Targeted poverty alleviation and corporate investment in poor counties: an empirical analysis from the geographical distribution of new subsidiaries of listed firms

ABSTRACT Location is key to corporate investment decisions. Many studies have examined firms' investment in regions with sound institutional environments, but little is known about their investment in poor regions. This paper examines the impact of Targeted Poverty Alleviation (TPA) on firms’ investment in poor regions using listed firms from 2007 to 2021. It finds that TPA mainly guides resource-dependent and labour-intensive firms to invest in poor regions through subsidiaries. Further analysis shows that firms establish subsidiaries in poor regions actively, not passively motivated by administrative orders. The heterogeneity analysis shows that, the higher the land and labour prices in firms’ location, the more likely they are to establish subsidiaries in poor regions. Signalling and resource effects are the main drivers of firms’ investment in poor regions. This paper provides evidence for the effectiveness of TPA and implications for firms to achieve common prosperity..


Introduction
Socialism aims to eliminate poverty, improve people's livelihoods, and gradually achieve common prosperity.The Chinese government has been committed to alleviating poverty and developing the countryside for a long time.However, the process has become more challenging over time, as the old poverty alleviation models have lost their effectiveness due to fragmented distribution and the depth of poverty among the poor.To improve the efficiency of poverty alleviation, China promoted Targeted Poverty Alleviation (TPA) in 2013, and allocated social resources specifically to poor regions.
TPA is different from the old way: blood-transfusion-based poverty alleviation.It emphasises the importance of stimulating the endogenous development of the poor, the crucial role of market mechanisms, and the key to industrialisation in alleviating poverty.Firms are the direct carriers of industries.Therefore, how to guide firms to invest and build factories in poor regions has become an important part of promoting TPA, which is jointly promoted by government, market, and society.Following the proposal for TPA, the government has issued a number of policies to encourage firms to invest in poor regions and participate in poverty alleviation.For example, 'Opinions on Innovative Mechanisms to Solidly Promote Rural Poverty Alleviation and Development', 'Opinions on Mobilizing Social Forces to Participate in Poverty Alleviation and Development', 'Opinions of the China Securities Regulatory Commission (CSRC) on Playing the Role of Capital Markets to Serve the National Strategy for Poverty Alleviation and Development', etc.After TPA implementation, have Chinese firms increased their investment in poor regions?
Figure 1 illustrates the practice of Chinese A-share listed firms.Between 2007 and 2021, the number of new subsidiaries established by non-financial listed firms within nationallevel poverty-stricken counties (NPCs) remained relatively stable, but increased significantly after the proposal of TPA in 2013.In 2019, for example, 774 A-share listed firms set up 1,854 new subsidiaries in NPCs.While capital and talent have migrated to developed regions, some firms have moved back to poor regions to invest.Poor regions have gradually become important locations for listed firms to invest in.The existing literature finds that firms prefer to invest in developed regions with a better industrial base and superior institutional environment, from the dimensions of returns to scale, industrial agglomeration, and regional institutional environment (Baldwin & Krugman, 2004;Delios & Henisz, 2003;Mclean et al., 2012).However, little literature has explored why firms establish subsidiaries in poor regions.
This paper examines the specific impact of TPA on corporate investment location and its mechanism within China's unique institutional background.We manually collect and collate data on the list of NPCs along with the establishment of new subsidiaries by firms.Our empirical results demonstrate that after TPA implementation, companies established more subsidiaries in poor counties, and these companies are mostly resource-dependent and labour-intensive.Further analysis shows that the establishment of subsidiaries in poor counties is an active pursuit of policy dividends, rather than a passive investment behaviour driven by administrative orders, and that firms are 'migrating birds'.In the mechanism analysis, we show that TPA leads to the information effect and the resource effect on subsidiaries' investments.
We extend the literature in several ways.First, our study broadens the research scope of corporate subsidiary location selection.Existing literature analyzes the location selection of subsidiaries based on factors such as tax avoidance, macro environment, infrastructure, industry associations, pollution transfer, and others (Benito et al., 2003;Cao & Jia, 2020;Ma et al., 2020;Single, 1999;Song et al., 2021).Based on China's unique TPA policy, this paper examines how TPA affects firms' subsidiary location selection, and enriches and develops research on this topic.
Second, our study enriches and expands the theoretical framework of corporate investment location.The literature has already studied that firms prefer to invest in developed regions with superior industrial bases and institutional environments based on returns to scale, industrial agglomeration, and regional institutional environments (Baldwin & Krugman, 2004;Delios & Henisz, 2003;Mclean et al., 2012).However, some firms have returned to poor regions to invest after TPA implementation.This paper proposes that TPA is compatible with industrial transfer in China.By combining their efforts, these two factors can lead industries that have lost their comparative advantages in developed regions to relocate to poor regions.This enriches and expands the theoretical connotation of corporate investment location.
Third, we contribute to the emerging literature on TPA's impact on firms' subsidiary investment.Studies have examined how TPA impacts firms in terms of financing constraints, innovation, risk, operation performance, and philanthropy (Deng et al., 2020;Ji et al., 2021;Liu et al., 2020;Zhen & Wang, 2021).We uncover a novel effect of TPA on firms at the level of subsidiary investment.We help the theoretical and practical communities to better understand the economic consequences of TPA, adding to relevant theoretical and empirical research.
Finally, this study has practical significance.Our findings provide valuable references to aid firms in achieving common prosperity during rural revitalisation.We also provide company-level evidence of TPA implementation effectiveness.
The remainder of this paper is organised as follows.Section 2 describes the institutional background and hypothesis development.Section 3 describes the sample, variable definitions, and empirical model specifications.Section 4 reviews the summary statistics and reports the main empirical results.Section 5 presents the further analysis.Finally, Section 6 offers a brief conclusion.

