Skip to content
BY 4.0 license Open Access Published by De Gruyter Open Access August 26, 2022

Why do zombie firms seldom die or resurrect? The effect of government subsidies on the survival duration of China’s zombie firms

  • Ming Deng and Jinbo Wang EMAIL logo
From the journal Economics

Abstract

One of the most important tasks in China’s supply-side structural reform is to effectively clean up zombie firms. Using industrial firm-level data in China from 1998 to 2007, we identified zombie firms in Chinese industrial sectors. Based on the identified zombie firms, we measured the survival duration of zombie firms and empirically studied the effect of fiscal subsidies on the survival duration of zombie firms by using survival analysis. Results revealed that fiscal subsidies significantly lengthened the survival duration of zombie firms. Specifically, fiscal subsidies not only reduced the possibility of zombie firms exiting the market, but more importantly, they reduced the possibility of zombie firms’ resurrection. Thus, fiscal subsidies do not contribute to the “death” or “resurrection” of zombie firms. These results are robust despite changing the identification method and the survival analysis method, and considering the endogeneity of fiscal subsidies. Our findings show that we should adhere to market-oriented mechanism in disposing zombie firms and cut off the channels that government transfuses to zombie firms by fiscal subsidies.

1 Introduction

As a key part of China’s economic restructuring, resolving surplus production capacities is the top priority of China’s supply-side structural reform. Many studies believe that the existence of zombie firms is an important cause of over-capacities (Geng, Liu, & Wu, 2021; Liu, Zhang, Zhang, & Wang, 2019; Shen & Chen, 2017). If market mechanism operates smoothly, then surplus production capacities will be automatically eliminated by market mechanism due to insufficient market demand (Shen & Chen, 2017). However, the existence of zombie firms broke this market feedback mechanism. Even when market demand is insufficient, these behindhand production capacities can rely on external support to survive. It means that effective cleaning up of zombie firms is a key step in resolving over-capacity. Moreover, judging from the reality in China, this step is imminent, because various calculation data (including the calculation later in this article) show that the proportion of zombie firms in Chinese firms is far from low.

To effectively clean up zombie firms, understanding the formation and exit mechanism of zombie firms is necessary. Existing literature shows that the focus of academic analysis is on the identification of zombie firms and their formation mechanism. However, only a few studies have been done on the exit mechanism of zombie firms (Carreira, Teixeira, & Nieto-Carrillo, 2022; Gouveia & Osterhold, 2018; Nurmi, Vanhala, & Virén, 2020). From the perspective of the formation of zombie firms, zombie firms mainly rely on low-interest loans provided by banks to survive (Caballero, Hoshi, & Kashyap, 2008), which is also the basis to identify zombie firms. In addition, government subsidies have been an important reason for the survival of zombie firms, particularly in a transitional economy such as China where government is able to intervene in the operation of market. China’s local governments often use fiscal subsidies to rescue zombie firms which are about to be bankrupt to ensure employment and maintain social stability. From the perspective of fiscal revenue and expenditure, as long as the cost of saving zombie firms is lower than the potential tax revenue, local governments have an incentive to help zombie firms survive. Despite the generality of fiscal subsidies to firms, fiscal subsidies and credit resources are likely to flow to low-efficiency companies, because government is affected by factors such as limited information and government failure (Robinson, 2009; Rodrik, 2008), which indirectly leads to the formation and continued survival of zombie companies.

However, to what extent do government’s fiscal subsidies affect the survival probability and exit probability of zombie firms? To what extent have the subsidies affected the survival period of zombie firms? Existing literature lacks in-depth discussion. Considering the impact of fiscal subsidies on the formation and exit of zombie firms, we identify zombie firms and measure the duration of zombie firms based on the data of industrial firms above the designated size in China from 1998 to 2007, and study the impact of fiscal subsidies on the duration of zombie firms by the survival risk model. The results show that fiscal subsidies significantly reduce the probability of zombie firms’ “death,” making them “stiff but not dead”; more importantly, fiscal subsidies significantly reduce the probability of zombie firms “resurrecting” to become normal firms, making them “stiff and dead.”

Compared with existing research, the contributions of this article lie in the following aspects. First, existing literature on the formation and exit mechanism of zombie firms is mostly based on bank credit. Only a few considerations have been taken from the perspective of government fiscal subsidies. This status does not match the reality that Chinese government provides various forms of fiscal subsidies to firms (Nurmi et al., 2020). Second, issues have been raised on the disposal of zombie firms. Most of the existing literature on exit mechanism use the absolute or relative quantity of zombie firms as research object. By contrast, we discuss the exit of zombie firms from the perspective of their survival risk, taking the survival period of zombie firms as the research object. Third, most of the existing literature focuses the research of zombie firms’ exit on “death” (i.e., exiting the market); however, we not only discuss the exit of zombie firms by “death,” but also by “resurrection” (i.e., returning to normal firms). The rest of this article is arranged as follows: in Section 2, we review the existing literature; in Section 3, we identify zombie firms in China’s industrial firms and measure the survival duration of zombie firms; in Section 4, we empirically analyze the impact of fiscal subsidies on the duration of zombie firms before they exit the market; in Section 5, we discuss the impact of fiscal subsidies on the probability of zombie firms “resurrecting” into normal firms; and Section 6 gives the conclusion.

2 Literature review

“Zombie firms,” which was first proposed by Kane (1987), generally refer to firms that cannot continue to operate on their own but are protected from bankruptcy due to external support, especially from government or banks. The most prominent zombie firm problem in history broke out in Japan in early 1990s. After the burst of economic bubble in the 1980s, a large number of Japanese firms fell into huge losses and even insolvency. However, due to the pressure from government’s industrial policy (Tett, 2004), banks concealed their own losses to meet the capital requirements of Basel Agreement (Sekine, Kobayashi, & Saita, 2003) and to avoid centralized liquidation of bankrupt firm assets. These measures resulted in collateral depreciation (Jaskowski, 2015) and other issues, thus forcing banks to continue to provide “Evergreen Lending” to firms that cannot resume normal operations and should enter bankruptcy procedures. With bank credit support, these firms continued to exist in Japanese economy and became zombie firms. In particular, Japan’s unique main banking system led to too close connections between banks and firms, which exacerbated the problems (Hoshi, 2006). Caballero et al. (2008) proposed a systematic method to identify zombie firms from the perspective of whether a firm has bank credit subsidies. Subsequent research conducted in-depth investigation on improving the impact, formation mechanism, and clean-up methods of zombie firms. In China, since central government proposed supply-side structural reforms in 2016 and used the disposal of zombie firms as the focus of “de-capacity” in supply-side structural reforms, zombie firms have attracted widespread attention from researchers and policymakers (Chang, Zhou, Liu, Wang, & Zhang, 2021; Jiang, Li, & Song, 2017; Lu, Li, & Qian, 2020; Shen & Chen, 2017; Woo, 2019). Nie, Jiang, Zhang, and Fang (2016) estimated that the proportion of zombie firms in China’s industrial firms declined by 30% during 2000–2013, and has remained stable after 2004. The proportion of zombie firms in industrial sectors during 2005–2013 was approximately 7.51%.

Zombie firms attract widespread attention because they seriously endanger the healthy operation of economy. First, zombie firms use non-market means to occupy a large amount of capital, land, labor, and other production factors, which distort market operation mechanism, cause distortions in resource allocation, and lead to the phenomenon of “bad money driving out good money” (Caballero et al., 2008; Nishimura, Nakajima, & Kiyota, 2005). Second, this form of market distortion not only lowers the productivity of zombie firms, but also depresses the growth of employment and investment in non-zombie firms, hinders the growth of normal firms and high-efficiency firms, impairs the overall efficiency and innovation level of the economy, reduces the aggregate productivity growth, and damages economic growth potential (Ahearne & Shinada, 2005; Kwon, Narita, & Narita, 2015). In addition, the backward production capacity of zombie firms has increased banks’ non-performing loan rate, which may cause systemic financial risks (Hoshi, 2006; McGowan, Andrews, & Millot, 2017). These consequences have been confirmed by the large number of zombie firms in Japan’s “lost decade” of the 1990s (Ahearne & Shinada, 2005; Caballero et al., 2008; Kwon et al., 2015; Peek & Rosengren, 2005).

