The effect of targeted poverty alleviation on consumption of poor rural households in China

Abstract

Using a survey data of 358 rural households from Jiangxi province of China. This article uses the regression adjusted inverse probability weighted estimation method (IPWRA) to study the impact of precision poverty alleviation on the consumption of rural impoverished households. The empirical results of ATE estimation indicate that precision poverty alleviation has a significant impact on the total consumption and food, clothing, education and living consumption of impoverished households, but has no significant impact on transportation, communication and miscellaneous consumption. Moreover, the poverty alleviation has the largest effect on consumption of education, followed by food.

1. Introduction

Poverty has always been a global issue[ [[1], [2], [3]]]. Despite significant progress in global poverty reduction, there are still a large number of poor people in the world who lack food, clean drinking water, and sanitation facilities. According to Fig. 1, in the early 1980s, nearly 2 billion people lived in extreme poverty, and in 2015, this number decreased to 735 million. Eliminating poverty and achieving prosperity has always been a common goal for all countries, especially developing countries[ [4,5]]. The academic community has given great attention to the definition, theory, measurement standards and types of poverty, as well as anti-poverty strategies [ [[6], [7], [8]]).

Fig. 1.

World population living in extreme poverty

Note:People are considered to live in extreme poverty when living on less than 1.90 international-$ per day.

Sources:OWID based on World Bank (2019) and Bourguignon and Morrisson (2002).

China is the world's largest developing country with a large number of impoverished people [ [9,10]]. Since the establishment of the People's Republic of China, the Chinese government has been committed to reducing poverty. The targeted poverty alleviation strategy has been implemented since 2013, aiming to help the remaining impoverished population lift themselves out of poverty and achieve a moderately prosperous society by 2020. These poverty alleviation efforts have achieved good results. As shown in Fig. 2, as of the end of 2018, the number of rural poor people decreased from 98.99 million at the end of 2012 to 16.6 million, resulting in a total decrease of 82.39 million people; The proportion of impoverished people has decreased from 10.2 % in 2012 to 1.7 %, an overall decrease of 8.5 percentage points.

Fig. 2.

The poverty-stricken population andthe poverty headcount rate

Note: The poverty line increases year by year. 2010 Standard is considered the current rural poverty standard, which is 2300 yuan (in 2010 constantprices) per person per year.2018 standard is 3535 yuan (in 2010 constantprices).

Sources: China National Bureau of Statistics

China has a large number of impoverished people and significant achievements in poverty alleviation, which has attracted widespread attention from the academic community. The relevant research focuses on the poverty standard line[ [11,12]], the definition and types of poverty [ [[13], [14], [15], [16], [17], [18]]], spatial differences in poverty [ [19,20]], and poverty mechanisms [ [20]], Poverty alleviation strategies or policies [ [[21], [22], [23], [24]]], poverty measurement [ [[25], [26], [27], [28]]), economic development and poverty reduction[ [29,30]], targeted poverty alleviation [ [31,32]]. However, the effect of targeted poverty alleviation on the well-being of poor people is little known mainly due to thedata unavailability, especially from the aspect of consumption.

In terms of the policy effects of precision poverty alleviation, scholars believe that precision poverty alleviation policies affect consumption through two main channels. First, it raises the income of poor families. Since the consumption function is concave, low-income groups have a higher marginal propensity to consume[ [[33], [34], [35]],]. Therefore, precise policy on poverty alleviation drives household consumption by increasing household income. Secondly, it reduces the background risk of the family. The policy of precise poverty alleviation, through various means, enhances the employment opportunities for the poor, increases the reimbursement ratio of health insurance, lessens the educational burden on poor families, effectively diminishes the background risk related to income and health for poor families, and subsequently boosts the consumption levels of poor families [ [36,37]]. Based on the above analysis, this paper tests the impact of the policy on household consumption. To better understand the impact of targeted poverty alleviation (TPA), this paper investigates empirically the relationship between TPA and the consumption behavior of rural poor households using a survey data from Jiangxi province.

