy uncertainty on rural consumption is posited to be contingent upon the degree of regional economic development.

Methodology and data

Model specification

Fixed-effects model

To test Research Hypothesis 1, which posits that EPU enhances rural residents’ essential consumption while suppressing their discretionary consumption, the following fixed-effects model is formulated:

$${\mathrm{ln}{\rm{RC}}}_{i,t}={\beta }_{0}+{\beta }_{1}{\mathrm{ln}{\rm{EPU}}}_{i,t}+{\beta }_{2}{{\rm{controls}}}_{i,t}+{\eta }_{t}+{\eta }_{{pro}}+{\varepsilon }_{i,t}$$
(1)
$${{ln}{{RSC}}}_{i,t}={\beta }_{0}+{\beta }_{1}{{ln}{{EPU}}}_{i,t}+{\beta }_{2}{{{controls}}}_{i,t}+{\eta }_{t}+{\eta }_{{pro}}+{\varepsilon }_{i,t}$$
(2)
$${{ln}{{RDC}}}_{i,t}={\beta }_{0}+{\beta }_{1}{{ln}{{EPU}}}_{i,t}+{\beta }_{2}{{{controls}}}_{i,t}+{\eta }_{t}+{\eta }_{{pro}}+{\varepsilon }_{i,t}$$
(3)

In the model equations, the subscript i denotes the province, and t denotes time. The dependent variable ln(RCi,t) signifies the consumption expenditure by rural residents in province i at time t, which is differentiated into essential consumption ln(RSCi,t) and discretionary consumption ln(RDCi,t). ln(EPUi,t) is the key independent variable, and Controlsi,t includes the control variables that affect rural residents’ consumption expenditures. Furthermore, ηt and ηpro represent the time and province fixed effects, respectively, whereas \({\varepsilon }_{i,t}\) denotes the error term.

Mediation model

To examine Hypothesis 2, which suggests that the degree of social security mediates the relationship between EPU and rural resident consumption patterns, this study adopts the research methodology of Wen and Ye (2014). It utilizes social security as a mediating variable and constructs mediation effect models, as presented in Eqs. (4) and (5).

$${{{SSR}}}_{i,t}={\beta }_{0}+{\beta }_{1}{{ln}{{EPU}}}_{i,t}+{\beta }_{2}{{{controls}}}_{i,t}+{\eta }_{t}+{\eta }_{{pro}}+{\varepsilon }_{i,t}$$
(4)
$${{ln}{{RC}}}_{i,t}={\beta }_{0}+{\beta }_{1}{{ln}{{EPU}}}_{i,t}+{\beta }_{2}{{{SSR}}}_{i,t}+{\beta }_{3}{{{controls}}}_{i,t}+{\eta }_{t}+{\eta }_{{pro}}+{\varepsilon }_{i,t}$$
(5)

In these model equations, SSRi,t is the mediator variable, and the meanings of the dependent variable, the independent variables, and other terms are consistent with those in Eqs. (1–3).

Threshold effect model

In addition to considering the mediating effect, this study also examines the nonlinear structural relationship between EPU and rural residents’ consumption expenditures by setting the level of social security as a threshold variable. This approach aims to investigate the differentiated impact of EPU on rural residents’ consumption expenditures with varying levels of social security. This paper refers to the threshold effect testing method developed by Hansen (1999) to establish the threshold model presented in Eq. (6).

$$\begin{array}{l}{{{ln}}{RC}}_{i,t}={\alpha }_{0}+{\alpha }_{1}{{ln}}{{EPU}}_{i,t}\,\cdot\, I({{SSR}}_{i,t}\le {\mu }_{1})+{\alpha }_{2}{{{ln}}{EPU}}_{i,t}\,\cdot\, I({{SSR}}_{i,t}\, > \,{\mu }_{2})\\\qquad\qquad+\,{\alpha }_{3}{{controls}}_{i,t}+{\varepsilon }_{i,t}\end{array}$$
(6)

In this context, SSRi,t represents the threshold variable, and \({\mu }_{i,t}\) denotes the threshold value. The meanings of the dependent variable, independent variables, and other terms are the same as those in Eqs. (1–3).