Institutional background
Poverty is a major social problem in the process of globalisation.Socialism aims to eliminate poverty, improve people's livelihood, and gradually achieve common prosperity.The Chinese government has a long history of alleviating poverty and developing the country.Even though China did not have a specific anti-poverty policy, land reform, and the planned economic system contributed to the alleviation of rural poverty on a large scale at the institutional and material levels.With the reform and opening-up, along with the establishment of the market economy, China's economic development and living conditions have generally improved.However, the reform dividend does not benefit everyone equally.The policy of pioneering development in eastern coastal areas leads to a rapid economic boom in these regions.The western and central regions lag in economic development and experience greater income inequality.Consequently, to reduce absolute poverty in the central and western regions, the government begins implementing poverty alleviation policies with a regional orientation.For example, the State Council's Leading Group for Agricultural Construction in the Three Western Regions, established in 1982, was China's first regional poverty alleviation agency.Accordingly, in 1984, the 'Circular on Helping Poverty-stricken Areas to Change Their Appearance as Soon as Possible' proposed poverty alleviation policies for remote mountainous areas, minority populated areas, and old revolutionary areas.Chinese poverty alleviation policies in the early years adopted policies such as government transfers and tax incentives, which benefited the poor in the redistribution process.Then, poverty alleviation evolved from widespread universal to targeted alleviation.As a result of the massive investment of financial, material, and human resources, China's poor population has dropped from 689 million in 1990 to 250 million in 2011.
However, the poverty alleviation process has become more challenging over time.The old poverty alleviation models have declined marginally in effectiveness due to the fragmented distribution of the poor and the depth of poverty (Chen et al., 2019;Jia et al., 2017).To improve the efficiency of poverty alleviation, China began to promote Targeted Poverty Alleviation (TPA).A preliminary concept of TPA was first proposed by General Secretary Xi Jinping during his inspection in Xiangxi, Hunan Province, in November 2013.The government issued 'Opinions on Innovative Mechanisms for Solidly Promoting Rural Poverty Alleviation and Development' in 2014, which elevated TPA to an important mechanism.The implementation of TPA was then promoted through a series of policies.
Unlike the previous poverty alleviation practices, TPA emphasises the market mechanism and promotes a large poverty alleviation framework involving the government, the market, and society.Firms are key players in the market and are responsible for direct industrial development.To encourage firms to invest in poor regions, the government has introduced a series of policies.For example, it has issued 'Opinions on Further Mobilizing the Power of All Social Sectors for Poverty Alleviation and Development', and provided financial discount and credit support to firms investing in poor regions.The 'Guidance on Three-Year Action to Win the Battle against Poverty' encourages the development of ecofriendly labour-intensive industries in poor regions and encourages firms to invest in these regions through subsidies and loans.Additionally, several local governments have provided policy support to firms engaged in industrial poverty alleviation through land, taxation, and subsidies.

Hypothesis development
Firms primarily invest in order to expand their production, and their investment is highly dependent on two key elements: information and capital.As a strategic measure of national poverty alleviation governance, TPA influences the location of corporate subsidiaries in both information and capital aspects.In one sense, TPA is a location-based policy that directly targets poor regions.It conveys economic signals of government's inclination towards resources in different areas, influencing corporate managers' judgement of regional development, and enhancing their confidence to invest in poor regions.This is known as the 'signalling effect'.TPA lowers firms' investment and operation costs in poor regions by providing them with government subsidies, tax preferences, and other policy dividends.It also improves firms' ability to obtain resources, thereby encouraging them to invest.Thus, the 'resource effect' is evident.

