Introduction

The efficacy of livelihood initiatives stands as a pivotal benchmark for evaluating governing philosophy, governance capabilities, and overall governance effectiveness (Evans and Karras, 1994; Pan et al. 2024). In pursuit of garnering more votes, politicians often pledge to bolster livelihood investment during election years (Lyu et al. 2018). However, central governments typically assess and promote local officials based on their performance over their tenure. Local officials may prioritize maximizing their political interests by manipulating indicators such as revenue growth (Ferreira et al. 2013), creating a cyclical relationship between economic indicators and policy decisions. This intriguing phenomenon prompts us to explore its underlying causes: how does this manipulation affect the welfare of voters?

The public’s access to the benefits of economic growth hinges on government investments in livelihood. During the policy formulation and implementation process, governments often face trade-offs between fostering economic growth and enhancing people’s well-being (Khalifa, 1998). On one hand, to drive economic development, local officials may allocate limited financial resources to projects yielding immediate returns (Arthur, 2008; Guo et al. 2021), such as infrastructure and manufacturing investments, while neglecting essential public goods that do not provide an immediate and substantial boost to economic growth, such as healthcare, education, and social security. On the other hand, local governments may adopt expansionary fiscal policies to stimulate economic growth, potentially leading to a vicious circle of future regional economic activity. Fiscal overextension increases the risk of future government debt defaults and incurs high-interest payments. Consequently, politicians may significantly reduce livelihood investment if local governments accumulate substantial debt during their tenure.

Hence, this false economic growth may merely serve as a political tool for politicians, rather than a genuine source of prosperity for residents. GDP serves as a crucial indicator for assessing the economic status and developmental trajectory of a country or region, with its manipulation reflecting economic distortion to some extent. Politicians often manipulate GDP data to showcase good governance performance, thereby leveraging political gains and advancing their careers (Ferreira et al. 2013). While prior research has examined the influence of GDP manipulation on corporate behaviors (Li et al. 2020; Cai et al. 2022; Lin et al. 2021), it remains unclear whether this manipulation extends its impact on public welfare. Given the trade-off between economic development and livelihood investment, the credibility of GDP can partially reflect the extent of local government’s efforts in improving people’s livelihoods.

The accuracy of official GDP statistics has emerged as a pivotal concern (Chow, 2006; He and Sun, 2013; Holz, 2014; Perkins and Rawski, 2008; Klein and Ozmucur, 2003; Clark et al. 2017). Errors within macroeconomic data, including GDP, stem from various sources, such as measurement inaccuracies (Grimm and Weadock, 2006) and the intricate nature of accounting procedures. Compared with developed countries, the accounting methodologies and statistical frameworks for GDP in developing countries tend to be less advanced. Some low-income countries exhibit severely undeveloped national economic statistical accounting systems, or, in extreme cases, lack credible census data altogether, thereby rendering economic statistics at a rudimentary level (Chen and Nordhaus, 2011) and increasing susceptibility to manipulation.

China, as the largest developing country globally, offers a compelling context for examining this phenomenon for three primary reasons. First, China has faced the significant controversy regarding the reliability of official GDP data (Wu, 2007; Angus, 1998; Cai, 2000; Rawski, 2001). Second, China’s development in the social service industry lags behind its economic growth compared to developed countries. Data indicates an inadequacy in fiscal expenditures towards livelihood enhancement in China, highlighting a need for improvement in people’s overall life satisfaction (Li and Raine, 2014; John and Ramani, 2011). Third, the political promotion system in China serves as an effective incentive model for officials. The enthusiasm exhibited by local governments towards livelihood investment warrants an in-depth investigation. Moreover, the fiscal decentralization system and official promotion mechanism predominantly rely on GDP as a primary evaluation criterion in China (Montinola et al. 1995; Jin et al. 2005), thereby providing a natural research environment.