Variable selection

Dependent variable

The data on rural resident consumption (lnRC) expenditures across various provinces are sourced from the National Bureau of Statistics. This consumption is bifurcated into two structural categories: essential consumption (lnRSC) and discretionary consumption (lnRDC). Essential consumption (lnRSC) typically refers to expenditures that satisfy basic living needs. This type of consumption is generally considered inelastic, as individuals continue to engage in such spending even in the face of high economic uncertainty to maintain a basic standard of living. In contrast, discretionary consumption (lnRDC) encompasses spending aimed at enhancing quality of life and enjoyment levels. This type of consumption is more influenced by consumer expectations and confidence and is often suppressed during periods of high economic uncertainty. For data processing purposes, essential consumption is operationalized as the aggregate expenditure on food, clothing, and shelter. Conversely, discretionary consumption is quantified by aggregating expenditures on household utilities and services; transportation and communication; healthcare; education; recreational activities; and other miscellaneous goods and services.

Core explanatory variables

To examine economic policy uncertainty (lnEPU) in China, we adopt the monthly Chinese economic policy uncertainty (CNEPU) index developed by Huang and Luk (2020). This index was developed by extracting key terms such as “economy,” “uncertainty,” and “policy” from prominent newspapers such as the Beijing Daily, Guangzhou Daily, Shanghai Morning Post, and Southern Metropolis Daily. The original monthly data were converted to an annual figure by taking the arithmetic mean.

Mediating variable and threshold variable

The social security (SSR) level is both the mediator variable and threshold variable in this study. The social security level is a metric that captures the proportion of the social security subsidy expenditure relative to the total general budget expenditure at the local level. The original data are sourced from the China Statistical Yearbook. The Chinese government is actively committed to enhancing the welfare of its citizens. This commitment has led to an improvement in the social security level, which, by extension, contributes to an increase in rural residents’ disposable income. As a result, the fundamental consumption needs of these residents are met.

Control variables

Building on prior research (Hu et al. 2013; Qiu and Liu 2019; Jiang and Zhang 2021), this study identifies six variables that influence rural resident consumption expenditures in the economic, demographic, social, and educational dimensions. These include the CPI, rural resident disposable income, the urbanization rate, the child dependency ratio, the elderly dependency ratio, and educational attainment. Collectively, these variables serve to partially alleviate endogeneity concerns. Data on rural residents’ disposable income were obtained from the National Bureau of Statistics of China. The data for the remaining control variables were extracted from the China Statistical Yearbook. Table 1 provides descriptions of the variables.

Table 1 Variable and data descriptions.
Full size table

(1) Consumer price index (CPI). The price level directly affects residents’ real purchasing power, and differences in price levels across regions may influence consumer spending decisions. As the price level in China continues to rise, residents’ consumption levels are impacted, leading to a gradual decline in purchasing power and suppressed consumption demand. To account for these various influences, the price level is included as a control variable in this study.

(2) Rural residents’ disposable income (lnINC). Income is the foundation of consumption. We use the disposable income of rural residents to reflect their consumption capacity and include it as a control variable. As disposable income increases, residents’ consumption expenditures typically change in the same direction; however, the incremental increase in consumption spending tends to decrease as income increases. To enhance the rationality of the research design, the natural logarithm of this variable is taken.

(3) Urbanization rate (URB). Urbanization signifies substantial changes in the quantity and structure of the rural population, often correlating with significant improvements in economic development levels (Zhu et al. 2011). This change may influence residents’ consumption patterns and preferences, which in turn can significantly impact consumption levels. Therefore, this variable is included in the model.