Considering that zombie firms cause serious harm to the economy, how did they form? Research on Japanese zombie firms found that banks which rescue zombie firms and maintain the continuous operation of zombie firms have internal and external motivations. The internal motivation lies in banks’ need to renew zombie loans to meet regulatory requirements due to the pressure of non-performing loans; the external motivation comes from government’s pressure on banks to rescue zombie firms to avoid massive unemployment (Peek & Rosengren, 2005). However, the formation and survival of zombie firms is not only derived from the blood transfusion from banking sector, but also from the “assistant hand” of government. For the sake of its own interests, government may prevent firms from withdrawing from the market by providing direct financial assistance or deregulation (Brown & Dinc, 2011; Hoshi & Kashyap, 2010; Jaskowski, 2015; Willam, 2014). As a transitional economy, government intervention in the operation of China’s economy is prominent. Resource allocation in China is not yet fully driven by market mechanism but frequently relies on administrative instruction. Therefore, many studies have discussed the formation of zombie firms from the perspective of China’s institutional feature and economic structure. Nie et al. (2016) categorized five reasons for the formation of Chinese zombie firms; first, the collusion between local governments and firms; second, the vicious competition among China’s local governments and state-owned firms; third, the aftermath of the 2008 Chinese government’s economic stimulus package; fourth, the demand shock caused by the reduction in external demand after the financial crisis in 2008; and fifth, the credit discrimination of banks. Chang et al. (2021) pointed out that a high degree of government intervention may increase the probability of a firm becoming a zombie firm.

In response to the different reasons for the formation of zombie firms, existing literature proposed corresponding suggestions for cleaning up zombie firms. Using Japanese micro-firm data, Fukuda and Nakamura (2011) believed that reducing the number of employees and corporate fixed assets can promote the resurrection of zombie firms. Nakamura (2017) pointed out that although structural restructuring can help resurrect zombie firms, structural restructuring that lacks innovation may cause long-term deflation. In addition, transparent accounting standards and strict banking supervision policies can promote the revival of zombie firms.

The above review reveals the following findings. First, compared with the research on the identification, formation, and harm of zombie firms, fewer studies discuss the exit mechanism of zombie firms. Second, an important difference between the formation of zombie firms in China and that of zombie firms in other countries is that Chinese government’s intervention in economic operations has played an important role. Therefore, discussing the survival duration and exit mechanism of zombie firms from the perspective of government intervention in economy, especially from the perspective of government subsidies, is a great supplement to the existing research on zombie firms.

3 Data and methodology

3.1 Data

The data used in this article come from the Annual Surveys of Industrial Firms conducted by Chinese National Bureau of Statistics for the period from 1998 to 2007. This survey covers all state-owned industrial firms and non-state-owned firms with sales above 5 million RMB, which accounts for approximately 90% of the total sales of Chinese industrial firms. The survey contains extremely ample firm-level information, such as firm code, corporate nature, corporate profits, corporate sales, government subsidies, and industry categories. However, this database also has many problems, such as sample mismatch, missing indicators, abnormal indicators, and sample selection and measurement error. Additionally, some firms in the database changed their firm code in the period from 1998 to 2007, which made corresponding before and after data difficult. Drawing lessons from Brandt, Van Biesebroeck, and Zhang (2012), we first match the data over time. The firm is first cross-linked according to the unique numerical firm IDs and firm name. Firms that fail to match are further matched according to “address code + name of legal representative” and “address code + phone number + year of opening.” Second, China’s national economic industry classification standards changed in 2002. For consistency, we unify the two industry classifications to the 2011 standard (GB/T4754-2011) in accordance with the industry classification principles disclosed by Chinese National Bureau of Statistics. In addition, to ensure the data quality of the empirical analysis, in line with Cai and Liu (2009), we delete the two types of outliers in the industrial firm database.

First, we delete the missing observations of important indicators. We delete the missing observations of key indicators such as total assets, number of workers, industrial production, net fixed assets, or sales. Among them, the data in 2004 were combined with the data in other years without matching and missed the data of industrial production. We use the following formula to estimate the industrial production in 2004:

Industrial production = Product sales Beginning inventory + Ending inventory Industrial intermediate input + Value-added tax .

Second, we delete incorrect records that do not comply with accounting standards and logical relationships. We delete observations with fixed assets less than 10 million, total assets less than 10 million, and employees fewer than 10. We also delete observations with abnormal accounting indicators, for example, total assets minus current assets less than 0, total assets less than net fixed assets, original fixed assets less than net fixed assets, accumulated depreciation less than current depreciation, and subsidy income less than 0.

3.2 Identification of zombie firms

The existing literature adopt the following methods for measuring zombie firms.

  1. CHK method. CHK method is proposed by Caballero, Hoshi, and Kashyap in 2008 (Caballero et al., 2008), whose core idea is that if a firm pays interest on its debts below the interest calculated by market minimum interest, then the firm has obtained credit concessions from a bank. Once the credit concessions cease, such firms are likely to go bankrupt. Therefore, whether a firm has obtained non-market credit from a bank can be used as a basis for judging whether a firm is a zombie firm. The identification method involves the following steps. First, calculate the minimum interest payable by the firm; second, obtain the actual interest paid by the firm according to its financial statements. The difference between the minimum interest payable and the actual interest paid by the firm is the interest subsidy obtained by the firm from bank. Finally, bank standardizes the interest subsidies for firms to obtain the interest rate differential, and the firms with interest rate differential less than 0 are defined as zombie firms.

  2. FN–CHK method. Fukuda and Nakamura (2011) pointed out that CHK method may lead to misjudgments. On the one hand, non-zombie firms may easily be misjudged as zombie firms because banks tend to provide preferential treatment to firms with better qualifications and lower risk of default. To support key or emerging industries, government tends to issue low-interest loans to growth-type firms (Nie et al., 2016). According to CHK method, these two types of firms will be mistaken as zombie firms. On the other hand, some real zombie firms may be missed because zombie firms can borrow credit in exchange for old debts through the extensions of maturing loans. However, such zombie firms cannot be identified by using CHK method. In view of this limitation, Fukuda and Nakamura (2011) added the “profit standard” and “evergreen loan standard” to CHK method to form FN–CHK method. The “profit standard” identifies firms whose pre-interest and tax income is higher than the lowest interest payable in CHK method as normal firms. The “evergreen lending standard” identifies firms whose debt-to-asset ratio was more than 50% in last year and continued increasing borrowing in current year as zombie firms.

  3. FN–CHK correction method. Nakamura and Fukuda (2013) pointed out that some zombie firms identified according to CHK method and FN–CHK method are “one-shot zombie firms,” that is, during the sample period, these firms were only recognized as a zombie firm in a certain year. Nie et al. (2016) pointed out that “one-off zombie firms” are likely to be normal firms that encountered short-term problems in their operations or suffered short-term external shocks. Therefore, Nakamura and Fukuda (2013) made corrections on the basis of FN–CHK method, which is called FN–CHK correction method. According to FN–CHK correction method, if a firm is identified as a zombie firm by FN–CHK method in year t − 1 and year t, then the firm is identified as a zombie firm in year t.

Although FN–CHK method is not perfect, it captures the essential characteristics of zombie firms obtaining continuous credit support in the case of sluggish returns. FN–CHK method is also the basis of other methods. FN–CHK correction method removes “one-off zombie firms” from zombie firms on the basis of FN–CHK method. Combined with the purpose of this article, we mainly analyze zombie firms identified by using FN–CHK method. The specific identification steps are as follows:

The first step is to calculate the minimum interest RA i,t required to be paid by firm i under normal operations in year t according to the following equation (1):

(1) RA i , t = rs t 1 BS i , t 1 + j = 1 5 rl t j / 5 BL i , t 1 ,

where BL i,t is the long-term bank loan of firm i at year t, which is directly measured by firm’s long-term debt. rs t and rl t are 90% of the average 1 year and 5 year long-term benchmark loan interest rates of banks in year t, respectively[1]. BS i,t is the short-term bank loan of firm i at year t. In view of the unavailable bank liabilities data in the Annual Surveys of Industrial Firms, we use short-term corporate liabilities minus accounts payable and other payables as short-term borrowings. The operating liabilities considered in this article include accounts payable, value-added tax payable, income tax payable, wages payable, and welfare fees payable. However, accounts payable from 1998 to 2003 are missing in the survey. We supplement these missing data by the following method: for firms that remain in the sample before 2004, we use the ratio of accounts payable to current liabilities in subsequent years to fill in the missing data; for firms that are not in the sample before 2004, we use the ratio of the average four-digit industry accounts payable to current liabilities to fill in the missing data. In the survey, except for accounts payable, other payables are the cumulative amount for the whole year, not the year-end balance. According to the payment time of each item, we regard the year-end balance of all corporate income tax payable as one-fourth of the annual cumulative amount (quarterly payment), the year-end balance of value-added tax payable, wages payable, and welfare fees payable as one-twelfth of the annual cumulative amount (monthly payment).