This paper investigates the influence of regional GDP manipulation on residents’ welfare. Utilizing calibrated satellite light data as proposed by Henderson et al. (2012), this study estimates the actual level of economic development across various regions. The findings indicate that regional GDP manipulation leads to a reduction in livelihood investment within the jurisdiction. Notably, regional GDP manipulation exerts a more pronounced effect on livelihood investment during the initial stage of officials’ tenure. However, this effect diminishes when officials operate in their hometowns. Further analysis reveals that the adverse effects of economic distortion exacerbate the contraction in livelihood investment during economic downturns (2008–2013).

Our contribution to research in this field is twofold. Firstly, it expands upon the current literature on livelihood investment by incorporating the influence of regional GDP manipulations. This paper offers fresh perspectives on livelihood input challenges, encompassing not only an evaluation of the efficacy of local government interventions but also an examination of the economic and political drivers behind these interventions. Secondly, our study enriches the literature on regional GDP manipulations. While prior research has delved into the effects of regional GDP manipulations on micro-level indicators such as corporate tax avoidance (Li et al. 2020), earnings management (Cai et al. 2022), corporate investment efficiency (Li et al. 2024) and analyst forecast accuracy (Lin et al. 2021), this study extends the scope of economic distortion analysis to the macro level by investigating the composition of local fiscal expenditures.

Literature review and hypothesis development

China exhibits a relatively centralized political power structure enabling the central government to realize its economic objectives through political incentives (Blanchard and Shleifer, 2001). Regional GDP growth rates play a pivotal role in the promotion mechanism of Chinese officials (Qian and Weingast, 1996). Therefore, the accuracy of China’s GDP data has been a subject of considerable scholarly inquiry (Bu, 1999; Sinton, 2001; Rawski, 2001; Movshuk, 2002). Shiau (2005) identified four primary sources of error in estimating China’s GDP, encompassing data distortions, inaccurate conversions, services underestimation, and sampling errors. Lyu et al. (2018) corroborated the manipulation of economic growth data by local governments using the breakpoint method, revealing significant discontinuity between real GDP growth rates and target growth rates.

The trade-off between economic growth and residents’ welfare needs leads to resource misallocation, as the pressure to achieve economic growth supersedes the necessity to develop social welfare programs. Regional economic development and officials’ promotion are closely intertwined (Bo, 1996; Li and Zhou, 2005). When officials prioritize GDP manipulation to enhance their prospects of promotion, livelihood investments may be neglected. China’s GDP manipulation is a subject of significant controversy (Perkins and Rawski, 2008; Klein and Özmucur, 2003; Holz, 2014; Clark et al. 2017). Some literature argues that China’s GDP statistics are overestimated. Meanwhile, inflated GDP figures can be deceptive, while livelihoods require genuine economic support. GDP manipulation can have an impact on regional government’s fiscal revenue projections, potentially resulting in excessive fiscal spending that exacerbates fiscal imbalances. Consequently, these imbalances can curtail the government’s capacity to invest in vital sectors such as education, healthcare, and social security.

GDP manipulation underscores a discernible disparity between local economic performance and anticipated targets, thus engendering heightened pressure on economic development in subsequent years. Beyond merely attaining growth objectives, local economies are compelled to augment their economic growth quotas to offset shortfalls from preceding years. The severity of GDP manipulation correlates with the magnitude of the economic development deficit requiring rectification, thereby intensifying the strain on economic development efforts. Consequently, we propose the following hypothesis:

Hypothesis 1: The greater the extent of regional GDP manipulation, the lower the investment in livelihoods.

There are three primary motivations for government intervention in economic data: mitigating recession (Wallace, 2016), gaining popular support to secure electoral success (Healy and Lenz, 2014), and maintaining social stability (Hollyer et al. 2015). Shi and Svensson (2006) empirically tested the 21-year data of 85 OECD countries and found a tendency for governments to increase fiscal deficits during election years, with a rise of approximately 1% compared to non-election years, indicating a more pronounced occurrence of fiscal manipulation. This phenomenon is particularly prominent in developing countries (Shi and Svensson,2006).