(4) The child dependency ratio (CDR) and the elderly dependency ratio (ODR) are calculated. These two dependency ratios reflect the household burden faced by rural residents, which may affect their consumption behavior. Empirical tests by Mao et al. (2013) indicate that an increase in the child dependency ratio leads to higher consumption expenditures among rural residents, whereas the impact of an increasing old-age dependency ratio on rural consumption is not significant. However, research by Gao et al. (2015) revealed that a rising old-age dependency ratio may contribute to an increase in the overall consumption rate.

(5) Educational attainment (EDU). Residents with different levels of education exhibit significant differences in their consumption decisions (Wang et al. 2019). Accurately understanding the differences in consumers’ educational levels and consumption needs can help the government formulate more targeted and differentiated consumption policies. The education level indicator is derived from five groups of data, including the number of individuals with primary, junior high, senior high, and above educational attainment, as well as the population aged six and above. Interpolation is used to fill in missing data from 2001 and 2010, aiming to minimize the loss of variable information.

Descriptive statistics

Descriptive statistics for the research variables are displayed in Table 2. The standard deviation of rural resident consumption expenditures is 4954.711, ranging from a minimum of 1052 yuan to a maximum of 27,205 yuan. This range suggests considerable variation in rural consumer spending and highlights the disparities in consumption levels among rural residents. The average expenditure on survival-type consumption is 4076.298 yuan, which is roughly double that of development-type and enjoyment-type consumption, which averages 2544.362 yuan. This difference emphasizes that rural consumption patterns are driven predominantly by survival needs. The proxy indicator for EPU has a broad range, with a minimum of 11.787 and a maximum of 150.716, indicating significant variability. With respect to the educational level of rural residents, the minimum and maximum values are 0.738 and 36.201, respectively, suggesting a notable increase in rural education levels over the study period. After reviewing the descriptive statistics for the other control variables, except for the price level, there are considerable differences across the sample data from 2000 to 2021.

Table 2 Descriptive statistics.
Full size table

Empirical results and robustness test

Baseline regression results and robustness test

Baseline regression results

The Hausman test was conducted, and the results suggested the need to employ a fixed-effects model for empirical regression. Accordingly, both survival-type and enjoyment-type consumption expenditures were included as dependent variables in the model for regression analysis. The regression results are displayed in Table 3.

Table 3 Results of the fixed-effects model.
Full size table

Model 1 displays the regression outcomes for the influence of EPU on rural residents’ consumption expenditures. The coefficient of the core explanatory variable is both statistically significant and positive, suggesting that heightened EPU is associated with an increase in rural resident consumption expenditures. This upswing in EPU might alter consumption patterns over time, potentially encouraging short-term consumer behaviors such as an emphasis on “living in the moment” (Fu and Lv 2022). Other control variables also influence rural residents’ consumption expenditures. First, there is a significant positive effect of rural resident income on rural resident consumption expenditures, in alignment with theoretical expectations and empirical findings. The per capita disposable income of rural residents is significantly and positively correlated with their consumption expenditures, indicating that disposable income can partially mitigate the impact of EPU. A stable income allows residents to sustain their motivation and capacity to consume despite risks associated with uncertainty. Second, the child dependency ratio significantly inhibits rural residents’ consumption expenditures. Given their single income sources and the relatively stable income growth experienced by rural residents, they tend to adopt a conservative approach to child-rearing decisions. In addition to covering basic necessities and funding their children’s primary education, there is a reluctance to increase expenditures on additional activities such as talent or sports development (Xu and Song 2019). Furthermore, traditional customs and habitual thinking often lead rural residents to consider long-term educational and marital prospects for their children (Wang 2009), prompting increased precautionary savings and, consequently, lower consumption expenditures. Third, the elderly dependency ratio significantly affects rural residents’ consumption expenditures. In contrast to their urban counterparts, most elderly rural individuals lack retirement income and depend primarily on basic pension insurance, which often falls short of meeting their living expenses. Because the majority of elderly rural individuals rely on family savings and support from their children, there is a notable absence of external social support. Consequently, an increase in the elderly dependency ratio is correlated with an increase in the overall rural resident consumption expenditure amount.