By the end of 2020, China declared itself completely free from absolute poverty. The significant progress in the fight against poverty underscores the success of targeted poverty alleviation policies, despite the relatively low average proportion of developmental consumption and enjoyment consumption in the total consumption of impoverished farmers in China. Consumption is a crucial catalyst in the process of economic development. It can directly and effectively promote economic growth, and the level of residents' consumption can directly reflect the quality of their current daily lives. So, we aim to assess the effectiveness of the targeted poverty alleviation policy. What impact will the poverty alleviation policy have on rural residents' consumption? Can a precise poverty alleviation policy change residents' consumption patterns and, consequently, boost their intrinsic motivation to achieve sustainable poverty eradication? This study can provide empirical references and data support for the strategy of "rural revitalization" and the enhancement of the consumption level and quality of impoverished households in the "post-2020″ period. It can also help and guide impoverished families in forming reasonable consumption habits, thereby enhancing the living standards and happiness of local underprivileged households. Hence, it is of significant theoretical and practical importance to investigate the effects of precise poverty alleviation policies on the consumption patterns of impoverished rural families.

The rest of this article is organized as follows: The next section introduces the theoretical analysis between poverty alleviation and consumption; The third section introduces empirical methods and survey data; the forth part provides empirical results; The fifth part is the research conclusion.

2. Theoretical frameworks

Currently, China has completed its precision poverty alleviation efforts, and the research focus has now shifted to evaluating the impacts of poverty alleviation. This evaluation primarily focuses on the impact of poverty alleviation policies on society, families, and individuals. For example, Xu and Gao studied the impact of precision poverty alleviation on the economic development of impoverished counties using the double difference method. They found that precise poverty alleviation significantly promoted local economic development, improved the local financing environment, and optimized the local industrial structure [38]. Liu and wang (2020) found that precision poverty alleviation is effective in increasing the per capita net income of poor rural families, but the actual effect demonstrates marginal diminishing returns [39]. Han et al. (2020) utilized the Propensity Score Matching with Difference in Differences (PSM-DID) method and discovered that the targeted poverty alleviation policy significantly increased the income levels of impoverished farming households through agricultural activities [40]. Xiao and Yan (2012) examined the level of satisfaction among rural impoverished groups with the existing poverty alleviation policy by assessing their understanding of the policy [41]. They used this satisfaction measure to assess the impact of poverty reduction initiatives. Zhang and Zhou (2019) used the poverty line as a threshold and found that the implementation of targeted poverty alleviation policies significantly reduced the per capita cost of living for impoverished rural households [36]. Wang and Xu (2019) argued that targeted poverty alleviation can increase the per capita income of families, reduce the incidence of poverty, and have a sustainable effect. By combing through the literature, it is evident that most studies are based on empirical analysis, generally demonstrating the effectiveness of precision poverty alleviation in enhancing the income level of poor families [42].

Based solely on the income criterion, academics' extensive study on poverty has revealed significant limits in assessing the impact of programs on reducing poverty. To assess the impact of policies on reducing poverty more thoroughly, several researchers incorporated the consumption component [ [43,44]], which led to improved recommendations for the precise identification and management of poverty [ [45,46]]. By applying the double difference method test, Yin and Guo (2021) discovered that the specific strategy for reducing poverty encouraged the growth of impoverished households' overall consumption [47].Xu et al.(2019) discovered that precision poverty alleviation encourages the increase of income and consumption of the poor. The evidence reviewed above indicates that the great majority of academics support the idea that policies aimed at reducing poverty through precision can encourage farm households to consume more [37].

Generally speaking, there are still few empirical studies of the policy benefits of precision poverty alleviation, and those that do exist have mostly concentrated on the effects of specific policies on income. The study that is currently available, however, has not given enough consideration to how pro-poor policies' ability to increase the income of impoverished households influences their consumption structure and behavior.