Public goods, such as education and social services, typically exhibit a delayed impact on economic growth. This temporal gap frequently results in a gradual erosion of governmental prioritization of public goods throughout its tenure. Officials bear full responsibility for their assigned tasks throughout their tenure in China (Xue et al. 2023). Brief tenures often result in short-sighted decision-making, whereas excessively lengthy tenures may shift their governance objectives. Even if these officials are committed to people’s livelihoods, the allure of economic growth rewards and their enthusiasm for economic development can eclipse this sense of accomplishment in prioritizing people’s livelihoods. Through an analysis of data encompassing Chinese officials across 30 provinces and cities from 1998 to 2013, we find that the average tenure spans merely 3.17 years. Once an official’s tenure exceeds this average, the probability of their promotion diminishes. Consequently, there is a reduced likelihood of fiscal spending cuts in livelihood sectors. The inverse relationship between regional GDP manipulations and investment in livelihood initiatives becomes less conspicuous in the latter half of an official’s tenure. Therefore, we posit the following hypothesis:

Hypothesis 2: The impact of regional GDP manipulation on livelihood investment is influenced by political cycles, particularly during the early stages of local officials’ tenures, where the negative correlation between the two variables is more significant.

People’s livelihood encompasses the overall framework of social existence and represents the state of social life under national governance. Various factors influence government livelihood investment, including public spending (Aschauer, 1989), budget deficits (Alesina and Perotti, 1996), corruption (Pak, 2001), and state capacity (Besley and Persson, 2010). The provision of livelihood services in China remains unsatisfactory despite China’s economy having achieved remarkably sustained high-speed growth since the 1980s (Blanchard and Shleifer, 2001; Maskin et al. 2000). There persist numerous challenges in areas such as healthcare, education, and social security. The underlying cause of this phenomenon can be traced to the control and safeguarding of local markets and resources by regional governments, as well as the officials’ strong capacity and motivations for regional economic development (Montinola et al. 1995; Jin et al. 2005). When the government undergoes frequent official turnover, local officials tend to prioritize short-term, high-impact, and swift-growing projects. In their pursuit of showcasing achievements, officials at all levels strive to surpass their targets during their limited tenure.

Geographical identity exerts a significant influence on an individual’s group behaviors (Akerlof and Kranton, 2000). Hodler and Raschky (2014) revealed that the luminosity of satellite lights in national leaders’ hometowns exhibited a marked increase during their tenure. This phenomenon is globally referred to as “regional favoritism”. In China, numerous studies have underscored the prevalence of regional favoritism among provincial governors and party secretaries, who consistently display a propensity towards prioritizing development initiatives within their respective hometowns. As these officials wield authority over the management and allocation of public resources, their sentimental attachment to their places of origin often motivates a concentrated focus on comprehensive hometown development, particularly in terms of enhancing local livelihoods. Consequently, we propose the following hypothesis:

Hypothesis 3: The impact of regional GDP manipulations on livelihood investment is contingent upon the official’s place of origin. Specifically, when officials operate within their hometowns, the adverse correlation diminishes in strength.

Research design

Sample selection and data sources

This paper uses 30 provinces and cities in China (excluding Hong Kong, Macao, Taiwan, and Tibet) as the research sample. The scope of livelihood investment is defined as education expenditures, social and employment security expenditures, and medical expenditures. Due to the comprehensive reform of government revenue and expenditure classification in 2007, major changes have taken place in financial accounting subjects. The calibre adjustments have been made. Specifically, education expenditure corresponds to the subject of “Education Expenses”, social security and employment expenditure corresponds to the subjects of “Social Security Subsidy Expenditure” and “Compensation and Social Welfare Expenditure”, and medical and health expenditure corresponds to “Health Expenditure”.