Model 2 includes the regression outcomes for the impact of EPU on rural residents’ essential consumption expenditures. The coefficient for the key explanatory variable is statistically significant and positive, which implies that increased EPU substantially increases the essential rural resident consumption expenditure amount. At elevated levels of EPU, there is an increased risk of substantial shifts in the supply and demand dynamics of goods. In anticipation of potential future shortages, rural consumers often opt to procure ample essential goods during the present period. This behavior can manifest as hoarding, reflecting a type of irrational consumer response.

Model 3 displays the regression findings that pertain to the influence of EPU on rural residents’ expenditures in relation to discretionary consumption. The coefficient of the core explanatory variable is statistically significant and negative, indicating that an increase in EPU markedly dampens the discretionary rural resident consumption expenditure amount. In the presence of heightened uncertainty, rural residents’ consumption behavior is influenced by their historical consumption patterns, demonstrating a pronounced hysteresis effect (Cui 2011; Wang et al. 2015). Consequently, rural consumers have difficulty augmenting their spending on discretionary consumption categories such as education, culture, and entertainment within a short timeframe. Furthermore, owing to the influence of factors such as ideological concepts and educational levels, rural residents are less inclined to acquire novel skills as a strategy to counter future risks when confronted with uncertainty. In contrast to investing in future-oriented skills and knowledge, they prioritize concerns about immediate economic losses that may arise from erroneous decision-making. As a result, they are more inclined to defer such expenses, bolster their savings, and adopt more cautious approaches to manage prospective uncertainty risks.

In conclusion, as EPU increases, there is a corresponding escalation in rural residents’ essential consumption expenditures. From the vantage point of the consumption structure, heightened EPU intensifies residents’ apprehensiveness regarding prospective future risks. This leads to an increase in essential consumption expenditures and a decrease in discretionary consumption expenditures. This shift is detrimental to the optimization and advancement of the consumption structure. This outcome supports Hypothesis 1.

Robustness test

To improve the reliability of the empirical findings, this study employed three robustness checks.

  1. (1)

    Recalculation of the core explanatory variable

    In the aforementioned regression analyses, the EPU index, which was initially calculated as an arithmetic mean, is now adjusted to an annual figure using the geometric mean following the methodological precedents of Li and Yang (2015) and Gu et al. (2018). The rationale for choosing the geometric mean is that it is less sensitive to extreme values, which helps us understand the general trend of the impact of EPU on rural household consumption rather than the short-term fluctuations driven by outliers. Additionally, to test the external validity of our results, we introduced the global economic policy uncertainty index (GEPU) as an alternative variable. This step is motivated by the fact that although China is the primary focus of the study, in the context of global economic integration, global economic policy uncertainty may also influence the consumption behavior of rural Chinese residents. The baseline regressions were rerun, and the findings are detailed in Table 4. The results consistently indicate that EPU significantly influences the dependent variables across all six models, with the directions of the effects remaining congruent. This consistency supports the conclusion that EPU is associated with increased essential consumption expenditures and decreased discretionary consumption expenditures.

    Table 4 Regression results of the recalculated EPU index.
    Full size table
  2. (2)

    Respecification of the dependent variable

    Initially, the analysis distinguished between two broad categories of per capita consumption expenditure from the China Statistical Yearbook: essential consumption, which is based on the first three categories (food, clothing, and housing), and discretionary consumption, which is based on the last five categories (household facilities, healthcare, transportation and communication, education and entertainment, and other goods and services). To reinforce the robustness of the empirical findings, this study introduces an alternative classification system proposed by Xu and Gui (2022) that segments consumption expenditure into three types—physical consumption (RC1), labor consumption (RC2), and mental consumption (RC3)—to more closely observe the impact of EPU on different types of consumption. The rationale for this classification is that different types of consumption may exhibit varying sensitivities to changes in the economic environment. By segmenting these categories, we can more accurately capture the specific effects of EPU on consumption. A comparison of the two methods for classifying consumption expenditures is displayed in Table 5.