The second step is to estimate the corporate interest income according to equation (2):

(2) RB i , t = ( AT i , t 1 AR i , t 1 AI i , t i 1 ) × rd t ,

where AT i,t , AR i,t , and AI i,t represent the firm’s current assets, accounts receivable, and inventory, respectively; and rd t represents bank’s one-year benchmark deposit interest rate in year t. Only data about firm’s actual net interest expenditure (firm’s interest expenditure minus interest income) are found in the survey. Hence, to compare firm’s actual interest expenditure with benchmark interest expenditure, we need to estimate firm’s interest income RB i,t . The estimation method of RB i,t entails the following steps: accounts receivable and inventory at year t − 1 are deducted from current assets at year t − 1, the difference is multiplied by 1 year benchmark bank deposit interest rate of year t.

The third step is to compare firm’s actual net interest expense Rc i,t with the calculated minimum net interest expense (RA i,t − RB i,t ). Additionally, the difference with firm’s previous borrowing B i,t−1 should be standardized to obtain interest difference:

(3) gap i , t = [ RC i , t ( RA i , t RB i , t ) ] / B i , t 1 ,

According to Caballero et al. (2008), gap i,t < 0 indicates that the firm has received subsidies, and its zombie firm index is 1; otherwise, the firm is not a zombie firm, and its zombie firm index is 0.

The fourth step is to modify gap i,t using the information of corporate profits to obtain gapadj i,t :

(4) gapad j i , t = ( EBIT i , t ( RA i , t RB i , t ) ) / B i , t 1 ,

where EBIT i,t is the firm’s profit before interest and tax. If profit before interest and tax is greater than minimum net interest expense (RA i,t − RB i,t ), that is, gapadj i,t > 0, then the firm is changed to a non-zombie firm. The “evergreen lending standard” proposed that if firm’s profit before interest and tax is less than minimum net interest expense, and firm’s liabilities exceed 50% of total assets, then t period liabilities are greater than t − 1 period debt, and the firm is determined as a zombie firm in period t (Fukuda & Nakamura, 2011). We define the above standard for identifying zombie firms as Z1 standard.

Table 1 shows the summary of zombie firms identified by Z1 standard. Generally, the number of zombie firms identified in 1999 was the biggest, and the proportion reached 18.27%. The number of zombie firms in later years showed a steady decline until 2005, which was in line with the findings of Nie et al. (2016). Among these zombie firms, the proportion of subsidized zombie firms showed an overall upward trend, reaching a peak in 2004. During this year, zombie firms that received government subsidies accounted for more than 16% of all zombie firms. Thereafter, it declined slightly but remained above 14%.

Table 1

Summary of zombie firms identified by Z1 standard

Year Number of all firms Number of zombie firms Ratio of zombie firms (%) Number of zombie firms with subsidies Ratio of zombie firms with subsidies (%)
1999 114,740 20,959 18.27 2,746 13.10
2000 128,355 17,893 13.94 2,310 12.91
2001 136,825 15,002 10.96 1,787 11.91
2002 144,395 14,895 10.32 1,905 12.79
2003 160,132 13,544 8.46 1,794 13.25
2004 213,332 13,553 6.35 2,180 16.08
2005 232,859 18,809 8.08 2,707 14.39
2006 221,824 18,463 8.32 2,608 14.13
2007 202,514 14,131 6.98 2,158 15.27

Data source: Chinese industrial firm database.

3.3 Duration of zombie firms

The identified zombie firms are used to measure the duration of zombie firms. First, we define the duration of a zombie firm. If a firm has been in a zombie status for s consecutive years at t, t + 1, t + 2, … t + s (neither “death” nor “resurrection”), then the duration of zombie firm during the period (t, t + s) is s years. The longer the duration of a zombie firm, the more difficult it is for zombie firm to return to a normal status (stiff but not alive) or to exit the market (stiff but not dead).

When using the data from the Annual Surveys of Industrial Firms, special attention should be paid to the problem of data censorship. Given that the firms in this article are industrial firms above designated size from 1999 to 2007, the status of firms outside 1999–2007 are impossible to know. Thus, if a firm was a zombie firm in 1999, then the exact status of this zombie firm cannot be known. If this problem is ignored, then the duration of zombie firms will be underestimated, namely, left censoring problem. We removed left-censored observations from the sample, that is, selected firms that were not zombie firms in 1999 but became zombie firms during 2000–2007. Therefore, the duration of zombie firms identified in this article using Z1 standard is up to 8 years. If a firm remains as zombie firm in 2007, then the exact time when the firm stopped bottling is also impossible to know, which is the so-called right censoring problem. Nonetheless, the survival analysis methods can well deal with right censoring.

Survival analysis usually uses survival function (survival rate) or hazard function (risk rate) to describe the distribution characteristics of individual survival duration. We construct survival function and hazard function of the duration of zombie firms to estimate the distribution characteristics of the duration of zombie firms in China. Let T represent the survival time of a firm in a specific period of time, with value t (t = 1, 2, 3…, i) representing the duration of a specific zombie firm. If a duration is complete, then it is recorded as c i = 1. Right censoring is recorded as c i = 0.

Probability function f(j) for the duration of zombie firm represents the probability that zombie firm exits market in year j (denoted by P). The formula is as follows:

(5) f ( j ) = P ( T = j ) ,

The corresponding survival rate represents the probability that the zombie status of a zombie firm exceeds t years, expressed by S(j). The definition is as follows:

(6) S ( j ) = P ( T > j ) = k = j + 1 f ( k ) .

The risk rate refers to the probability of zombie firm exiting market at year j (to eliminate zombieization) on the premise that the firm has been in zombie status for j years. Risk rate is expressed by h(j):

(7) h ( j ) = P ( T = j | T j ) = f ( j ) / S ( j 1 ) .

Let n k represent the number of firm segments that remain in zombie status at the beginning of the kth year, and d k is the number of zombie firm durations that only last until year k. The non-parametric estimation of survival rate can be given by the following Kaplan–Meier (KM) formula:

(8) S ˆ ( j ) = k j n k d k n k .

According to the relationship between the survival rate and the risk rate in equation (7), we can estimate the value of risk rate, which is expressed by equation (9):

(9) h ˆ ( j ) = d j / n j .

3.4 Empirical model of factors affecting the duration of zombie firms

The commonly used estimation methods to determine the factors affecting the duration are Cox proportional hazard model and discrete-time survival risk model. Hess and Persson (2012) pointed out that discrete-time survival risk model is better than proportional hazard model. Discrete-time survival model can deal with node problems more effectively because the former allows easier control of unobserved heterogeneity and does not need to meet “proportional risk” assumptions required by Cox proportional hazard model. Therefore, we use discrete-time survival risk model to estimate the influencing factors of zombie firms’ duration, specifically, we use logit model to estimate the factors. Suppose the total duration of zombieization duration in the sample is n (n = 1, 2, 3, …), for a specific duration i (i = 1, 2, 3, …), we need to study its survival status until jth (j = 1, 2, 3, …) year; hence, for each duration in the sample, the time range that needs to be observed is 1 to j. Assuming that T i is the duration of zombie firm i, the risk rate of duration period i in the jth year refers to the probability of this zombie firm exiting from market in the jth year under the premise that this firm keeps zombie status for at least j years.

Using c i,j to indicate whether period i is censored in the jth year, if duration is right censored (that is, whether a firm was a zombie firm in the jth year, including exit and resurrection as a normal firm is uncertain), then c i,j = 0. If the period is complete (that is, the zombie firm is interrupted in the jth year), then c i,j = 1. According to Jenkins (1995), the log-likelihood function ln L of the available sample is:

(10) ln L = i = 1 n c i , j ln h i , j 1 h i , j + i = 1 n k = 1 j ln [ 1 h i , j ] .