The DMSP/OLS night light data released by the National Geographic Data Center of the National Oceanic and Atmospheric Administration (NOAA) spans from 1992 to 2013. Because the provincial fiscal expenditure data were missing from 1992 to 1997, the final sample period is set as 1998–2013.Footnote 1 The provincial-level economic data and official characteristic data in this paper come from the China Securities Markets and Accounting Research Database and the China Statistical Yearbook. Some missing data are supplemented by the provincial statistical yearbooks.

Calibration of night light data

The original DMSP/OLS night light data presents several challenges that need to be addressed through calibration. Firstly, a notable issue is the upper limit of the data value. With a saturation value of 63 for night light brightness data, this threshold proves inadequate for capturing the radiance of robust economic activities. As urbanization progresses and populations concentrate in major cities, the accuracy of observations diminishes over time (Baum-Snow et al. 2017). In China, particularly in the eastern coastal regions, the brightness of light has far exceeded the saturation level, resulting in significant observational errors and compromising empirical results.

To mitigate this, NOAA has recently released correction data for night light radiation, which eliminates the saturation effect. However, the data is discrete in terms of years. Following Hu et al. (2022), we use this correction data to address saturation issues in stable light data. Given the gradual nature of urban development and the relative stability of city layouts in the short term, we replace lighting layouts in saturated areas with data from nearby years. Leveraging the credibility of non-saturated data, we establish relationships between stable light data and radiation correction data in non-saturated areas, applying these relationships to restore stable light data in saturated regions.

The second challenge is the interannual discontinuity of light data. The replacement of new and old satellites may cause differences in information collection, resulting in incomparable data in some years. For the interannual inconsistency of the DMSP-OLS stable light data, we utilized the pseudo-invariant feature (PIF) calibration method. Based on geographical location, uniform range distribution, and distance from the mainland, Wu et al. (2013) selected Mauritius, Puerto Rico, and Okinawa as PIF regions. They used the radiance-calibrated data from 2006 as the reference image and applied a power function model for correction. Due to the simplicity of Wu’s method, the CCNL and PCNL datasets adopted it for the interannual correction of the original DMSP-OLS images. The regression model is as follows:

$${{DN}}_{c}+1=a{\left({{DN}}_{m}+1\right)}^{b}$$

where DNc is the corrected pixel value, DNm is the original pixel value, and a and b are the model coefficients. Wu et al. (2013) provided correction factors for the years 1992 to 2010, while Zhao et al. (2022) calculated correction factors for the remaining years using a similar approach.

Furthermore, data processing from two satellites in the same year can yield incomparable observations. In such instances, we employ the approach proposed by Elvidge et al. (2009), estimating the following equation to reconcile these disparities.

$${{DN}}_{i,t}=\left\{\begin{array}{c}0,{{DN}}_{i,t}^{1}\times {{DN}}_{i,t}^{2}=0\\ ({{DN}}_{i,t}^{1}+{{DN}}_{i,t}^{2})/2,{{DN}}_{i,t}^{1}\times {{DN}}_{i,t}^{2}\ne 0\end{array}\right.$$
(1)

Where DN refers to the digital number of the grid brightness value.

Lastly, the assumption of consistent growth presents a challenge as satellite sensors age and photographing effects worsen, potentially leading to fluctuations in light data. Following Elvidge et al. (2009), we address it by correcting grids showing a decrease in radiance over time, assuming early high values are likely due to noise. For consistently stable and bright grids, correction is only made if the subsequent year is dimmer than the previous one, otherwise, the high DN value of the previous year is retained.

$${{DN}}_{i,t}=\left\{\begin{array}{c}0,{{DN}}_{i,t+1}=0\\ {{DN}}_{i,t-1},{{DN}}_{i,t+1} > 0\,{and}\,{{DN}}_{i,t-1} > {{DN}}_{i,t}\\ {{DN}}_{i,t+1},{other}\end{array}\right.$$
(2)

The situation before and after light data processing is shown in Fig. 1.

Fig. 1: National average satellite light DN value.
figure 1

This figure reports the national average satellite light DN value before and after night light data calibration. Specifically, we solve the problem of the upper limit of the data value, the problem of more than one observation in the same year, the noise of interannual discontinuity and the assumption of consistent growth during the data processing.