    Table 5 Two classification methods of consumption expenditure.
    Full size table

    The re-estimated empirical results are detailed in Table 6. They reveal that EPU exerts a significant positive influence on physical consumption and a significant negative effect on both labor and mental consumption. These results are congruent with earlier conclusions.

    Table 6 Regression results after respecification of the dependent variable.
    Full size table
  3. (3)

    Exclusion of data from directly administered municipalities

    Beyond variable adjustments, this study further refined the sample by excluding data from the four municipalities (Beijing, Tianjin, Shanghai, and Chongqing). The rationale for this exclusion is that the economic development level, social security system completeness, and consumption habits of residents in these municipalities may significantly differ from those in other provinces. By excluding these data, we aim to reduce heterogeneity in the model, allowing the research results to more accurately reflect the consumption behavior of residents in rural areas across China. The regression results presented in Table 7 indicate that the coefficients associated with EPU are statistically significant across all three models. Additionally, the direction of these coefficients is consistent with our prior conclusions, supporting the link between EPU and consumption by rural residents.

    Table 7 Results excluding data from directly administered municipalities.
    Full size table

Instrumental variable approach

To address the potential bidirectional causality between EPU and household consumption, we employed an instrumental variable (IV) approach using the two-stage least squares method. Drawing on the methodology of Qin and Zhang (2025), we selected the annualized volatility of Brent crude oil futures pricesFootnote 1 as the IV. The rationale is twofold: (1) As a global benchmark for commodity pricing, fluctuations in Brent crude oil prices significantly correlate with China’s domestic EPU by influencing energy policy adjustments. (2) International oil prices are determined by global supply and demand dynamics (China acts as a price taker), which ensures that there is no direct linkage to household consumption decisions, and thereby the exogeneity requirement is satisfied.

Table 8 reports the IV regression results. The empirical findings confirm the validity of the IV. The first-stage regression shows that the coefficient of the IV is significant at the 1% level (β = 0.068, SE = 0.009). The Cragg–Donald Wald F statistic (64.19) far exceeds the Stock–Yogo weak identification test critical value (16.38), which precludes weak instrument concerns. The null hypothesis of the Anderson Canon LM test is rejected at the 1% level, which confirms the relevance of the instrument. After accounting for endogeneity, the impact of EPU on rural household consumption remains significant at the 1% level, which corroborates the robustness of the baseline results. The IV estimates reveal that the positive effect of EPU on consumption decreases from an OLS estimate of 0.453–0.222, which highlights the direction and magnitude of the endogeneity bias. Specifically, OLS may overstate the stimulative effect of EPU due to reverse causality (e.g., consumption growth may prompt short-term policy adjustments, creating spurious correlations) or omitted variables (e.g., unobserved precautionary savings or employment uncertainty). By isolating exogenous oil price shocks, the IV approach mitigates these biases and yields a more precise estimate. The corrected results suggest that while EPU still drives rural consumption through “immediate consumption” behavior, its effect is 51% weaker than the initial estimated effects, which reflects a dynamic balance between policy-induced consumption stimulation and suppression mechanisms.

Table 8 Results of the instrumental variable method.
Full size table

Mediation effect analysis

By conducting baseline regression analysis and robustness tests, this study provides compelling evidence that EPU hinders the advancement of rural consumption structures. This paper builds on the relevant literature to hypothesize that the social security level may serve as a mediating factor in the relationship between EPU and consumption by rural residents. Consequently, by estimating the mediation model and employing the Sobel test, we explore the mechanism by which EPU influences rural consumption expenditures. The outcomes of the test are presented in Table 9.