The log-likelihood function in equation (10) is obtained at the level of duration, which is only suitable for continuous methods. To convert equation (10) into a form suitable for discrete method, we define a dichotomous variable y i,k to reflect the interruption of duration i in year k. If the withdrawal of a firm as zombie firm in the jth year (c i,j = 0) is uncertain, then y i,1 = y i,2 = … = y i,j−1 = y i,j = 0. If duration i is the first interruption that occurs in year j (c i,j = 0), then y i,1 = y i,2 = … = y i,j−1 = 0. That is:

(11) c i , j = k = 1 j y i , k .

According to equations (10) and (11) can be further transformed into:

(12) ln L = i = 1 n k = 1 j y i , k ln h i , k 1 h i , k + i = 1 n k = 1 j y i , k ln [ 1 h i , k ] = i = 1 n k = 1 j [ y i , k ln h i , k + ( 1 y i , k ) ln ( 1 h i , k ) ] .

This setting indicates that risk rate h i,k of duration i in year k is essentially the probability of y i,k = 1; hence, 0 ≤ h i,k ≤ 1. According to the setting of discrete independent variable model, let

(13) h i , k = F ( α + v i + β X + ε i , k ) ,

where F(·) is the distribution function. Given that we use logit model for analysis, (·) refers to logistic distribution. The logit transformation of equation (13) can be obtained by equation (14):

(14) Logit ( h i , k ) = ln h i , k 1 h i , k = α + v i + β X + ε i , k ,

where α is a constant term, X is a set of control variables, and β is its coefficient. The disturbance terms v i and ε i,k are subject to normal distribution. In the control variable set X, in addition to fiscal subsidies (Sub), we introduce the following control variables (Agarwal & Gort, 2002; Audretsch & Mahmood, 1995): firm age (Age), firm size (Size), firm’s capital intensity (K/L), firm’s wage level (Wage), and firm’s ownership (SOE). Fiscal subsidy is measured by the ratio of government subsidy received by a firm to its total assets. Firm age is obtained by subtracting the year of firm’s establishment from the year of observation. Firm size is measured by the natural logarithm of employee number. Capital intensity is measured by the natural logarithm of firm’s assets per capita. Wage level is measured by the logarithm of firm’s per capita wages payable. Profit rate is measured by the ratio of firm’s operating profit to firm’s operating income. Growth is measured by the growth rate of firm’s total operating income. Debt-to-asset ratio is measured by the ratio of firm’s total liabilities to its total assets. If a firm is a state-owned firm, then the value of the variable SOE takes 1; otherwise, variable SOE takes 0. In addition, we control year fixed effect, industry fixed effect, and regional fixed effect. The industry refers to two-digit code and the region refers to province where the firm is located.

4 Impact of fiscal subsidies on duration of zombie firms prior to death

4.1 Stylized facts

Unlike normal economic duration, zombie duration can end in two ways: one is exiting market, namely, zombie firm “dies”; the other is becoming a normal non-zombie firm, which can be called “resurrection.” In this section, we first study the impact of fiscal subsidies on the duration of zombie firm before dying to determine whether fiscal subsidies accelerate or delay the clearing of zombie firms. Using all zombie firms identified in this article and excluding left-censored data, we can obtain the survival rate of each year by KM survival estimation. Given this section studies the impact of fiscal subsidies on the “death” probability of zombie firms, so we only consider those zombie firms who end zombie status by exiting market, but not consider those zombie firms who end zombie status by “resurrecting,” which means that there is no multiple duration issues in this section[2]. Table 2 lists the basic information about the distribution of zombie firms, including the average, median, and survival rates of 1, 3, and 5 years.

Table 2

Estimation of the duration of zombie firms prior to death

Duration (year) Survival rate
Mean Median 1 year 3 years 5 years
All zombie firms 1.842 2 0.733 0.406 0.243
Zombie firms with subsidies 1.950 2 0.795 0.493 0.326
Zombie firms without subsidies 1.797 2 0.708 0.370 0.207

Table 2 shows that the average duration of zombie firms before “death” is 1.842 years, with a median value of 2 years. This result indicates that most zombie firms will exit the market within 2 years after they fall into zombification. The average duration of subsidized zombie firms is 1.95 years, whereas the duration of unsubsidized zombie firms is 1.797 years.

4.2 Baseline result

Using the data of identified zombie firms, we estimate equation (14), and the estimation results are shown in Table 3. Column (1) only introduces fiscal subsidies, which shows that fiscal subsidies significantly reduce the risk rate (probability of death) of zombie firms, indicating that fiscal subsidies are not conducive to zombie firms’ exit from market and prolong the duration of zombie firms. Column (2) gives the results of odds ratio which represents the ratio of the probability of a zombie firm quitting zombie status and the probability of it remaining in zombie status. If the odds ratio is less than 1, then the explanatory variable has negative impact on the probability of a zombie firm becoming a normal firm, thus prolonging the duration of the zombie firm; if the odds ratio is greater than 1, then the explanatory variable has positive impact on the probability of zombie firms becoming normal firms. The results in column (2) show that odds ratio of fiscal subsidies is less than 1, indicating that fiscal subsidies have a negative impact on zombie firms leaving the zombie status and extend the duration of zombie firms. The coefficient regression results given in column (3) introduce other control variables, which show that the effect of fiscal subsidies on the risk rate of zombie firms remains significantly negative, and the corresponding odd ratio is still less than 1.

Table 3

Impact of fiscal subsidies on zombie firms’ death probability

(1) Coefficient (2) Odds ratio (3) Coefficient (4) Odds ratio
Sub −4.5418*** 0.0107*** −4.1472*** 0.0158***
(0.667) (0.007) (0.655) (0.010)
Size −0.3379*** 0.7133***
(0.012) (0.009)
Age 0.0009*** 1.0009***
(0.000) (0.000)
SOE 0.7260*** 2.0668***
(0.026) (0.054)
K/L 0.0372*** 1.0379***
(0.008) (0.008)
Wage −0.4300*** 0.6505***
(0.015) (0.010)
Year fixed effect Yes Yes Yes Yes
Industry fixed effect Yes Yes Yes Yes
Region fixed effect Yes Yes Yes Yes
Observations 98,664 98,664 97,785 97,785

Note: robust standard errors are in parentheses; *** p < 0.01.

Among the control variables, firm size (Size) has a significantly negative impact on the exit of zombie firms, because large zombie firms usually have strong anti-risk capabilities and can survive for a long time even in zombie status. Firm age (Age) has a significant positive influence on the exit behavior of a firm. The possible explanation is that the longer the establishment of a firm, the more likely it is to form a path dependence on the system, and more difficult it is to adopt advanced technology than new firms, and the more difficult it is to respond to changes in demand in a timely manner, and the more difficult it is to transform and upgrade industrial structure than new firms. In addition, the impact of firm’s wage on the exit of zombie firms is significantly negative. The possible explanation is that firms with higher wages are more likely to hire highly skilled labor and adopt knowledge-intensive technology, and their production efficiency is higher. Hence, they can survive for a long time even in zombie status. They are not easy to withdraw from the market.

4.3 Robustness check

We conduct robustness analysis from two aspects. First, in benchmark regression, the distribution function F(·) is set as logistic distribution. In addition to logistic distribution, the distribution function F(·) can also be normal distribution and extreme value distribution in discrete-time survival risk model. These two distributions correspond to probit model and cloglog model (complementary log–log), which are commonly used in existing studies. Therefore, we use probit model and cloglog model to estimate parameter β again, and the estimation results are shown in Table 4. The results in Table 4 show that the direction and significance of coefficients of fiscal subsidies are consistent with the benchmark regression. Therefore, changing the estimation method will not affect the robustness of benchmark regression.

Table 4

Robustness check (changing estimation method)

(1) Cloglog regression (2) Probit regression
Sub −4.1833*** −1.8193***
(0.637) (0.276)
Size −0.3440*** −0.2031***
(0.011) (0.007)
Age 0.3327*** 0.2024***
(0.011) (0.006)
SOE 0.3455*** 0.2052***
(0.026) (0.015)
K/L 0.0221*** 0.0156***
(0.007) (0.004)
Wage −0.3531*** −0.2293***
(0.013) (0.009)
Year fixed effect Yes Yes
Industry fixed effect Yes Yes
Region fixed effect Yes Yes
Observations 97,732 97,732

Note: robust standard errors are in parentheses; *** p < 0.01.