The "signaling effect" of TPA
Corporate investment is a complex process with a long duration, high risk, and high uncertainty.Investing is difficult for firms due to information asymmetry, as it makes it harder for management to discern future development trends.This increases investment risks (Rao et al., 2017).To avoid huge losses to production and operation, firms will try to obtain effective information for investment decisions.Meanwhile, related studies have suggested that firms will hedge the investment risk resulting from uncertainty by investing in areas explicitly supported by policies (Criscuolo et al., 2019;Hong et al., 2021;Huang & Lv, 2016;Tan et al., 2017).
As a type of location-oriented policy, TPA aims to redirect resources to poor regions.In essence, it is an economic signal released by the government, indicating that resources are favoured to be distributed to poor regions.TPA provides firms with reliable external information about regional investment and reduces their information uncertainty.This signal affects firms' management directly, as it influences their judgement of the market situation, enhances their security in investing in poor regions, and increases their willingness to invest in poor regions.

The "resource effect" of TPA
The history and geographical location of poor regions make them significantly less attractive to corporate investments.To encourage corporate investments in poor regions, both the central and local governments have implemented various policies to reduce the unfavourable factors and increase the benefits of investing in these regions.These policies include government subsidies, tax incentives, and other preferential measures that aim to motivate firms to invest in poor regions and alleviate their lack of willingness to do so.
TPA policies can improve firms' access to resources.Government subsidies are provided to firms that invest in poor regions in order to achieve the goal of TPA.Guizhou Province, for example, provides financial subsidies to subjects involved in poverty alleviation activities. 1 The availability of resources is an important factor in motivating firms to invest in regions supported by location-oriented policies (Liu et al., 2019;Wei & Yuan, 2010;Yuan & Zhu, 2018).According to existing literature, firms' participation in TPA can assist them obtaining government subsidies (Han & Wu, 2021;Zhen & Wang, 2021).Firms' access to scarce resources such as government subsidies for investing in poor regions gives them a competitive advantage in profitability and solves the problem of their reluctance to invest in poor regions.
TPA policies also can reduce firms' investment costs in poor regions.Investment, like any other production and business activities, aims for profit maximisation.When making investment decisions, firms need to consider both costs and benefits.Investing in poor regions involves not only direct costs such as preparation and transportation, but also transaction costs due to market segmentation and new environment.Various government preferential policies reduce firms' investment cost and operation costs in poor counties.These policies include tax incentives, social insurance subsidies, vocational training subsidies, loan subsidies, risk compensation, etc.For instance, Anhui Province and Hubei Province have provided preferential policies on credit, land use, and taxation to firms investing in poor regions,2 which are sufficient to compensate for the additional costs of investment and operations in these regions.In addition, policies may facilitate the approval of projects that invest in poor regions.Research has shown that the relaxation of the approval system can reduce transaction costs and improve the efficiency of firms (Guo & Shao, 2021;Wang et al., 2020).Furthermore, the prices of labour, land, and other factors are relatively low in poor regions, and the operating costs of firms are low.Therefore, the cost advantage firms receive from TPA policies, combined with the lower factor prices in poor regions, guides them to invest in poor regions.
Based on the analysis of both the 'signalling effect' and the 'resource effect', we propose the following hypothesis: H1: After the implementation of Targeted Poverty Alleviation (TPA), firms will establish more subsidiaries in poor counties.

Sample selection and data source
This study selects non-financial firms listed on the Shanghai and Shenzhen stock exchanges from 2007 to 2021 as the sample.We exclude special treatment companies, coded as ST (the company has suffered losses for 2 consecutive years), *ST (the company has suffered losses for 3 consecutive years), and PT (the company waiting for delisting).In this study, we use geographic data of newly established subsidiaries of listed firms, data of poor counties, and financial data of listed firms.
We obtain the geographic data of newly established subsidiaries of listed firms from the Chinese Research Data Services (CNRDS) database, and standardise their geographic location as 'a certain county (district) in a certain city in a certain province'.When the subsidiaries' geographic information are not available in the CNRDS database, we manually retrieve it by searching 'Tianyancha' or 'Baidu'.In addition, following Cao et al. (2018), we identify new subsidiaries as those that are reported by listed firms in the current year but not in the previous year, to reduce the influence of historical investment on the analysis.
We identify poor counties based on the list published by the State Council Leading Group Office of Poverty Alleviation and Development. 3Financial data are derived from the China Stock Market and Accounting Research (CSMAR) database or the CNRDS database.To reduce the influence of outliers, all of the continuous variables are winsorised at the 1 st and 99 th percentiles.We use Stata 16 for the regression analysis.