Full size image

Measurement of regional GDP manipulation

This study aims to address the challenge of objectively measuring the authenticity of local economies. Economists have widely used satellite nighttime light data for measuring real growth in regional economies (Chen and Nordhaus, 2011; Henderson et al. 2012; Hodler and Raschky, 2014; Nordhaus and Chen, 2015). The incorporation of global nighttime light data into economic research was pioneered by Henderson et al. (2012). The growth rate of night light data exhibits strong positive relevance with GDP growth (see Fig. 2).

Fig. 2: Comparison of GDP growth and lighting growth.
figure 2

This figure shows the national average GDP growth rate (solid line) and national average night light growth rate (dotted line) from 1992 to 2013.

Full size image

Compared with traditional methods, using satellite light data as a proxy indicator of GDP data has the following advantages. Firstly, it enhances objectivity. Night light data, derived from the satellite-mounted DMSP/OLS sensor, is mainly acquired through technical means, thereby substantially reducing the interference of human factors. Secondly, it enhances comparability. Satellite light data is unaffected by price levels, ensuring consistency across different regions and time periods. Thirdly, night light data provides a more comprehensive reflection of regional economic conditions, capturing economic activities in informal and underground markets (Sutton and Costanza, 2002). Fourthly, it addresses the insufficiency of GDP accounting in remote areas. Satellite light data, with its wide geographical coverage and robust photoelectric amplification capabilities, can effectively compensate for the limitations of traditional GDP accounting methods, particularly in remote regions (Donaldson and Storeygard, 2016). Following Henderson et al. (2012), we estimate the following equation:

$${g}_{i,t}={\omega}_{0}+{\omega}_{1}{{light}}_{i,t}+{{{\upeta}}}_{i}+{{{\updelta}}}_{t}+{\varepsilon}_{i,t}$$
(3)

where gi,t represents the official GDP growth rate, lighti,t is the growth rate of light brightness per unit area in the jurisdiction, η is the non-observed province effect, δ is the year effect, the subscripts i and t represent different provinces and years, and ε is the stochastic disturbance term.

To further eliminate the possible measurement errors in the light data, we refer to Chen and Nordhaus (2011) and Henderson et al. (2012), combining the official GDP growth rate with the true GDP growth rate measured by night light data. The comprehensive estimating equation for measuring the real economic growth is:

$${\hat{g}}_{i}^{\prime} =\rho {g}_{i}+(1-\rho ){\hat{g}}_{i}$$
(4)

Where \({\hat{g}}_{i}^{\prime}\) represents the real GDP growth rate, \({g}_{i}\) represents the official GDP growth rate, and \({\hat{g}}_{i}\) is the fitting value of the GDP growth rate calculated from the light data of model (3). ρ represents the weight parameter. According to Henderson et al. (2012) and Xu et al. (2015), we calculate ρ = 0.586.Footnote 2GDPDIS is the difference between \({\hat{g}}_{i}^{\prime}\) and \({g}_{i}\). It is the main explanatory variable. Positive GDPDIS means that the official economic growth is higher than the actual economic growth. Note that light and GDP are different measurement methods. The gap between them may not be completely explained by GDP manipulation, but this is the most effective way under the currently achievable technology. It does not affect the horizontal comparison of GDP manipulation among provinces.