The second robustness analysis is changing the method of identifying zombie firms. The identifying method in benchmark regression is based on Z1 standard, we change the method by two ways. First, we use FN–CHK correction method proposed by Nie et al. (2016) to identify zombie firms. If a firm has been identified as a zombie firm by Z1 standard for two consecutive years, then it is identified as a zombie firm by FN–CHK correction method. We record this method as Z2 standard. Second, we make adjustment on Z1 standard according to actual profit method. We identify firms whose net profit is less than 0 for three consecutive years after deducting financial subsidies as zombie firms, and record this method as standard Z3. Table 5 shows the estimated results based on zombie firms identified by these two standards. The results in Table 5 show that even if the method of identifying zombie firms is changed, the impact of the fiscal subsidy remains significantly negative.

Table 5

Robustness checking (changing the identification standard of zombie firms)

Z2 standard Z3 standard
(1) Coefficient (2) Odds ratio (3) Coefficient (4) Odds ratio
Sub −3.6607*** 0.0251*** −1.8841*** 0.3459***
(1.132) (0.028) (0.310) (0.098)
Size −0.3470*** 0.7008*** −0.4630*** 0.5998***
(0.021) (0.015) (0.014) (0.009)
Age 0.3426*** 1.4085*** 0.3516*** 1.4109***
(0.020) (0.029) (0.014) (0.020)
SOE 0.4791*** 1.6016*** 0.4725*** 1.4856***
(0.046) (0.074) (0.033) (0.049)
K/L 0.0066 1.0141 −0.0099 1.0386***
(0.013) (0.013) (0.009) (0.009)
Wage −0.4161*** 0.6576*** −0.4703*** 0.6084***
(0.026) (0.017) (0.020) (0.012)
Year fixed effect Yes Yes Yes Yes
Industry fixed effect Yes Yes Yes Yes
Region fixed effect Yes Yes Yes Yes
Observations 35,187 84,184 84,184 84,184

Note: robust standard errors are in parentheses; *** p < 0.01.

4.4 Treatment of endogeneity

The benchmark regression may face endogeneity issues. On the one hand, governments will help firms that are about to go bankrupt through fiscal subsidies for the sake of sustaining employment. On the other hand, fiscal subsidies and the survival time of zombie firms may be affected by common factors. It means that government’s fiscal subsidies to zombie firms are not exogenous, but depend on the survival status of zombie firms. To solve this problem, we use propensity score method (PSM) to select samples of zombie firms that have not received fiscal subsidies. We match a group of unsubsidized zombie firms with similar characteristics to firms that have received fiscal subsidies. Then, we use the data of subsidized zombie firms and matched unsubsidized zombie firms for survival risk analysis. These two groups of zombie firms only differ in whether they receive fiscal subsidies; hence, the difference in survival time between these two groups can be attributed to the effect of fiscal subsidies.

Specifically, we divide zombie firms into a treatment group and a control group according to whether they receive fiscal subsidies. The treatment group includes zombie firms that have received fiscal subsidies. We define a binary dummy variable as Sub i,t = {0, 1} as following: when zombie firm i receives fiscal subsidies in period t, Sub i,t takes 1; otherwise it takes 0. Therefore, the impact of fiscal subsidies on the survival duration of zombie firm i in period t is T i , t 1 T i , t 0 . Among them, T i , t 1 is the survival duration of subsidized zombie firm i in period t and T i , t 0 is the survival duration of subsidized zombie firm i if it does not receive fiscal subsidy in period t. The average treatment effect of fiscal subsidies on the duration of zombie firms is:

(15) E ( T i , t 1 T i , t 0 | S u b i , t = 1 ) = E ( T i , t 1 | S u b i , t = 1 ) E ( T i , t 0 | S u b i , t = 1 ) .

However, E ( T i , t 0 | S u b i , t = 1 ) in equation (15) is unobservable, so we need to construct a “counterfactual” observation value. This observation value can be obtained by fitting the duration of zombie firms that have not received fiscal subsidies. We use nearest neighbor matching to match zombie firms in the control group that are similar to zombie firms in the treatment group. The covariate used to calculate the propensity score should simultaneously affect the fiscal subsidy of the zombie firms and the duration of the zombie firm. We select firm size (Size), firm age (Age), firm capital intensity (K/L), firm debt ratio (Debt), and firm ownership (SOE) as matching covariates. We then use logit method to estimate equation (16):

(16) P = Pr { S u b i , t = 1 } = Φ { Z i , t } ,

where Z is a set of covariates. Estimating equation (16) can obtain probability prediction value P ˆ , and the principle of nearest neighbor matching can be expressed as:

(17) Ω ( i ) = min j P ˆ i P ˆ j , j ( S u b = 0 ) ,

where P ˆ i and P ˆ j represent the probability prediction values of treatment group and control group, respectively. Ω(i) represents the matching set of zombie firms from control group corresponding to zombie firms of treatment group. For each treatment group i, only unique control group j enters set Ω(i). To test the reliability of matching results, we conducted a balance test on matching results. The condition of matching balance is Sub i t X i , t | P ( X i , t ) . Given the probability P(X i , t) of zombie firm obtaining fiscal subsidies, whether zombie firm actually obtains government subsidies and its feature vector are independent of each other. Table 6 shows the results of the balance test. The p-value of the t-test used to test whether the difference between treatment group and control group is significant is greater than 10% after matching, which indicates that after matching, the matching variables have no significant differences between treatment group and control group. In addition, after matching, the absolute value of the mean deviation between the treatment group and control group is less than 10%, which is lower than 20%, namely, the standard proposed by Rosenbaum and Rubin (1985) for a good matching effect. Therefore, the matching effect is in line with the requirement.

Table 6

Balance test of matching variables

Mean
Treatment group Control group Standard bias (%) t statistics p-value
Size Unmatched 5.2685 5.6733 −23.42 −14.75 0
Matched 5.2685 5.3014 −1.81 −1.17 0.121
Age Unmatched 16.5714 18.1610 −18.86 −10.41 0
Matched 16.5714 16.5056 −0.79 −0.40 0.345
K/L Unmatched 9.2276 11.7412 −27.61 −15.67 0
Matched 9.2276 9.3006 −8.13 −0.44 0.330
Wage Unmatched 2.3550 2.7008 −5.26 −3.28 0.001
Matched 2.3550 2.3511 0.06 0.04 0.484
SOE Unmatched 0.2704 0.2173 11.90 5.97 0
Matched 0.2704 0.2685 4.31 0.21 0.417

Note: samples in matching are the zombie firms identified by the FN–CHK method.

Using the matched samples, we estimate equation (14) again, and the results are shown in Table 7. In Table 7, we not only list the regression results based on logit model, but also the results of cloglog regression and probit regression. The regression results show that in these three types of regression, fiscal subsidies have a significant inhibitory effect on the mortality of zombie firms.

Table 7

Regression results based on matched samples

(1) Logit regression (2) Cloglog regression (3) Probit regression
Sub −4.0027*** −5.2689*** −4.5715***
(0.466) (1.162) (0.830)
Size −0.7123*** −0.8881*** −0.8305***
(0.173) (0.149) (0.094)
Age 0.0013*** 0.0044*** 0.0065***
(0.000) (0.007) (0.001)
SOE 0.1434*** 0.1588*** 0.2283***
(0.036) (0.027) (0.051)
K/L 0.0096*** 0.0143*** 0.0075***
(0.003) (0.003) (0.002)
Wage −0.1233* −0.1859*** 0.0943***
(0.067) (0.045) (0.011)
Year fixed effect Yes Yes Yes
Industry fixed effect Yes Yes Yes
Region fixed effect Yes Yes Yes
Observations 38,154 38,154 38,154

Note: robust standard errors are in parentheses; *** p < 0.01, * p < 0.1.

5 Impact of fiscal subsidies on the duration of zombie firms prior to resurrection

5.1 Stylized facts

Section 4 shows that fiscal subsidies reduce the “death” probability of zombie firms and lengthen the duration before zombie firms’ “death.” However, firms’ exit from zombie status has two forms: one is exiting from the market, in which zombie firms “die”; the other is becoming normal non-zombie firms, that is, zombie firms “resurrect.” The goal of fiscal subsidies is to neither retain firms in zombie status nor clear zombie firms from the market. The ideal result is helping zombie firms to become normal firms and continue to generate profits. Therefore, studying the impact of fiscal subsidies on the duration of zombie firms’ “resurrection” makes more sense.