Table 9 Results of the mediation effect analysis.
Full size table

Table 9 illustrates how social security mediates the impact of EPU on consumption by rural residents. As shown in Column (1), there is a significant increase in rural resident consumption expenditures due to EPU, with a coefficient estimate of 0.18, which represents the total effect. According to Column (2), the regression results reveal that EPU positively affects the social security level. In response to heightened uncertainty risks, governments often extend policy support to low-income groups to stabilize the economy, resulting in increased social security subsidies and an enhanced overall = social security level. Column (3) shows that both EPU and the social security level positively influence consumption by rural residents, with a direct effect of 0.143. This indicates that an increase in fiscal spending on social security by the government leads to higher social security income for residents, which in turn mitigates the pressure from uncertainty risks and ensures the basic consumption capacity and demand of rural residents. In summary, social security exerts a partial mediating effect on the relationship between EPU and consumption by rural residents, confirming Hypothesis 2. Using the formula a × b/c to calculate the mediating effect, it is determined that social security plays a 20.8% mediating role (Fig. 1).

Fig. 1: This figure presents the main and mediation effects in our study.
figure 1

Path (a) shows the impact of economic policy uncertainty on the social security level. Path (b) shows the impact of the social security level on rural resident consumption expenditures. Path (c) depicts the main effect of economic policy uncertainty on rural resident consumption expenditures. The arrows indicate the direction of the relationships, and the coefficients on each arrow represent the strength of the effects. Besides, ***indicate significance at the 1% level.

Full size image

Threshold effect analysis

Considering that the impact of EPU on rural resident consumption expenditures may vary at different social security levels, this study further investigates the existence of this nonlinear structural relationship by using the social security level as a threshold variable. Drawing on the research of Hansen (1999), we employed the bootstrap resampling method to sequentially test for the presence of single and multiple thresholds. The results of the threshold effect tests are presented in Table 10.

Table 10 Results of the threshold effect tests.
Full size table

As shown in Table 10, the estimated value of the single threshold for the social security level is 2.4605, which is significant at the 1% level and falls within the 95% confidence interval [2.415, 2.484]. In contrast, the F statistics for the double and triple thresholds did not pass the 5% significance test. On the basis of the above analysis, this study uses the social security level as the threshold variable and conducts a single threshold regression estimation via panel data from 2000 to 2021. The results are presented in Table 11.

Table 11 Results of threshold effect analysis.
Full size table

The threshold model test results confirmed the existence of a nonlinear structural relationship between EPU and rural resident consumption expenditures. As shown in Table 11, the regression coefficients of EPU on rural resident consumption expenditures under different social security levels are 0.597 and 0.543, both of which are significant at the 1% level. This finding indicates that as the social security level increases, the impact of EPU on rural resident consumption expenditures decreases. When the social security level is less than 2.4605, the impact of EPU on rural resident consumption expenditures is positive and significant at the 1% level, with an estimated coefficient of 0.597. When the social security level exceeds 2.4605, the effect remains significantly positive, but the estimated coefficient decreases to 0.543. This suggests that as social security levels rise, the stimulating effect of EPU on rural resident consumption expenditures weakens, showing a diminishing marginal effect.

The possible reasons for this include the fact that with the development of socioeconomic levels, the Chinese government has continuously worked to increase the well-being of its citizens, promote social justice, and increase fiscal spending on social security. As a result, the social security level has steadily improved, which has in turn increased rural residents’ disposable income, reducing the pressure and inhibitive sentiment caused by uncertainties regarding economic policies. Thus, rural residents are more willing to improve their quality of life, including spending on clothing, food, and other essentials, thus promoting growth in household consumption. Therefore, before crossing the threshold, essential consumption by rural consumers reaches a relatively stable state. After surpassing the threshold, however, there are two reasons for the diminishing marginal consumption effect: first, essential consumption has already stabilized, so even if disposable income continues to increase, there is no further growth in survival-related spending. Additionally, owing to factors such as the mindset and spending habits of rural consumers, they are unlikely to significantly increase their spending on development and leisure consumption. Second, with rising prices, the real wealth effect diminishes, as income growth is offset by price inflation, which can dampen the willingness of rural consumers to spend.