Empirical model

To verify the effect of local economic growth manipulation on livelihood investment, we estimate the following equation:

$${{SOCISRV}}_{i,t}={\alpha }_{0}+{\alpha }_{1}{{GDPDIS}}_{i,t-1}+{\beta }_{1}{{Controls}}_{i,t}+{\varphi }_{t}+{\gamma }_{i}+{\mu }_{i,t}$$
(5)

Where \({{SOCISRV}}_{i,t}\) represents livelihood investment, specifically the proportion of education expenditure, medical expenditure, and social security expenditure in public budget expenditure. \({{GDPDIS}}_{i,t-1}\) is the economic reality of area i at year t-1. Larger GDPDIS represents a higher degree of manipulation in official GDP data. \({\varphi }_{t}\) and \({\gamma }_{i}\) represent non-observation factors related to the year and region. \({\mu }_{i,t}\) represents the stochastic disturbance item. The subscripts i and t represent the region and year. \({{Controls}}_{i,t}\) represents a series of control variables, including the level of regional economic development, the degree of decentralization of fiscal expenditure, the degree of fiscal self-sufficiency, the degree of government competition, the regional industrial structure, the preference of residents’ public goods demand, and the heterogeneity of officials (age, gender, educational background, source of employment). All variables and their definitions are detailed in Table 1.

Table 1 Variable measurements.
Full size table

Empirical analyses

Descriptive statistics

Table 2 reports descriptive statistics for the variables. The average value of SOCISRV is 31.69%, the minimum is 19.50%, and the maximum is 43.23%. The standard deviation is 5.032, indicating that there are big differences in livelihood investment. The maximum indicates the highest degree of economic growth manipulation does not exceed 4.33%. The standard deviation was 0.7211, indicating that there are structural differences in economic authenticity among different provinces.

Table 2 Descriptive statistic.
Full size table

Table 3 shows the Pearson correlation coefficients between the main variables. The correlation coefficient between SOCISRV and GDPDIS is −0.125. The significantly positive at levels above 1%. This is preliminary evidence of a negative correlation between the two.

Table 3 Correlation coefficient matrix.
Full size table

Figure 3 reports the proportion of China’s national social service expenditure to GDP from 1998 to 2013. The three social service expenditures have been increasing steadily over the last decade, of which education expenditure accounted for the largest proportion, with a maximum of 3.7%, which is still below the budget target of 4%. Medical expenditure accounted for the smallest proportion, increasing gradually from 0.48 to 1.38%. Social security expenditure reached 2.33% in 2013, narrowing the gap with education expenditure gradually.

Fig. 3: The proportion of social service expenditures.
figure 3

This figure reports the proportion of national social service expenditures from 1998 to 2013.

Full size image

Table 4 reveals that the overestimation of GDP is not as severe as some existing literature suggests. The primary factor contributing to the gap between economic statistics and reality lies in the estimated true GDP growth rate. The discrepancy is mitigated by the saturation correction applied to the light data in this study, resulting in a closer alignment between actual GDP calculations and there corresponding luminosity.

Table 4 True economic growth rate: the average of 1993–2012 (%).
Full size table

China’s official average economic growth rate from 1993 to 2012 stood at 10.20%, with the true average growth rate slightly lower at 10.13%, indicating a manipulation rate of 0.54%. Consistent with findings by Xu et al. (2015), regions such as Inner Mongolia and Tianjin exhibit more pronounced data manipulation. In the eastern region, the average difference is 0.30%, whereas in the central and western regions, the average difference is negative. It implies that the actual economic growth rates in these regions surpass the officially reported figures. Such disparities may stem from underdeveloped statistical systems and accounting methodologies, leading to the omission of certain economic activities from official statistics, thereby understating the actual economic performance. Additionally, artificial factors may also play a role, with some regions underreporting economic income to secure greater central financial support. For instance, in October 2015, the National Audit Office disclosed cases in Guangxi Province where over 3000 individuals underreported their per-capital income and personal assets, failing to meet the standards for poverty alleviation filing.Footnote 3

Main results

Table 5 reports the effect of regional GDP manipulation on livelihood investment. We include different control variables and fixed effects in different models. Hypothesis 1 is supported by the coefficients on GDPDIS are negative. This implies that regional GDP manipulation has a significant negative effect on livelihood investment.