In Section 4, once a zombie firm “dies,” it is unlikely to re-enter the market. Even if it re-enters, it does not use the same corporate code. However, if a zombie firm becomes a normal firm, it may become a zombie firm again, and after becoming a zombie firm again, the same firm code is still used in the Annual Surveys of Industrial Firms. Therefore, in this section, we will face the problem of multiple duration of zombie status. To solve the problems of multiple durations, we select firms that continuously operated from 1998 to 2007, which means that these firms, whether they were zombie firms or not, had not withdrawn from market in the 10 years, namely, did not “die” in the 10 years. Besede and Prusa (2006) pointed out that when experienced multiple durations, the distributions of durations were similar if we regarded the first duration as the only duration or different durations of the same firm was regarded as several independent durations. Therefore, we can treat multiple duration periods of the same firm as independent durations.

Given that the sample size of firms that have survived for 10 consecutive years is much small in the total sample, the actual profit method (i.e., Z3 standard to identify zombie firms) requires three consecutive years of data, which will further reduce the research sample. This section mainly uses Z1 standard (i.e., FN–CHK method) to identify zombie firms to ensure enough samples. Table 8 shows the information of zombie firms identified by Z1 standard from firms that continuously existed in the database from 1998 to 2007.

Table 8

Ratio of fiscal subsidies and zombie firms prior to resurrection

Year Number of all firms Number of zombie firms (%) Ratio of zombie firms Number of zombie firms with subsidies (%)
1999 5,240 16.20 760 14.50
2000 3,593 11.11 557 15.50
2001 3,351 10.36 503 15.01
2002 3,121 9.65 517 16.57
2003 3,040 9.40 508 16.71
2004 3,095 9.57 571 18.45
2005 3,227 9.98 672 20.82
2006 3,282 10.15 610 18.59
2007 3,116 9.64 581 18.65

Data source: Chinese industrial firm database.

In Table 8, the trend of the proportion of zombie firms is close to that in Table 2, but the proportion of subsidized zombie firms is significantly higher than that in Table 2. The possible reason is that the longer the duration, the higher the number of firms that may obtain fiscal subsidies from government. Unlike the change in the proportion of zombie firms, the number of subsidized zombie firms has increased significantly after 2004, reaching a peak of 672 in 2005. The proportion of subsidized zombie firms has also exceeded 20%. These results confirm that government has increased its support for zombie firms since 2004.

Table 9 shows the duration distribution of zombie firms before their “resurrection,” which reveals that approximately 34.69% of the firms had only a single 1 year zombie status. The proportion of single zombie firms with a duration of 2 years is 8.38%. These firms have been in trouble for a long time, but they can still become normal firms through their own efforts or with the help of external forces. Only a few zombie firms have a single duration of more than three years, accounting for less than 5% of the sample. Moreover, the proportion of zombie firms with multiple durations to all zombie firms exceeds 50%, which indicates that a large number of firms can leave zombie status quickly after falling into zombie status. However, this status is temporary. Switching between zombification and non-zombification shows that the economic benefits of these firms are not high, and exiting from the market or fully returning as normal firms is difficult.

Table 9

Duration distribution of zombie firms prior to resurrection

Duration Number of durations Percent Cumulative percent
Single (1 year) 3,789 34.69 34.69
Single (2 years) 915 8.38 43.06
Single (3 years) 321 2.94 46.00
Single (4 years) 135 1.24 47.24
Single (5 years) 63 0.58 47.81
Single (6 years) 29 0.27 48.08
Single (7 years) 19 0.17 48.25
Single (≥8 years) 14 0.13 48.38
Multiple 5,639 51.62 100.00
Summary 10,924 100

Data source: Chinese industrial firm database.

The KM estimation provides the duration distribution of all samples without left censoring, which is shown in Table 10. According to the overall sample, zombie firms’ duration is short, and the probability that the duration exceeds 1 year is 40.4%, that is, 59.6% of firms can become normal firms in merely 1 year. Although the existence of multiple durations means that some firms will become zombie firms again after they exit the zombie status, firms do not easily fall into a long-term zombie status.

Table 10

Estimation of the duration of zombie firms prior to resurrection

Duration Survival rate Number of durations Number of zombie firms
Mean Median 1 year 3 years 5 years
Full sample 1.45 1 0.404 0.116 0.047 10,924 7,861
Single duration 1.48 1 0.400 0.130 0.064 5,285 5,285
First duration 1.46 1 0.365 0.089 0.032 7,861 7,861

Note: we only use the “First duration” if a firm has several durations as a zombie firm; the “Single duration” refers to a firm that does not become a zombie firm again if it has previously transformed from a zombie firm to a normal firm.

5.2 Baseline result

Next we use the previous discrete-time survival model to study the impact of fiscal subsidies on the duration of zombie firms before “resurrection.” Column (1) in Table 11 only introduces fiscal subsidy, which shows that the coefficient of fiscal subsidy is negative at the significance level of 1%, indicating that the more fiscal subsidies a zombie firm receives, the more difficult it is to leave zombie status through “resurrection.” Column (3) introduces more control variables, and it shows that fiscal subsidies still significantly suppress the probability of zombie firms “resurrecting.” Columns (2) and (4) provide the estimation results of odds ratio. The odds ratio of fiscal subsidy remains significantly less than 1, which shows that fiscal subsidy significantly inhibits zombie firms’ probability of “resurrection.” Why do fiscal subsidies inhibit the “resurrection” of zombie firms? We think the possible explanation is that firms receiving fiscal subsidies have no incentive to innovate technology and improve production. Additionally, they can survive for a long time with government’s subsidies without exiting market, resulting in zombie status.

Table 11

Impact of fiscal subsidies on zombie firms’ resurrection probability

(1) Coefficient (2) Odds ratio (3) Coefficient (4) Odds ratio
Sub −4.0265*** 0.0178*** −4.0004*** 0.0183***
(0.984) (0.018) (0.986) (0.018)
Age 0.0004* 1.0004*
(0.000) (0.000)
Size −0.1362*** 0.8727 ***
(0.037) (0.032)
K/L 0.0644*** 1.0665***
(0.024) (0.026)
Wage −0.1754*** 0.8391***
(0.030) (0.025)
SOE 0.0948 1.0994
(0.07) (0.082)
Year fixed effect Yes Yes Yes Yes
Industry fixed effect Yes Yes Yes Yes
Region fixed effect Yes Yes Yes Yes
Observations 61,249 61,249 60,633 60,633

Note: robust standard errors are in parentheses; *** p < 0.01, * p < 0.1.

5.3 Robustness check

To examine the robustness of the estimation results in Table 11, we conducted a variety of robustness tests. First, we changed the identification standard of zombie firms from Z1 standard to Z2 standard. Based on zombie firms identified by Z2 standard, we reused discrete-time survival risk model to estimate the probability of zombie firms “resurrecting” as normal firms, which is exhibited in Table 12, and the impact of fiscal subsidies on the probability of zombie firms’ “resurrection” remains significantly negative.

Table 12

Impact of fiscal subsidies on the probability of resurrection of zombie firms (Z2 standard)

(1) Coefficient (2) Odds ratio
Sub −5.5977*** 0.0037***
(2.169) (0.008)
Age 0.0006 1.0005
(0.000) (0.000)
Size 0.4141*** 1.5130***
(0.083) (0.125)
K/L 0.5361*** 1.7094***
(0.058) (0.099)
Wage −0.1245** 0.8829**
(0.056) (0.050)
SOE −0.0861 0.9175
(0.142) (0.130)
Year fixed effect Yes Yes
Industry fixed effect Yes Yes
Region fixed effect Yes Yes
Observations 21,759 21,759

Note: robust standard errors are in parentheses; *** p < 0.01, ** p < 0.05.

In addition, in this section, we deleted the left-censored data, which helps in improving the accuracy of the estimation results, but it will lead to a large reduction in the sample size. To resolve this problem, we use all samples with left censoring and the same method (logit regression) to analysis. Table 13 shows the corresponding regression results, which show that the coefficient of fiscal subsidy is still negative and significant at the 1% level.