In order to explore the mechanism how poverty alleviation affects the consumption of impoverished population by which precisions, this study theoretically elucidates the relationship between precision poverty alleviation and the consumption behavior of rural impoverished households. The aim of poverty alleviation is to increase the income of poor households and improve their living standards. There are many poverty alleviation measures, which can be divided into two categories: one is to improve the social security of the poor population through government transfer payments to meet their basic living needs, mainly including medical care, education and housing, and the other is to help poor population increase their income, mainly including employment training, industrial subsidies, employment opportunities and financial support, etc. Many studies have documented well that the improvement of social security positively affect consumption through savings, income growth, investment behavior and assets allocation [ [48,49]].Therefore, the first category of poverty alleviation measures has an indirect positive effect on consumption. The second group of poverty alleviation measures leads to an increase the income of poor people. According to the classic consumption theory, consumption increases with income. Hence, the second group of measures has a direct positive impact on consumption. To sum up, we can claim that poverty alleviation measurescan stimulate the consumption of poor people as illustrated in Fig. 3.

Fig. 3.

The mechanism between poverty alleviation measures and consumption.

3. Data and method

3.1. Data

The survey data used in this paper are collected from southern Jiangxi province. This area was the core area of the former Central Soviet Area during the China Agrarian Revolutionary War. Since the founding of New China, the Southern Jiangxi Soviet Area has gradually lost the advantages of transportation and industry, and formed a large concentrated and fragmented poverty-stricken area. China's definition of rural poverty is based on strict criteria, and the poverty line has been raised year after year, from 2300 yuan per person per year in 2010 to 3535 yuan in 2018. According to this standard, we classify households with annual per capita income below the poverty line as poor, and those above the poverty line as non-poor. The data released by the Ministry of Finance of Jiangxi Province shows that by the end of 2017, there are 269 poverty-stricken villages in Jiangxi Province, of which nearly half of the population lives in southern Jiangxi. (The survey questionnaire received verbal consent from all participants. The investigation project has been approved by the Academic Committee of Gannan Normal University with approval number GJJ160959). We used a wide range of national and international questionnaire design techniques on the same subject when creating the questionnaire; the specifics are in the Appendix.

The survey was conducted between August 2017 and January 2018. A three-stage sampling procedure was used to collect data. Firstly, Xingguo County, which has a high density of impoverished population in Gannan, was chosen. Secondly, 30 villages distributed in the county were randomly selected. Finally, based on the farmer information provided by the village committee, approximately 20–30 households in each village were randomly selected to form data on 561 representative households (impoverished and non-impoverished).

Before the investigation, we conducted training for the investigators and conducted a pilot test to redesign some survey questions. A detailed and structured questionnaire survey was conducted, with face-to-face interviews conducted by census takers and supervised by one of them. Data on household demographic characteristics, housing, transportation, income, consumption structure, social security, and cultural life were collected through interviews with household heads.

In order to improve the data quality, we filter the data by following methods. First, the outliers who aren't in conformity with normal conditions are deleted. Second, the observations with consistent answers for all questions are dropped. After the filter, the final dataset is composed of 358 households including 211 poor households and 147 non-poor households.

3.2. Method

To establish the relationship between targeted poverty alleviation and consumption of poor households, we assume that the consumption can be expressed as a function of a poor household dummy ($D_i$),a vector of explanatory variables ($X_i$)and an error term (${\sigma }_i$),as indicated in equation (1).

$$Y_i=f\left(D_i,X_i\right)+{\sigma }_i$$ (1)

Where $Y_i$ represents the family consumption expenditure. ${\mathit{D}}_{\mathit{i}}$ is equal to 1 if a family is identified as poor household, and zero otherwise.