When verifying the robustness of the threshold effect, we further controlled for external macro factors that could affect the results. Drawing on the research of Ye and Zhang (2024) and Caldara and Iacoviello (2022), we included both per capita GDP and the global geopolitical risk index (GPR) in the model for re-estimation. The rationality of this design is as follows: per capita GDP controls for the potential effect of regional economic level differences on consumption, and GPR captures the spillover effects of international conflicts, trade frictions, and other exogenous shocks on China’s policy formulation (Shi et al. 2024), which ensures that the threshold effect of EPU is not disturbed by global systemic risk. Table 12 shows that after the addition of the control variables, the threshold effect of EPU remains significant. When the social security level is below 2.4605, the impact of EPU on rural resident consumption is positively significant at the 1% significance level, with the estimated coefficient changing from 0.597 to 0.542; when the social security level is above 2.4605, the impact of EPU on rural resident consumption remains significantly positive, but its estimated coefficient changes from 0.542 to 0.539. Compared with the initial threshold effect model, the adjusted model’s core coefficient changes by less than 5%, and the null hypothesis of the threshold value stability test (bootstrap P = 0.000) is rejected. This finding indicates that the threshold effect remains robust after we control for per capita GDP and GPR and further validates the reliability of our research findings.

Table 12 Results of threshold effect analysis after the addition of control variables.
Full size table

Heterogeneity test

Heterogeneity tests were performed by incorporating interaction terms, and the results are detailed in Table 13. These results support the hypothesis of heterogeneity, indicating that the influence of EPU on consumption by rural residents differs across regions on the basis of their level of economic development. According to the findings, in Model 1, the interaction term between EPU and the economic development level is significantly negative. This suggests that an increase in EPU stimulates total consumption expenditure through its impact on survival-type consumption. However, in regions with greater economic development, this positive effect is less pronounced, whereas in less-developed regions, the positive effect is greater. Model 2 reveals an insignificant interaction term between the EPU index and the economic development level. This suggests that there is no statistically clear correlation between EPU and essential consumption expenditures. Model 3 shows a significantly negative interaction term between the EPU index and the economic development level. Additionally, in Model 3, EPU is negatively correlated with discretionary consumption expenditures. This implies that as the local economic development level increases, the negative impact of EPU on discretionary consumption expenditures decreases. In more economically developed regions, due to the ratchet effect and consumption inertia, consumers do not experience a significant decrease in the overall discretionary consumption expenditure level despite environmental changes. In contrast, in regions with lower economic development, the negative effect of EPU is more pronounced.

Table 13 Heterogeneity test results by economic development level.
Full size table

Discussion

The primary findings of this study demonstrate that an escalation in EPU notably increases rural resident subsistence consumption expenditures while concurrently suppressing expenditures on development and enjoyment. This finding aligns with the results of Luttmer and Samwick (2018), who reported that policy uncertainty leads to an increased tendency to save, but further reveals the unique consumption behavior patterns in developing countries. Unlike developed countries, rural Chinese residents exhibit more pronounced “stockpiling behavior,” which may be attributed to (1) a relatively unstable supply in rural markets (Maruejols et al. 2023) and (2) insufficient financial inclusion, leading to a lack of risk management tools (Pham 2024). Additionally, the conclusions of this study support the view of Bergman and Worm (2021) that policy uncertainty has a more significant negative effect on middle- to low-income consumers. Notably, our results differ from those of Morikawa’s (2019) study on urban residents in Japan. Japanese consumers are more inclined to maintain developmental consumption, likely because of their more comprehensive social security system (Adema 2014). This comparison validates the hypothesis of Bairoliya and Miller (2021) regarding the impact of urban–rural disparities in social security on consumption rigidity. Compared with Sahabi’s (2017) research on the impact of policy uncertainty on consumption expenditures, this paper further explores the mediating role of social security. The results show that the presence of social security can significantly mitigate the adverse effects of policy uncertainty on consumption, particularly for rural and low-income groups. This finding is consistent with Wang (2018) and empirically validates the “social safety net” hypothesis proposed by Bergman and Worm (2021). Notably, the use of IV methods (with international crude oil price volatility as the instrument) effectively addresses endogeneity concerns and ensures the reliability of the results (Wei 2019). Furthermore, the threshold effect model in this study reveals that when social security levels exceed a certain threshold, the inhibitory effect of EPU on consumption diminishes, indicating that social security acts as a buffer to reduce the risk perception and promote consumption expenditures. This finding supports the research of Wang and He (2024) and further corroborates the results of Chen et al. (2024) on the impact of social security on consumption behavior in rural China.