Table 13

Impact of fiscal subsidies on the probability of resurrection of zombie firms (modified sample)

(1) Coefficient (2) Odds ratio
Sub −3.9146*** 0.0199***
(0.760) (0.015)
Age 0.0002* 1.0002*
(0.000) (0.000)
Size 0.0218 1.0220
(0.027) (0.028)
K/L 0.2415*** 1.2732***
(0.020) (0.025)
Wage −0.1454*** 0.8647***
(0.022) (0.019)
SOE 0.0879 1.0919*
(0.052) (0.057)
Year fixed effect Yes Yes
Industry fixed effect Yes Yes
Region fixed effect Yes Yes
Observations 106,949 106,949

Note: robust standard errors are in parentheses; *** p < 0.01, * p < 0.1.

Table 9 shows that a firm may have several durations as zombie firms. In the benchmark regression in Table 11, we set all the durations as independent durations. However, correlation may exist among zombie durations of a firm in different periods, which will lead to the unsatisfied basic assumption of sample independence. To treat this problem, we will analyze the robustness of the results in this section from two ways. First, for firms with multiple zombie durations, we only introduce the first duration and eliminate the remaining durations. Second, we only introduce the duration of zombie firms with a single duration. Although these two treatment methods will eliminate a large number of samples, they can ensure that the durations are independent of each other. Table 14 shows the regression results considering only the first period of duration and single duration, which shows that the result is unchanged.

Table 14

Impact of fiscal subsidies on the probability of resurrection of zombie firms (based on modified duration)

Only with the first duration Only with single duration
(1) Coefficient (2) Odds ratio (3) Coefficient (4) Odds ratio
Sub −5.4265*** 0.0044*** −4.0488*** 0.0174***
(1.188) (0.005) (1.570) (0.027)
Age 0.0001 1.0001 0.0005* 1.0005*
(0.000) (0.000) (0.000) (0.0003)
Size −0.4133*** 0.6614*** −0.1115* 0.8945*
(0.043) (0.028) (0.059) (0.052)
K/L 0.0561** 1.0577** 0.2021*** 1.2239***
(0.029) (0.030) (0.040) (0.049)
Wage −0.2842*** 0.7526*** −0.2415*** 0.7855***
(0.030) (0.023) (0.046) (0.036)
SOE 0.2941*** 1.3419*** 0.0139 1.0140
(0.079) (0.106) (0.071) (0.072)
Year fixed effect Yes Yes Yes Yes
Industry fixed effect Yes Yes Yes Yes
Region fixed effect Yes Yes Yes Yes
Observations 52,521 52,521 37,712 37,712

Note: robust standard errors are in parentheses; *** p < 0.01, ** p < 0.05, and * p < 0.1.

Finally, we use PSM method used in Section 4 to match firms that have not received fiscal subsidies. We also use the matched samples to estimate the benchmark regression equation again. The result is exhibited in Table 15, which indicates that the inhibitory effect of fiscal subsidies on the “resurrection” of zombie firms is robust.

Table 15

Impact of fiscal subsidies on the probability of resurrection of zombie firms (based on matched samples)

(1) Coefficient (2) Odds ratio
Sub −2.7816*** 0.2700***
(0.736) (0.009)
Age 0.0777 0.9415
(0.081) (0.6170)
Size 0.0008*** 1.0005***
(0.002) (0.005)
K/L 0.1455*** 1.2590***
(0.030) (0.121)
Wage −0.2808*** 0.7931***
(0.044) (0.206)
SOE 0.0715 1.1439
(0.068) (0.8885)
Year fixed effect Yes Yes
Industry fixed effect Yes Yes
Region fixed effect Yes Yes
Observations 30,114 30,114

Note: robust standard errors are in parentheses; *** p < 0.01.

6 Summary

An important task for China to promote supply-side structural reforms is to handle zombie firms properly. In view of the negative impact of zombie firms, in the context of China’s current promotion of high-quality economic development, the disposal of zombie firms has become particularly meaningful. To handle zombie firms properly, in addition to restraining the formation of new zombie firms, accelerating the exit of zombie firms is necessary. The ways of cleaning up zombie firms include zombie firm exiting from market in the form of “death” and becoming normal firms in the form of “resurrection.” A key reason why government maintains the survival of zombie firms is that zombie firms bring jobs. From this point, the best approach to clean up zombie firms is to “resurrect” them and allow them to operate as normal firms.

The existing literature focuses on discussing how to suppress the formation of zombie firms; less attention is given on how to promote the exit of zombie firms. Based on the micro data of Chinese industrial firms from 1998 to 2007, we use different methods to identify zombie firms. Further, using the data of identified zombie firms, we measure the survival duration of zombie firms, and study the impact fiscal subsidies on the survival probability of zombie firms by survival model. The research shows the following findings:

First, from 1999 to 2007, zombie firms accounted for the highest proportion of all industrial firms in 1999, reaching 18.27%. The proportion gradually declined but increased again after 2005. Among these zombie firms, the proportion of zombie firms that received government subsidies showed an overall upward trend, reaching a peak in 2005, and then falling slightly, but it was also over 14%.

Second, for zombie firms who exit by “death,” their average survival period is 1.842 years, with a median value of 2 years, indicating that most zombie firms will exit the market within 2 years after they fall into zombification. The duration of subsidized zombie firms is longer than that of unsubsidized zombie firms. For zombie firms which exit by “resurrection,” nearly 35% of them have a single 1 year zombie status, indicating that the duration of zombie firms that can be “resurrected” is relatively short. In addition, the number of samples with multiple durations exceeds 50%, which indicates that a large number of firms can quickly leave the zombie status after falling into the zombie status. However, they also easily fall into the zombie status again.

Third, government-provided fiscal subsidies extend the duration of zombie firms. Specifically, fiscal subsidies not only reduce the probability of zombie firms withdrawing from the market, but more importantly, they reduce the risk of zombie firms turning into normal firms. Probability makes zombie firms “stiff but not dead” and “stiff but not alive.” After using different zombie firm identification methods and different estimation methods, and considering the endogenous nature of fiscal subsidies, this result remains robust.

The initial goal of the government when subsidizing low-efficiency zombie firms was to help firms overcome difficulties, turn losses into profits, and transform them into normal firms. However, the current subsidies provided by the Chinese government not only prevent zombie firms from exiting the market by way of “death,” but also prevent zombie firms from turning losses into profits and becoming normal firms.

Based on our results, we believe that, first, when governments provide subsidies to zombie firms, they must accurately assess the profitability and growth of zombie firms. The subsidy should be used cautiously for firms that have obviously lost vitality. For low-efficiency firms that have no chances of making profit, the government should cut off the channel of “blood transfusion” to zombie firms, so that they can be automatically eliminated in the market competition. For zombie firms with better technology and higher risk tolerance, strict performance standards should be formulated, and regular inspections should be made to prevent firms from accepting subsidies to simply survive the status quo without making technological changes. In this way we can not only improve the efficiency of the use of government subsidies but also effectively clean up zombie firms. Second, and more importantly, given that fiscal subsidies have indeed significantly inhibited zombie firms from exiting the market by way of “death” and from “resurrecting” into normal firms, therefore, we suggest that we should guide the market to give full play to the fundamental and decisive role of the market mechanism in dealing with zombie firms. Finally, government intervention should be reduced in the business activities of firms, especially zombie firms.

Acknowledgments

The authors gratefully acknowledge the financial support from the Fundamental Research Funds for the Central Universities of China (No: 2072021052) and the National Natural Science Foundation of China (No: 71973111).

  1. Conflict of interest: The authors have no conflict of interest to disclose.

  2. Article note: As part of the open assessment, reviews and the original submission are available as supplementary files on our website.