Measures and empirical model specification
To identify how TPA impacts listed firms' subsidiaries in poor counties, we use a difference-in-difference (DID) model.Since TPA affects all firms, we cannot simply compare the number of subsidiaries before and after TPA implementation in poor counties, as this may reflect a time trend rather than the effect of the policy.For this reason, we need to find a suitable control group that is not affected by TPA or affected to a lesser degree.We use the difference in the degree of TPA impact on listed firms as a source of variation to determine the effect of this policy.
Different firms have different requirements for regional resource endowments, which are reflected in their investment preferences.Technology-intensive firms require highly qualified staff and an environment with strong R&D capabilities to drive innovation and development.They therefore prefer to locate in developed regions.In contrast, resources such as land, minerals, and labour are abundant in poor regions, which are more attractive to labour-intensive and resource-dependent firms.Accordingly, TPA policy guides corporate investment more for resource-dependent and labour-intensive firms.As shown in Figure 2, most listed firms with subsidiaries in poor counties have stronger resource and labour dependence, except for information transmission, software, and information technology service firms.The effect of TPA policy depends on whether the factor demand of the firm matches the resource endowment of the poor counties.We use this difference to identify the effect of the 'one-size-fits-all' policy.
On this basis, this paper constructs a double difference model to test whether TPA policy implementation and factor demand of firms compatible with the resources of poor regions.We define a treatment group as resource-dependent and labour-intensive firms whose factor demands match the resource endowments of the poor regions, and a control group as firms whose poverty alleviation policy has a different guiding effect.We observe the change in the number of subsidiaries of listed firms in poor regions before and after the implementation of the TPA policy (first difference), comparing this change among firms with different levels of influence (second difference), to identify how the TPA policy affected the establishment of subsidiaries in poor counties.The specific model is as follows: Where i denotes the firm, t denotes the year, k denotes the control variable, γ i is the firm fixed effect, δ t is the year fixed effect, and ϑ p is the region fixed effect.The regression coefficient β1 indicates, before and after the TPA policy, the change in the number of new subsidiaries established by resource-dependent and labour-intensive firms in poor regions, in comparison with other firms.If our hypothesis is true, β1 should be significantly positive.
The dependent variable is the number of subsidiaries established by the firm in poor counties in year t (PoorSub) and its natural logarithm plus one (LnPoorSub).
The treatment group and the control group.Treatment groups are resourcedependent and labour-intensive firms with factor demands matching poor regions' resource endowment.Treat takes the value of 1 and 0 for other firms as the control group.The classification criteria for resource-dependent and labour-intensive firms follow Lu and Dang (2014), Wang and Guo (2015). 4 Policy shock dummy variable (Post).Even though the concept of TPA was introduced in November 2013, the TPA policy was implemented in 2014 and beyond.Post equals 1 for observations in 2014 and later years, 0 for otherwise.
Following Cao et al. (2018), Cao and Jia (2020), we also control for a series of variables that affect new subsidiaries established by the firm.These variables include corporate characteristics, corporate governance, industry competition, and the macroeconomic environment.These are firm size (Size), leverage level (Lev), performance (Roa), growth (Growth), age (Age), SOE (Soe), shareholding ratio of the largest shareholder (Top1), shareholding ratio of institutional investors (Inshold), whether the chairman and CEO are the same person (Dual), the proportion of independent directors (Indr), Herfindahl index (HHI), GDP per capita (GDP).In addition, we control for firm, year, and province fixed effects.Table 1 presents definitions of all of these variables.The natural logarithm of total assets at the end of the year Lev

Descriptive statistics
The ratio of total liabilities to total assets at the end of the year Roa The ratio of annual net profit to total assets at the end of the year Growth The annual percentage revenue growth of a firm Age The natural logarithm of 1 plus the number of years since a company's establishment Soe A dummy variable that equals 1 if the firm is state-owned, and 0 otherwise Top1 The shareholding ratio of the largest shareholder at the end of the year Inshold The proportion of the firm's shares held by institutional investors Dual A dummy variable that equals 1 if the chairman and the CEO are the same person and 0 otherwise Indr The percentage of independent directors on the board HHI Calculated by squaring the operating income share of each firm in the industry and adding all firms up GDP The GDP per capita of county j in year t

Main results
Table 3 presents the empirical results for our main analysis as specified in Model (1).Columns (1) and ( 2) display the results without control variables, while columns (3) and (4) display the results including all variables.The coefficients on Treat×Post are significantly positive at the 1% level in all four columns, suggesting that TPA implementation guides firms to establish more subsidiaries in poor counties.This guiding effect is stronger for resource-dependent and labour-intensive firms that match poor regions' factor demand and resource endowment.These results supports our hypothesis.The coefficient is also significant economically.After the implementation of TPA, the number of new subsidiaries established by listed firms in poor counties increased by 28.95% ( = 0.077/0.266)compared with the unaffected control group.