Table 5 Regional GDP manipulation and livelihood investment.
Full size table

Political cycle

Table 6 presents the effect of regional GDP manipulation on livelihood investment under the sub-samples of different governors’ political cycles. Compared with the secretary of the provincial party committee, the governor presides over the use of fiscal funds and economic policy formulation in the province, which has a greater effect on local fiscal expenditure. The governor is responsible for the implementation of social service security.

Table 6 Regional GDP manipulation and livelihood investment—tenure of governor.
Full size table

Hypothesis 2 is supported by the negative and significant coefficients of GDPDIS. Column (1) reports the results for the first year of the governor’s appointment. The coefficient on GDPDIS is −0.264, which is insignificant. Because local officials did not understand the economic situation in their region when they took office, the livelihood investment has not been reduced much. The descriptive analysis shows that the average official tenure is only 3.18 years. Column (2) reports the results based on the sub-samples in the first three years of tenure. The coefficient is −0.699, which is significant at the 1% level. Column (3) reports the coefficient was not only smaller (−0.312) but also less statistically significant in the sub-samples of 3–5 years’ tenure.

In general, the effect of regional GDP manipulation on livelihood investment is affected by the political cycle, with the peak appearing in the second year. The negative effect of regional GDP manipulation on livelihood investment in the early stage of office is more significant. When financial resources are limited and local officials have short-term tenure, it has become a rational choice for local officials to reduce social service financial resources to invest in economic development in the early stage.

Officials’ place of origin

To test Hypothesis 3, we divided the sample into two groups, the governor and party committee secretary. Columns (1) and (2) of Table 7 report the impact of regional GDP manipulation on livelihood investment in the governor group. Column (2) adds the interaction of GDPDIS and LOCAL_sz. The coefficient of the interaction term is −0.726, which is significant at the 5% level. It shows the effect of GDP manipulation on livelihood investment when governors’ workplaces are different from their birthplaces is stronger than when they work in their hometowns (LOCAL_sz = 0). Columns (3) and (4) report the impact of regional GDP manipulation on livelihood investment in the provincial party committee secretary group. The interaction coefficient is −1.676, which is significant at the 5% level, denoting that when provincial party committee secretaries hold office in their hometown, the effect of regional GDP manipulation on livelihood investment is significantly smaller. This evidence supports Hypothesis 3.

Table 7 Regional GDP manipulation and livelihood investment —regional favoritism.
Full size table

Robustness checks

Similar results have been found following robustness checks, which allow us to be confident about the empirical method used in the analysis.

Dynamic panel model

We add the lag term of SOCISRV to a dynamic panel model and choose the DID-GMM and SYS-GMM methods to estimate the model. We estimate the following equation:

$$\begin{array}{l}{{SOCISRV}}_{i,t}={\alpha }_{0}+{\theta }_{1}{{{SOCISRV}}_{i,t-1}+\alpha }_{1}{{GDPDIS}}_{i,t-1}\\\qquad\qquad\qquad\quad+\,{\beta }_{1}{{Controls}}_{i,t}+{\varphi }_{t}+{\gamma }_{i}+{\mu }_{i,t}\end{array}$$
(6)

Fiscal decentralization, local self-sufficiency, and government competition influence fiscal status. Fiscal status will also affect the variables above. We set them as endogenous variables. Poor livelihood investment may force people with high requirements for social services to move to areas with higher levels of public services. Considering the obstacles of the household registration system and the low possibility of inter-provincial migration, we attempt to set the residents’ livelihood needs as a predetermined variable. Disturbance items are unlikely to affect the economic data of the previous year, so the lagging GDPDIS is taken as an exogenous variable. The adjustment of industrial structure often takes a long period. Even if the interference item influences the industrial structure, it will not be reflected in the short term. Therefore, we set the proportion of the secondary and tertiary industries in GDP as exogenous variables. In addition, we set all dummy variables of officials’ characteristics (age, gender, education, career) as exogenous variables. Table 8 reports the dynamic panel regression results. We obtain regression results consistent with those in Table 5.