References

Agarwal, R., & Gort, M. (2002). Firm and product life cycles and firm survival. American Economic Review, 92(2), 184–190.10.1257/000282802320189221Search in Google Scholar

Ahearne, A. G., & Shinada, N. (2005). Zombie firms and economic stagnation in Japan. International Economics and Economic Policy, 2(4), 363–381.10.1007/s10368-005-0041-1Search in Google Scholar

Audretsch, D. B., & Mahmood, T. (1995). New firm survival: new results using a hazard function. Review of Economics and Statistics, 77(1), 97–103.10.2307/2109995Search in Google Scholar

Besede, T., & Prusa, T. J. (2006). Product differentiation and duration of US import trade. Journal of International Economics, 70(2), 339–358.10.1016/j.jinteco.2005.12.005Search in Google Scholar

Brandt, L., Van Biesebroeck, J., & Zhang, Y. (2012). Creative accounting or creative destruction? Firm-level Productivity Growth in Chinese Manufacturing, 97(2), 339–351.10.1016/j.jdeveco.2011.02.002Search in Google Scholar

Brown, C. O., & Dinc, I. S. (2011). Too many to fail? Evidence of regulatory forbearance when the banking sector is weak. Review of Finance Studies, 4, 1378–1405.10.1093/rfs/hhp039Search in Google Scholar

Caballero, R. J., Hoshi, T., & Kashyap, A. K. (2008). Zombie lending and depressed restructuring in Japan. American Economic Review, 98(5), 1943–1977.10.3386/w12129Search in Google Scholar

Cai, H., & Liu, Q. (2009). Competition and corporate tax avoidance: Evidence from Chinese industrial firms. Economic Journal, 119(537), 764–795.10.1111/j.1468-0297.2009.02217.xSearch in Google Scholar

Carreira, C., Teixeira, P., & Nieto-Carrillo, E. (2022). Recovery and exit of Zombie firms in Portugal. Small Business Economics, 59(2), 491–519.10.1007/s11187-021-00483-8Search in Google Scholar

Chang, Q., Zhou, Y., Liu, G., Wang, D., & Zhang, X. (2021). How does government intervention affect the formation of Zombie firms?. Economic Modelling. 94, 768–779.10.1016/j.econmod.2020.02.017Search in Google Scholar

Disney, R., Haskel, J., & Heden, Y. (2003). Restructuring and productivity growth in UK manufacturing. Economic Journal, 113(489), 666–694.10.1111/1468-0297.t01-1-00145Search in Google Scholar

Fukuda, S., & Nakamura, J. (2011). Why did “Zombie” firms recover in Japan?. World Economy, 34(7), 1124–1137.10.1111/j.1467-9701.2011.01368.xSearch in Google Scholar

Geng, Y., Liu, W., & Wu, Y. (2021). How do Zombie firms affect China’s industrial upgrading?. Economic Modeling, 97, 79–94.10.1016/j.econmod.2021.01.010Search in Google Scholar

Gouveia, A. F., & Osterhold, C. (2018). Fear the walking dead: Zombie firms, Spillovers and exit barriers. OECD productivity working papers 13.Search in Google Scholar

Hess, W., & Persson, M. (2012). The duration of trade revisited: Continuous-time versus discrete-time hazards. Empirical Economics, 43(3), 1083–1107.10.1007/s00181-011-0518-4Search in Google Scholar

Hoshi, T. (2006). Economics of the living dead. Japanese Economic Review, 57(1), 30–49.10.1111/j.1468-5876.2006.00354.xSearch in Google Scholar

Hoshi, T., & Kashyap, A. K. (2010). Will the U.S. bank recapitalization succeed? Eight lessons from Japan. Journal of Financial Economics, 97(3), 398–417.10.3386/w14401Search in Google Scholar

Jaskowski, M. (2015). Should Zombie lending always be prevented?. International Review of Economics & Finance, 40, 191–203.10.1016/j.iref.2015.02.023Search in Google Scholar

Jenkins, S. P. (1995). Easy estimation methods for discrete‐time duration models. Oxford Bulletin of Economics and Statistics, 57(1), 129–136.10.1111/j.1468-0084.1995.tb00031.xSearch in Google Scholar

Jiang, X., Li, S., & Song, X. (2017). The mystery of Zombie enterprises – “Stiff but deathless.” China Journal of Accounting Research, 10, 341–357.10.1016/j.cjar.2017.08.001Search in Google Scholar

Kane, E. J. (1987). Dangers of capital forbearance: The case of the FSLIC and “Zombie” S&Ls. Contemporary Economic Policy, 5(1), 77–83.10.1111/j.1465-7287.1987.tb00247.xSearch in Google Scholar

Kwon, H. U., Narita, F., & Narita, M. (2015). Resource reallocation and Zombie lending in Japan in the 1990s. Review of Economics, 18(4), 709–732.10.1016/j.red.2015.07.001Search in Google Scholar

Liu, G., Zhang, X., Zhang, W., & Wang, D. (2019). The impact of government subsidies on the capacity utilization of Zombie firms. Economic Modelling, 83, 51–64.10.1016/j.econmod.2019.09.034Search in Google Scholar

Lu, L., Li, X., & Qian, Z. (2020). Monetary policy, financial development and the financing of Zombie firms: Evidence from China. Economic and Political Studies, 8(2), 141–164.10.1080/20954816.2020.1730542Search in Google Scholar

McGowan, M. A., Andrews, D., & Millot, V. (2017). Insolvency regimes, Zombie firms and capital reallocation. OECD economics department working papers, No. 1399.Search in Google Scholar

Nakamura, J. (2017). Evolution and recovery of Zombie firms: Japan’s experience. In Japanese firms during the lost two decades. Springer Briefs in Economics. New York: Springer.10.1007/978-4-431-55918-4_2Search in Google Scholar

Nakamura, J., & Fukuda, S. (2013). What happened to “Zombie” firms in Japan? reexamination for the lost two decades. Global Journal of Economics, 2(2), 1–18.10.1142/S2251361213500079Search in Google Scholar

Nie, H., Jiang, T., Zhang, Y., & Fang, M. (2016). China’s Zombie firms: Cause, Consequence, and cure. Beijing: China Social Sciences Press (in Chinese).Search in Google Scholar

Nishimura, K. G., Nakajima, T., & Kiyota, K. (2005). Does the natural selection mechanism still work in severe recessions? examination of the japanese economy in the 1990s. Journal of Economic Behavior & Organization, 58(1), 53–78.10.1016/j.jebo.2004.03.008Search in Google Scholar

Nurmi, S., Vanhala, J., & Virén, M. (2020). The life and death of Zombies: Evidence from government subsidies to firms. Research Discussion Papers 8/2020. Bank of Finland.10.2139/ssrn.3601386Search in Google Scholar

Peek, J., & Rosengren, E. S. (2005). Unnatural selection: Perverse incentives and the misallocation of credit in Japan. American Economic Review, 95(4), 1144–1166.10.3386/w9643Search in Google Scholar

Robinson, J. A. (2009). Industrial policy and development: A political economy perspective. Paper prepared for the 2009 World Bank ABCDE Conference, Seoul, Korea, 22–24, 2009.Search in Google Scholar

Rodrik, D. (2008). Normalizing industrial policy. Commission on growth and development, Working Paper No. 3.Search in Google Scholar

Rosenbaum, P. R., & Rubin, D. B. (1985). Constructing a control group using multivariate matched sampling methods that incorporate the propensity score. American Statistician, 39(1), 33–38.10.1017/CBO9780511810725.019Search in Google Scholar

Sekine, T., Kobayashi, K., & Saita, Y. (2003). Forbearance lending: The case of Japanese firms. Monetary & Economic Studies, 21(2), 69–92.Search in Google Scholar

Shen, G., & Chen, B. (2017). Zombie firms and over-capacity in chinese manufacturing. China Economic Review, 44, 327–342.10.1016/j.chieco.2017.05.008Search in Google Scholar

Tett, G. (2004). Saving the sun: Japan’s financial crisis and a wall Stre. New York: Harper Collins.Search in Google Scholar

Willam, D. (2014). Zombie banks and forbearance lending: Causes, effects, and policy measures. Leipzig: University at Leipzig.Search in Google Scholar

Woo, W. T. (2019). China’s soft budget constraint on the demand-side undermines its supply-side structure reforms. China Economic Review, 57, 101111.10.1016/j.chieco.2017.09.010Search in Google Scholar

Received: 2021-12-05
Revised: 2022-07-19
Accepted: 2022-07-26
Published Online: 2022-08-26

© 2022 Ming Deng and Jinbo Wang, published by De Gruyter

This work is licensed under the Creative Commons Attribution 4.0 International License.

Downloaded on 4.2.2023 from https://www.degruyter.com/document/doi/10.1515/econ-2022-0029/html
Scroll Up Arrow