Given that impoverished families are not randomly selected, they are determined by the government based on the family situation. In this case, the use of OLS method may lead to selective bias in the estimation results. In order to address the issue of potential selectivity bias, this article adopts the regression adjusted inverse probability weighted estimator (IPWRA) method instead of the propensity score model (PSM), as the IPWRA estimator has dual robustness to ensure consistent results, which can not only reduce bias caused by specification errors in PSM, but also solve the problem of selectivity bias. IPWRA estimation can be carried out in three steps. Firstly, use the following Popit model to calculate propensity scores,as indicated in equation (2).

$${\hat{p}}_i\left(X_i\right)=\mathrm{Pr}\left(\mathrm{D}=1|X_i\right)=\frac{\mathrm{Exp}\left(\hat{\delta }{x}_i\right)}{1+\mathrm{Exp}\left(\hat{\delta }{x}_i\right)}$$ (2)

As the name suggests, inverse probability weighting refers to weighting the results by the reciprocal of the individual's probability of receiving treatment. Assuming that the resulting model is given by a formal linear regression function,as indicated in equation (3).

$$Y_i={\alpha }_i+{\phi }_i{Z}_i+{\sigma }_i$$ (3)

Where $Z_i$ includes the poor households dummy ($D_i$)and the vector of the explanatory variable (X_i). Then, we use linear regression to estimate coefficients using inverse probability weighted least squares $\left(\alpha ,\phi \right)$,as indicated in equations (4), (5).

$${\min }_{a_{0,}{\phi }_0}\sum _i^N\frac{\left(Y_i-{\alpha }_0-{\phi }_0{Z}_i\right)}{\left(1-{\hat{p}}_i\left(X_i\right)\right)}if{D}_i=0$$ (4)

$${\min }_{a_{1,}{\phi }_1}\sum _i^N\frac{\left(Y_i-{\alpha }_1-{\phi }_1{Z}_i\right)}{{\hat{p}}_i\left(X_i\right)}if{D}_i=1$$ (5)

Thus, the final step is to calculate the average treatment effect on treated (ATE) as indicated in equation (6).

$${\text{ATE}}_{IPWRA}=E\left(Y_1-Y_0\right)=\frac{1}{N_w}\sum _i^{N_w}\left[\left({\hat{\alpha }}_1-{\hat{\alpha }}_0\right)-\left({\hat{\phi }}_1-{\hat{\phi }}_0\right)Z_i\right]$$ (6)

Among them $\left({\hat{\alpha }}_1,{\hat{\phi }}_1\text{,}\right)$ is the estimated inverse probability weighted parameter of the result function for impoverished households. $\left({\hat{\alpha }}_0,{\hat{\phi }}_0\text{,}\right)$ is a parameter estimated by the inverse probability weighted least squares method for non-impoverished households. $N_w$ represents the number of impoverished households.

3.3. Variable selection and description

In our survey data, the dependent variable, the consumption of poor rural households, mainly includes food, clothing, transportation, education, communications, living expenses and other. The consumption of each item is measured by the annual household expenditure. In terms of control variables, this paper includes householder characteristics and family characteristics. The cohort of householder characteristics contains 3 variables, namely age, health situation and education. The array of family characteristics has 4 items including gross household income, per capita income, land assets and household size. Table 1 defines and statistically describes these variables.

Table 1.

The definitions and summary statistics of the variables.

Variable	Description	Mean
Dependent variables
Total Food	Annual total expenditure Annual expenditure on food	32369.97 13880.9
Clothing	Annual expenditure on clothing	516.22
Transportation	Annual expenditure on transportation	3867.99
Education	Annual expenditure on education	1067.81
Communication	Annual expenditure on communication	2659.04
Living	Annual expenditure on fees for living	1740.26
Miscellaneous	Annual expenditure on other items	8404.46
Independent variables
Poverty	1 if poor household is identified, 0 otherwise	0.58
Age	Age of household head. 1 = age<25; 2 = [26,30]; 3 = [31,35]; 4 = [36,40]; 5 = [41,50]; 6 for age>50	5.26
Education	Education level of household head. 1 = illiteracy; 2 = primary school; 3 = junior middle school; 4 = high school; 5 = college and above	2.66
Health	Health situation of household members. 1 = healthy; 2 = minor ailments; 3 = major diseases	2.02
Land Household size	Land size cultivated in mu (1 mu = 1/15 ha) The number of household members	3.32 4.56
Income	Annual household income	35303.83
Per capital income	The ratio of household annual income to the number of household members	8839.72

As shown in above table, the average annual total expenditure of rural households is 32369.97 yuan. Regarding the consumption structure, the proportion of food is the highest on average, followed by miscellaneous and transportation, clothing expense is the lowest. In our survey data, poor households accounted for 58 % of the total sample. The average age of household head is over 40 years old with low average education. The per capita cultivated land is very low, less than 1 mu. According to data released by NBS, the per capita income of rural households in China was 13432 yuan in 2017.Whilethe per capita income of the sample households was 8839.72, much lower than national level in the sample year. Table 2 reports the mean difference in selected variables between poor households and non-poor households.