Parallel trend assumption test
The basic premise for DID to effectively identify causal relationships is that the parallel trend hypothesis holds.Referring to Li et al. (2021), we use the event study method to test whether the parallel trend is met.The test results are shown in Figure 3, where the ordinate represents the size of the coefficient and the dashed line represents the confidence interval.Before the implementation of TPA, the difference in the number of new subsidiaries established in poor counties between the treatment group and the control group is relatively small, while after the implementation of TPA, the difference between them gradually widened, indicating that our DID method meets the requirements of parallel trends.From the dynamic effect, it can be seen that the number of new subsidiaries established in poor counties has fluctuated since 2014 and reached a relatively high point in 2019.

Choosing to establish subsidiaries in poor counties: active or passive
We believe that after the implementation of TPA, the establishment of subsidiaries by firms in poor regions is an active investment behaviour that pursues policy dividends.However, this result may be a passive behaviour of firms under the counterpart policy of TPA.Since the central government promotes TPA vigorously, local governments are assessed on their TPA performance.Local departments and officials who fail to implement TPA will be held accountable.It has given local governments in developed areas a huge push to promote firms in their jurisdictions to alleviate poverty in counterpart regions, easing their own TPA pressure.When the government wants corporate social responsibility, it may ask firms for donations in the form of 'public welfare apportionment' or 'task assignment'.It is possible that listed firms' establishment of subsidiaries in poor counties is a passive investment behaviour mobilised by local governments, rather than an active effort to seek policy dividends through resource allocation.
To further explore whether the motivation for firm investment in poor regions is an active layout or a 'passive' investment, we utilises the counterpart policy.China has implemented a counterpart policy since the 1980s in order to reduce the economic gap between eastern and western regions and alleviate poverty in poor regions. 5In the counterpart policy, developed provinces provide 'point-to-point' assistance to poor provinces.The characteristics of counterpart policy are consistent with the essence of TPA.After the implementation of TPA, the counterpart policy has become its important policy component.Assuming that the establishment of subsidiaries in poor regions is a passive investment, it is likely that new subsidiaries will originate mainly from counterpart provinces rather than from other provinces.Conversely, it indicates that the establishment of subsidiaries by firms in poor regions is an active investment behaviour.
Accordingly, we further divide the subsidiaries in poor counties into counterpart subsidiaries and non-counterpart subsidiaries according to whether they are sourced from listed firms in corresponding counterpart-assisted regions.The counterpart policy list consists of two parts: one is the East-West Cooperation Poverty Alleviation List; the other is, the State Council Poverty Alleviation Office's List of Paired Assistance for 'Joining Hands to Achieve Prosperity' formulated in 2017.According to the list, 267 economically developed areas in eastern China will cooperate with 390 poor counties in western China to achieve prosperity.The changes in the number of subsidiaries in poor counties from these two different sources are shown in Figure 4.As shown by this figure, after TPA, listed firms have established large numbers of subsidiaries in poor counties from their counterpart regions, but they grow at a slower rate and have a lower absolute number than those from noncounterpart regions.This preliminary finding indicates that after TPA, listed firms that establish subsidiaries in poor regions mainly come from regions outside counterpartassistance relationships.In addition, this article uses the number of subsidiaries established by listed firms from non-counterpart regions (NopairPoorSub) and its logarithm plus one (LgNopairPoorSub) as dependent variables and conducts regression according to Model (1).Table 4 shows the results.The coefficients of Treat×Post are still significantly positive at the 1% level.This indicates that after TPA implementation, firms' establishment of more subsidiaries in poor counties cannot be fully explained by 'passive' investment behaviour.Instead, they are more likely to be 'active' investment behaviour pursued by firms.

Placebo test
After implementing TPA, new subsidiaries established in poor counties may also be attributed to other factors at the regional, annual, and company levels that have not been considered.To eliminate these interferences, we follow Cai et al. (2016) approach to conduct the placebo test.Specifically, a processing group was randomly designated from the sample, assuming that these firms are resource-dependent or labour-intensive as treatment groups, and the remaining firms are control groups.Model (1) was subjected to 500 random sampling regressions.The regression results in Figure 5 show that, the proportion of significant positive and negative coefficients of the interaction term is small, which means that the virtual processing effect constructed in this paper does not exist.These findings suggest that resource-dependent and labour-intensive firms have increased their new subsidiaries in poor counties, rather than due to other unobservable factors.