Table 8 Dynamic panels.
Full size table

Dependent variables

We use different calculation methods to measure the social service investment. SOCISRV_1 represents the logarithm of the total social service expenditure after deflation by the consumer price index, and SOCISRV_2 represents the natural logarithm of the per-capital social expenditure. The total social service expenditure is deflated using the consumer price index.

The regression results are shown in Table 9. The model controls the general control variables (regional economic development level, industrial structure, fiscal self-sufficiency, fiscal status, population dependency ratio, government competition ratio) and characteristic variables of the provincial officials (gender, education, age, career). To ensure the consistency of variables, the GDPDIS is the average per capita GDP. Using the logarithm method of total expenditure to test the results, we obtain regression results consistent with those in Table 5.

Table 9 Different methods to measure livelihood investment.
Full size table

Satellite light data: excluding some provinces

The continuous ground flame formed by gas flares in oil extraction processes may interfere with the night light data, thereby affecting the light brightness value and the estimation result. Because of the inability to obtain accurate data on the geographic scope of China’s oil production, following Xu et al. (2015), this paper excludes provinces with larger oil production scales based on the data from the China Statistical Yearbook over the years, namely Heilongjiang, Guangdong, Shaanxi, Xinjiang, Shandong, and Tianjin. We re-run the major regression using the sub-sample. Table 10 reports that the coefficients in columns (1) to (4) are all significantly negative at levels above 1%.

Table 10 Results of excluding some provinces.
Full size table

Further research: the effect of the financial crisis

We are interested in the differential effect of GDPDIS on SOCISRV during different economic environments. When economic growth slows, the pressure of local officials on economic development becomes greater. The reduction of livelihood investment will intensify. When the economic situation eases, it is easier to beat the growth target. Therefore, the reduction of livelihood investment caused by GDP manipulation will decrease.

Table 11 reports the regression results. The coefficient after 2007 is −0.185, which is much higher than the coefficient before 2007. 2007 was a turning point in China’s economic development. Under the influence of the financial crisis, the economic development trend declined after 2008. We believe that in the sub-sample of 2008–2013, there has been a structural change in the impact of regional GDP manipulation on livelihood investment.

Table 11 The effect of financial crisis.
Full size table

Conclusion

The negative impact of regional GDP manipulation on the national economic level is immeasurable. People’s livelihood not only reflects the living conditions of the entire society but also embody the people’s aspirations for a better life. It encompasses national strategies and goals. There persists an evident imbalance and inadequacy in livelihood investment, and regional GDP manipulation undoubtedly exacerbates this challenge.

From the perspective of local government financial management behaviors, this paper examines the impact of regional GDP manipulation on livelihood investment. We find that high levels of regional GDP manipulation significantly reduce livelihood investment. Moreover, the impact of regional GDP manipulation on livelihood investment is influenced by the tenure and birthplace of local officials. Particularly in the early stages of an official’s tenure, regional GDP manipulation has a more pronounced effect on livelihood investment. However, this impact diminishes when the official’s place of employment is their hometown. Furthermore, we observe that during periods of economic slowdown (2008–2013), the reduction in livelihood investment due to regional GDP manipulation is further exacerbated.

The media exposure of local GDP manipulation scandals in early 2018 highlighted the immediate need to address incentive manipulations. Minimizing information asymmetry through optimizing statistical systems and technological innovation emerges as a short-term solution. The incorporation of provincial GDP statistics under the leadership of the National Bureau of Statistics and the implementation of a statistical system with “quasi-vertical management” can effectively mitigate information asymmetry and incentive manipulations. In the long term, policy interventions should increase the weight of livelihood indicators in evaluating governance performance. Our findings carry significant policy implications, providing supporting evidence for policies aimed at optimizing mechanisms for evaluating officials. It is undeniable that the adverse effects of the imbalanced development of social services stem from the prioritization of economic growth. Therefore, enhancing cadre evaluation methods and prioritizing the improvement of social services in evaluation criteria are crucial steps towards modernizing the national governance system and enhancing its capabilities in the future.