Table 2.

Mean difference in selected variables between poor and non-poor households.

Variables	Non-poor	Poor	Mean Diff
Dependent variables Total	38434	36323	2110.86***
Food	16338	15070	1268.31***
Clothing	577.4	470.8	106.60*
Transportation	5183	3465	1718.03***
Education	983.4	1135	−151.1*
Communication	3398	2124	1273.52***
Living	2065	1499	565.88**
Others	8975	7424	1550.76***
Independent variables Age	5.160	5.330	−0.17*
Education	2.130	1.900	0.22**
Health	1.690	2.280	−0.59***
Land	2.63	3.43	−0.8**
Household size	5.070	4.200	0.86***
Income	39405	31890	7515.31*
Per capita income	9175	8557	618.3*

Note: *, **, and *** represent statistical significance at the 10 %, 5 %, and1% levels, respectively.

The above table shows that poor and non-poor households are systematically different in terms of some characteristics. For instance, relative to non-poor households, poor households have a much less expense on all kinds of items except education. However, this result cannot be used to make inferences regarding the impact of targeted poverty alleviation on consumption of rural households without controlling for other confounding factors. With regard to independent variable poor householders are older and worse educated that non-poor householders. Moreover, the poor households have worse health situation, larger cultivated land and fewer members. At last, as expected, both income and per capita income of poor households are lower than that of non-poor households.

4. Empirical results and discussion

4.1. Determinants of being poor household

Table 3 reports the parameter estimates of the determinants of being a poor household by the government using Probit model. Both coefficients and marginal effects are reported for better interpretation. The positive and statistically significant coefficients and marginal effects of the health situation and land size suggest that households with worse health situation and larger land size are more likely to be identified as poor households. However, the household size, income and per capita income have negative and statistically significant effects on the probability of being poor households, implying that households with more members and higher income are less likely to become poor households.

Table 3.

Determinants of poor households: probit model.

	(1)	(2)
Variables	Coefficients	Marginal effects
Age	0.08	0.03
	(0.09)	(0.03)
Education	0.03	0.01
	(0.08)	(0.03)
Health	0.51c	0.20c
	(0.10)	(0.04)
Land	0.08c	0.03c
	(0.03)	(0.01)
Household size	−0.25c	−0.10c
	(0.07)	(0.03)
Income	−0.04b	−0.01b
	(0.03)	(0.01)
Per capita income	−0.01a	−0.004a
	(0.04)	(0.02)
Constant	−0.43
	(0.59)
Observations	358	358

Notes: Robust standard errors in parentheses.

^ap < 0.1.

^bp < 0.05.

^cp < 0.01.

4.2. Matching quality

Before analyzing the effectiveness of policies, we check the quality of matching, as the reliability of IPWRA results determines the quality of our matching. Firstly, Fig. 4 shows the common support of propensity score distribution and nearest neighbor matching. A visual examination of the estimated propensity score distribution of households receiving policy support and those not receiving policy support indicates that the hypothesis of shared support is met. Next, Fig. 5, Fig. 6 shows the kernel density of propensity scores before and after matching. From the graph, we can see that there is a significant difference between the two groups before matching. However, after matching, the nuclear densities of the two groups were similar. Finally, the standardized mean difference for all covariates from the overall covariate balancing test reduces from 30.5 % pre-matching to 9.1 % post-matching, which is in line with the suggestion of Rosenbaum and Rubin (1983) that the standard deviation should be less than 20 %.In addition, the joint significance of all covariates was rejected before matching (p > chi2 = 0.000), but was rejected after matching, as shown by likelihood ratio test (p > chi2 = 0.717). To test the successful balance of the distribution of covariates between treated and untreated households through low mean deviation and joint non significance of covariates.