Propensity-score-matched (PSM) sample
A control group was constructed based on PSM.The previous section identified resourcedependent and labour-intensive firms that are able to match the resource endowments of poor counties as the treatment group, while the control group consists of other firms.To avoid possible 'selection bias' and enhance the robustness of the results, we use propensity score matching (PSM) to construct a reconstructed control group for regression analysis.Specifically, we use Logit regression, and the control variables in the baseline regression as matching covariates.The control group is constructed using a 1:1 nearest neighbour matching method without replacement.Table 5 reports inter-group differences after matching.Column (1) represents the matched treatment group, while column (2) represents the control group sample constructed through the PSM method.Column (3) shows the results of mean difference test, while column (4) shows T-value of mean difference test.As a result of matching, there are few differences between the two groups.This indicates that PSM method has effectively reduced differences between treatment and control groups.Table 6 shows the results of regression on samples constructed using the PSM method.The coefficient of Treat×Post are positively significant, supporting the findings of previous analyses.

Alternative samples and alternative measure of dependent variables
We repeat our main regression model using alternative samples and alternative measures of dependent variables.This is to ensure that our main findings are not affected by a specific sampling procedure or measure of variables.Table 7 shows the results.
Alternative samples.We exclude the sample from the year when TPA was proposed.Since TPA was first proposed at the end of 2013, we use 2014, when it was implemented, as the impact year.However, this processing method ignored that some firms may have already been affected by policies at the end of 2013.Therefore, we exclude observations from 2013 for regression analysis.The results are shown in columns ( 1) and ( 2) of Table 7, and the coefficient of Treat×Post is still significant at the 1% level.Alternative measure of dependent variables.The previous regression used the absolute number of newly established subsidiaries in poor counties as a measurement indicator.According to Cao et al. (2018), we use the relative amount as a robustness test.That is, the proportion of subsidiaries in poor counties (PoorSubrat), the number of newly established subsidiaries by company i in poor counties divided by the number of newly established subsidiaries of firm i in year t.In column (3) of Table 7, the coefficient of Treat×Post is still significant and positive.The main results are robust to alternative proxies for dependent variables.

Tobit model
To avoid biased estimation results obtained by OLS, we use Tobit model as a robustness check.Since some listed firms did not establish subsidiaries in poor counties in some years, the values of the dependent variables PoorSub and LnPoorSub are 0 at this time.The results are shown in Table 8, and the coefficient of Treat×Post is significantly positive at the 1% level, still supporting our research hypothesis.

Heterogeneity analysis
Although TPA guides all enterprises to invest in poverty-stricken areas, the factor requirements of resource-dependent and labour-intensive firms are more in line with poor regions' resource endowment.Therefore, TPA has a stronger guiding role for these two types of firms.On the basis of this notion, we construct experimental and control groups to explore the heterogeneity of TPA policies on firm investment guidance.We conduct heterogeneity analysis from the perspectives of land and labour.
If resource-dependent and labour-intensive firms have higher enthusiasm for investing in poor regions, it is mainly because resource endowment in poor regions can better meet the factor demand of these two types of firms.The higher the land and labour factor prices in the firm's location, the more obvious the comparative advantage of poor counties, and the greater the incentive for firms to invest there.Following this logic, we acquired the land and labour factor prices of the listed firm's location.We classified them into the high factor price group and low factor price group according to whether they exceed the median price of that year, and performed separate regressions.If listed firms invest in poor counties based on regional comparative advantage, the higher the land and labour factor prices in the listed firm's location, the stronger their motivation to establish subsidiaries in poor counties.
For the land factor price, we obtained the land transaction area and transaction price of all the listed firms' locations from the 'China Land Market Network'. 6It publishes the transfer information of each industrial land according to the 'Regulations on the Transfer of State-owned Land Use Rights by Bidding, Auction and Listing' issued by the Ministry of Land and Resources in 2006.A substitute variable LandPrice for land factor price is calculated by using the average local land transaction price within a year.For the labour factor price, we obtained the 'average annual wage of employees on duty' in the listed firm's location from the EPS Global Statistical Analysis Platform and the 'China County Statistical Yearbook'.This measure is used as a substitute variable LaborPrice for labour factor price.
Land and labour prices in poor regions are relatively low.Therefore, the higher the labour and land prices in the listed firm's location, the more obvious the factor comparative advantage in poor regions, and the stronger the attraction of poor regions to companies.The samples are divided into high price group and low price group according to whether they are greater than or equal to the median of LandPrice and LaborPrice of that year.The results of grouped regression are shown in Tables 9 and 10.
Table 9 shows the results of grouped regression based on whether the land factor price (LandPrice) in the listed firm's location is higher than or equal to the median.The dependent variable in columns (1) and ( 2) is the number of subsidiaries established by the firm in poor counties in that year (PoorSub), and the dependent variable in columns (3) and ( 4) is the logarithm value of PoorSub plus 1.In columns (1) and (3) with higher LandPrice and columns (2) and (4) with lower LandPrice, the coefficient of Treat×Post is significantly positive.However, the coefficients of Treat×Post in columns (1) and (3) are higher than that in columns ( 2) and (4), and the difference between groups is significant.This indicates that the higher the land factor price in the listed firm's location, the more obvious the land comparative advantage in poor counties.In addition, the more motivated listed firms are more prone to establish subsidiaries in poor counties.
Table 10 shows the results of grouped regression based on whether the labour factor price (LaborPrice) in the listed firm's location is higher than or equal to the median.In columns (1) and (3) with higher LaborPrice and columns (2) and ( 4) with lower LaborPrice, the coefficient of Treat×Post is significantly positive.But, the coefficients of Treat×Post in columns (1) and (3) are higher than that in columns (2) and ( 4), and the difference between groups is significant.This implies that the higher the labour factor price in the company's location, the greater the labour comparative advantage in poor counties, and the more motivated companies are to establish subsidiaries there.