This study deepens the theoretical understanding in several dimensions. First, with respect to the mechanism of action, traditional research often emphasizes the singular inhibitory effect of EPU on consumption (Bahmani-Oskooee and Maki 2020), whereas this study identifies significant structural differentiation characteristics. This finding expands the macroeconomic uncertainty theory proposed by Acemoglu et al. (2007) by providing a new perspective for understanding consumption behavior in developing countries. Second, concerning the role of social security, existing studies mostly treat it as a simple moderating variable (Shi et al. 2024), whereas this study reveals its dual roles as both a mediating variable and a threshold variable. This refinement enhances the policy buffer theory framework proposed by Chen et al. (2017).

The findings suggest that EPU significantly affects consumption by rural residents. Policymakers should consider enhancing policy stability and transparency while improving awareness among rural consumers. Additionally, advancing the rural revitalization strategy requires the promotion of industry integration and the development of infrastructure, which will boost production efficiency and generate more job opportunities. Strengthening the social security system and expanding its coverage are also necessary to address the diverse needs of residents. Finally, coordinating regional resource allocation and improving the education environment will enhance rural human capital, increasing the consumption capacity and confidence of residents and ultimately driving growth in consumption in rural areas.

The study acknowledges several limitations. This research primarily concentrates on two broad categories of the consumption structure without examining other potential influencing factors, such as culture, technology, and the environment. Moreover, while the regression analysis revealed a correlation between EPU and consumption by rural residents, establishing a causal relationship requires further investigation.

For future research, broadening the scope to include factors such as the environment, technology, and regional economic shocks to construct a more comprehensive theoretical framework for understanding rural resident consumption behavior is recommended. In addition, investigating how various proxies for social security, including social insurance coverage rates, social welfare expenditures, or direct transfers to households, influence consumer behavior within the context of policy uncertainty would be valuable. Concurrently, as the global economic landscape evolves, examining the differences in consumption behavior in response to policy uncertainty across different countries and regions, as suggested by Bahmani-Oskooee and Maki (2020), is worthy of attention.

Conclusion

In this study, a comprehensive analysis of how EPU influences rural consumption was conducted using baseline regression, mediation effect analysis, and threshold effect analysis for empirical investigation.

First, with respect to the consumption structure, EPU appears to stimulate essential consumption among rural residents while simultaneously suppressing discretionary consumption. Rising EPU can intensify people’s perception of risk, leading them to prioritize essential spending over discretionary purchases. This shift is counterproductive to the increase in consumption by rural residents. Second, the influence of EPU on rural resident consumption expenditures is contingent upon a region’s level of economic development. EPU can increase total consumption by affecting essential spending; however, this positive impact is muted in more economically developed regions and pronounced in less-developed regions. Conversely, the negative impact on discretionary consumption expenditures is mitigated in areas with greater economic development and exacerbated in those with less development. Third, the level of social security serves as a mediating factor in the relationship between EPU and consumption by rural residents. Consumption by rural residents is significantly impacted not only by EPU but also by the level of social security provided. Fourth, EPU has a nonlinear effect on rural resident consumption expenditures. As the level of social security increases, the induced effect of EPU on consumption by rural residents weakens, showing a phenomenon of diminishing marginal effects.