Fig. 4.

Propensity score distribution and common support.

Fig. 5.

Kernel density of propensity score before matching.

Fig. 6.

Kernel density of propensity score after matching.

4.3. Treatment effects of poverty alleviation

Table 4 reports the IPWRA results with regard to the treatment effects of the targeted poverty alleviation on the consumption of poor rural households. The average treatment effects (ATE),untreated potential outcome means (POM), treated POM and growth rate are reported. In our study, the ATE measures the difference between the average expenditure if all households were to be identified as poor households and the average expenditure if none of households had been identified as poor households. The value of ATE is calculated as the difference between treated POM and untreated POM. The growth rate is computed as the ratio of treatment effects to untreated potential outcome means.

Table 4.

Treatment effects of poverty alleviation on consumption.

	Treated POM	Untreated POM	ATE	ATE(%)
Total	37,992c (1539)	31,697c (2386)	6295b (2219)	19.8 %
Food	15,586c (1222)	13,056c (1525)	2530a (1529)	17.9 %
Clothing	444.8c (53.63)	409.1c (77.51)	35.79a (88.76)	8.7 %
Transportation	4531c (711.3)	5787c (1167)	−1256.0 (1200)	−21.7 %
Education	1208.3b (222.4)	712.6b (313.4)	495.7b (213.6)	69.6 %
Communication	2983c (262.2)	3189c (412.7)	−529.9 (437.8)	−16.6 %
Living	1818c (163.1)	1626c (231.2)	192.7a (277.2)	11.8 %
Miscellaneous	6527c (704.8)	9605c (802.4)	−3078 (1169)	−32.0 %

Notes: Robust standard errors in parentheses.

The percentages are computed as the ratio of treatment effects to potential outcome means.

^ap < 0.1.

^bp < 0.05.

^cp < 0.01.

As can be observed from Table 4, the positive and significant ATE estimate in the first line show that the implementation of the targeted poverty alleviation policy has a significant positive impact on the total expenditure of rural households.To be exact, poverty alleviation measures increase the total annual consumption of poor households by 6295(19.8 %), which is equivalent to spending 524 more per month. This result is consistent with the conclusions of most academic studies on precision poverty alleviation policies and rural household consumption, such as Liu (2021), who utilized CFPS 2010–2018 data and found that precision poverty alleviation can promote the increase of rural household consumption [50]. Hu (2014) further found that transfer income can promote the increase of all consumption expenditures of rural residents, but this promotion shows a nonlinear threshold phenomenon with the change of income level [51].

However, the effect of poverty alleviation on the consumption of poor households differs with consumption categories. In particular, the poverty alleviation measures have significant impact on food, clothing, education and living consumption, but have no significant impact on transportation, communication and miscellaneous consumption. The ATE estimates reveal that the causal effect of poverty alleviation was to increase food consumption by 2530(17.9 %), clothing consumption by 35.79 (8.7 %), education consumption by 495.7 (69.6 %), and living consumption by 192.7 (11.8 %) on average, respectively. The results show that poverty alleviation measures have the largest effect on education, followed by food, which can possibly be explained by the direction of poverty reduction that is basic needs, like food and education, should be satisfied in priority.

The results obtained may exhibit certain biases compared to the research conclusions of some scholars. For instance, Bing and Zhao (2019) argued that the precision poverty alleviation policy has the greatest effect on promoting residential and educational expenditures, followed by communication, other, and clothing expenditures, with food and transportation expenditures coming last [52]. Gong (2013) also believed that precision poverty alleviation mainly promotes non-food consumption [53]. However, many scholars also agree that poverty alleviation policies significantly stimulate expenditures on food and other items [54]. This may be due to differences in the research methods and data employed by various scholars, leading to certain biases in the results obtained.