Mechanism analysis: "signaling effect" and "resource effect"
Investment is the primary activity of firms to expand and reproduce, and firms' investment activities highly depend on two key factors: information and capital.Based on this, we argue that the TPA policy, on the one hand, improves firms' willingness and confidence to invest in poor regions.This is done by transmitting economic signals of resource inclination to these regions.On the other hand, TPA policies such as land transfer and tax preferential reduce the cost of firms' investment in poor regions.This improves the ability to obtain key resources such as government subsidies and encourages firms to invest in poor regions.We examine the mechanisms of TPA policies that lead firms to invest in poor regions from the perspectives of 'signalling effect' and 'resource effect'.'Signaling effect' of TPA.To avoid huge losses caused by failed investment decisions, firms will hedge risks by investing in areas clearly supported by policy.TPA conveys economic signals of resource inclination to poor regions to the market, provides firms with definite external information about regional investment, reduces information uncertainty, strengthens management's sense of security and confidence about investing in poor regions, and therefore increases their willingness to do so.To reflect the impact of TPA on firms' subsidiary investment expectations in poor regions, we analyse the tone in 'Management Discussion and Analysis' (MD&A).Listed firms' annual report is a crucial basis for investors, financial institutions, and other stakeholders to make investment decisions.Among these, MD&A provides an overview of the firm's actual operating situations and a future outlook by the management, which contains a certain amount of information.A more positive tone of management implies that management is full of confidence in firms' development (Xie & Lin, 2015;Zeng et al., 2018).We analyse the impact of TPA policy on management's investment expectations through the tone of MD&A.Referring to Xie and Lin (2015) and Zeng et al. (2018), the tone of management (Tone) is calculated by using the number of positive words (Positive) and negative words (Negative) in MD&A as Model (2).
The value of Tone ranges from [−1,1], and the higher the value, the more confident the management is in the firm's investment plans.
In order to analyse the mechanism of TPA policy affecting firm investment more precisely, we further rearrange the sample into resource-dependent and labourintensive firms.Then, we divide them into the treatment group and control group according to whether they set up subsidiaries in poor counties.This can eliminate the differences caused by industry, and estimate the net impact of TPA on whether firms decide to invest in subsidiaries in poor counties.Specifically, resource-dependent and labour-intensive firms that establish subsidiaries in poor counties are the treatment group, where Treat1 takes 1.Resource-dependent and labour-intensive firms that did not establish subsidiaries in poor counties are the control group, where Treat1 takes 0. After the firm establishing subsidiaries in poor counties, Post1 takes 1, and before establishing subsidiaries in poor counties, it takes 0. Model (3) is designed for regression as follows.
Where i denotes the firm, t denotes the year, k denotes the control variable, γ i is the firm fixed effect, δ t is the year fixed effect, and ϑ p is the region fixed effect.
The results are shown in column (1) of Table 11, and the coefficient of the interaction term Treat1 × Post1 is significantly positive.This means that firms that set up subsidiaries in poor counties are more optimistic about investment projects, thus proving the 'signaling effect'.
'Resource effect' of TPA.Both central and local governments provide preferential policies to encourage firms to invest in poor regions.These policies include tax incentives, social insurance subsidies, vocational training subsidies, loan interest subsidies, risk compensation, etc.These directly lower the investment cost and operating cost of firms in poor counties.Meanwhile, factor prices in poor regions are low, which also lowers the cost of firms to invest in poor regions.In addition, the government provides subsidies to firms that invest in poor regions.This increases the willingness of firms to invest in poor regions.To examine the effects of establishing subsidiaries in poor regions on firm costs and government subsidies, operating cost rate (Cost) and government subsidy (Subsidy) are used as dependent variables, with Model (3) used for regression.
The results are shown in columns (2) and (3) of Table 11.When operating cost rate (Cost) is used as the dependent variable, the coefficient of Treat1×Post1 is significantly negative.This implies that firms that establish subsidiaries in poor counties can obtain the resources of these counties at lower costs, and receive various government subsidies to reduce their operating costs.Subsidy refers to Luo et al. (2014), and is measured by the natural logarithm of the annual government subsidy obtained by listed firms.When government subsidy (Subsidy) is used as the dependent variable, the coefficient of Treat1*Post1 is significantly positive.This means that firms that set up subsidiaries in poor regions can receive more support from the government.

Conclusion
Based on the manually collected geographical data of newly established subsidiaries of listed firms from 2007 to 2021, this paper finds that Targeted Poverty Alleviation (TPA) guides firms to invest in poor regions.The internal mechanism is that TPA i) improves firms' willingness and confidence to invest in poor regions, ii) reduces the cost of enterprises' investment in poor regions through government subsidies, tax incentives, and other policies, and iii) improves the investment attractiveness of poor regions.Distinguishing the motivation for firms' investment in poor regions, this paper finds that the establishment of subsidiaries in these regions is an active investment behaviour, rather than a passive obedience under administrative orders.
Further analysis shows that when the land and labour factor prices in the firms' location are higher, the firms have more motivation to establish subsidiaries in poor regions.
The implications of this study are mainly in three aspects: First, the government should clarify its role as a 'platform' for industrial poverty alleviation.To 'build a good platform', the government needs to make further investments in improving the infrastructure of poor regions, and releases local comparative advantages.For firms, poverty alleviation is not their primary business purpose.Making firms passive participants in poverty alleviation through administrative orders may distort firms' behaviour norms.Therefore, the government should respect firms' independent choice and encourage them to participate in poverty alleviation according to their own advantages and market demand.
Secondly, local governments should enact preferential policies on land, employment, taxation, subsidies, etc., in accordance with regional comparative advantages.Local governments also should enhance the attractiveness of poor regions from the policy support perspective, and develop advantageous industries in a tailor-made way.This can promote the integration of interests between firm development and poor region development.Underdeveloped regions need to fully utilise their comparative advantages (Lin et al., 2006).
Finally, the government should eliminate market segmentation barriers to crossregional resource flows to better utilise regional comparative advantages.Market segmentation severely hinders local resources flowing to other places or external resources entering the local market.Therefore, the government needs to reduce local protectionism and market segmentation, and clear the channels for resource flow.The industrial activity can then be transferred from developed regions to poor regions.

Figure 1 .
Figure 1.New subsidiaries of listed firms to national-level poor counties, 2007 to 2021.

Figure 2 .
Figure 2. Industry distribution of listed firms with new subsidiaries in poor counties, 2007 to 2021.

Figure 4 .
Figure 4. New subsidiaries of listed firms to poor counties, 2007 to 2021: by counterpart source.

Table 2
shows the summary statistics of the main variables.The mean number of new subsidiaries established by listed firms in poor counties (PoorSub) is 0.266, with a maximum of 7 and a minimum of 0, and a standard deviation of 0.816.This indicates that there is a variation in the number of subsidiaries established by listed firms in poor counties.The mean value of Treat is 0.376, which means that nearly 38% of firms in the sample belong to resource-dependent or labourintensive firms.Other control variables are generally consistent with previous research.

Table 1 .
Variable definitions.PoorSub The number of new subsidiaries established by firm i in poor counties in year t LnPoorSub The natural logarithm of one plus the number of new subsidiaries established by firm i in poor counties in year t Treat A dummy variable that equals 1 if the firm is a resource-dependent or labour-intensive firm and 0 otherwise Post A dummy variable that equals 1 for observations in 2014 and later years, 0 for otherwise Size

Table 3 .
Effect of TPA on firms' subsidiary investments in poor counties.

Table 4 .
TPA and firms' subsidiary investments in poor counties: excluding counterpart policy.

Table 5 .
Variable differences after PSM method.

Table 7 .
Alternative samples and alternative measure of dependent variables.

Table 9 .
TPA and firms' subsidiary investments in poor counties: advantage in land price.

Table 10 .
TPA and firms' subsidiary investments in poor counties: advantage in labour price.

Table 11 .
The influence of firms' subsidiary investments in poor counties.are based on robust standard errors adjusted for firm-level clustering.*, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.