4.4. Robustness checks

In order to offer more evidences to verify the relationship between targeted poverty alleviation and consumption of poor households, we execute several robustness checks including propensity score match (PSM) estimates, regression adjustment (RA) estimates, inverse-probability weights (IPW) estimates and augmented inverse-probability weights (AIPW)estimates as shown in Table 5.

Table 5.

The results of robustness checks.

	PSM	RA	IPW	AIPW
Total	8037b (2618)	7329c (2560)	6581c (2343)	7948c (2470)
Food	4629c (1391)	3988c (1015)	4013c (1080)	3939a (1039)
Clothing	38.66 (128.18)	4.22b (84.66)	21.58a (80.01)	27.42a (81.28)
Transportation	−1576 (581.9)	−1485 (596.6)	−1357 (526.1)	−1306 (538.1)
Education	575.6c (408.67)	339.9b (321.9)	375.9b (284.8)	387.2b (289.3)
Communication	−611.2 (474.5)	−596.4 (288.1)	−549.7 (293.1)	−540.9 (295.2)
Living	268.98a (180.7)	224.5b (171.8)	221.9a (187.9)	259.5a (175.8)
Miscellaneous	−1667 (2220)	−3671 (1354)	−3711 (1369)	−3713 (1407)

Notes: Robust standard errors in parentheses.

ATEs are reported in each column.

^ap < 0.1.

^bp < 0.05.

^cp < 0.01.

As shown in Table 5, we find consistent results between RA, IPW, AIPW and IPWRA methods as suggested by their close average treated effects. However, the average treated effects estimated by RA, IPW, AIPW and IPWRA are slightly lower than that estimated by PSM. This finding suggests that in our case, the misspecification may result in a positive bias, leading to overestimated ATEs in the PSM model estimations.

5. Conclusion

This article uses survey data from 358 households in Xingguo County, Jiangxi Province to study the impact of targeted poverty alleviation policies on the consumption of rural impoverished households. We used the Inverse Probability Weighted Estimator (IPWRA) method with regression adjustment in the baseline analysis to explain potential selection bias and model mis designation bias caused by systematic differences between impoverished and non impoverished households. We also performed several robustness checks on the baseline results, including propensity score matching (PSM), regression adjustment (RA), inverse probability weighting (IPW), and enhanced inverse probability weighting (AIPW).

The empirical results of ATE estimates indicate that the targeted poverty alleviation measures have a significant impact on the consumption of poor households in terms of total consumption and food, clothing, education and living consumption, but have no significant impact on transportation, communication and miscellaneous consumption. In particular, the causal effect of poverty reduction is on average to increase the total consumption of poor households by 6295 (19.8 %),food consumption by 2530(17.9 %), clothing consumption by 35.79 (8.7 %), education consumption by 495.7 (69.6 %), and living consumption by 192.7 (11.8 %), respectively. We find poverty alleviation measures have the largest effect on education, followed by food, which can possibly be explained by the direction of poverty reduction that is basic needs, like food and education, should be satisfied in priority.

Most of the academic research on the precise poverty alleviation policy and the consumption of poor rural households has used microdata, and due to the differences in the research methods and research data used by various scholars, the results obtained have a certain degree of bias, such as the impact of the poverty alleviation policy on various consumption categories. However, the vast majority of scholars agree that the precise poverty alleviation policy can promote the increase in consumption of poor farmers. In this paper, survey data from 358 households were used, which is a small sample size. Although it has obtained research conclusions consistent with those of academics, it has also limited the in-depth study of the relationship between precision poverty alleviation and household consumption, and it is hoped that more research data will be available in the future to facilitate a more in-depth analysis of this topic.

Funding disclosure

This research was supported by Jiangxi Provincial Department of Education Science and Technology Project(Grant No.GJJ160959)

Data availability statement

The data is available from the corresponding author upon request.

CRediT authorship contribution statement

Gu Qunfang: Resources, Investigation, Supervision, Validation, Writing – original draft. Huang Xiaobing: Formal analysis, Data gathering, Conceptualization, Writing – review & editing.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following is the Supplementary data to this article.