Inventory investment and the choice of financing in China: Does city-level financial development play a role?

Transcription

1 Inventory investment and the choice of financing in China: Does city-level financial development play a role? Junhong Yang * Management School, University of Sheffield, Conduit Road, Sheffield S10 1FL, United Kingdom Alessandra Guariglia Department of Economics, University of Birmingham, JG Smith Building, Edgbaston, Birmingham B15 2TT, United Kingdom Yuchao Peng School of Finance, Renmin University of China, Beijing , China Yukun Shi School of Business, University of Leicester, University Road, Leicester LE1 7RH, United Kingdom Preliminary Version: October 2016 Abstract We investigate the extent to which Chinese firms make use of different formal and informal channels to finance their inventory investment. To this end, using a panel of 224,604 firms over the period , we estimate error-correction inventory investment models augmented with bank loans and trade credit. We find that both sources of finance have a significant and positive impact on inventory investment. Furthermore, we observe that the impact of bank credit (trade credit) is greater in cities characterized by higher (lower) financial development. This suggests that in the presence of weak financial development, trade credit substitutes for formal bank loans. We further show that this substitution effect is more pronounced for private firms, firms located in southeastern regions, and financially constrained firms. Our results are robust to the use of a variety of specifications and estimation methods. JEL classifications: D92; E22; G3; O16 Keywords: Financing choice; Trade credit; Bank Loans; Inventories; Financial development; Financing constraints * Corresponding author. addresses: junhong.yang@sheffield.ac.uk (J. Yang), a.guariglia@bham.ac.uk (A. Guariglia), yukun.shi@leicester.ac.uk (Y. Shi), yuchaopeng@hotmail.com (Y. Peng) 1

2 1. Introduction It is widely accepted that inventories are frequently used to absorb economic shocks due to their low adjustment costs. For this reason, they are the most the volatile component of a country s GDP (Caglayan et al. 2012). A few studies have investigated the extent to which firm-level variations in inventory investment can be associated with changes in financial variables, such as cash flow. They found that the sensitivities of inventory investment to financial variables are generally high, especially if compared to those of fixed investment 1. A growing literature shows that financial development significantly affects firms decisions on how to finance their activities. In particular, well-developed financial markets reduce the costs of external finance faced by firms (Rajan & Zingales 1998; Fisman & Love 2003; Ge & Qiu 2007), making it easier for firms to finance their activities using bank loans or issuing shares. By contrast, informal sources of finance such as trade credit have been found to be prevalent in less developed financial markets or for firms facing tighter financial constraints (Fisman & Love 2003; Blasio 2005; Guariglia & Mateut 2006; Mateut et al. 2006; Ge & Qiu 2007; Cull et al. 2009). In this paper, we examine the role of (formal) bank lending and (informal) trade credit in financing Chinese firms inventory investment, differentiating cities on the basis of their financial development. Our aim is to assess whether the use of formal financing prevails in cities characterized by better financial development, whilst informal finance can be used as a substitute in those cities with poorer financial development. The Chinese setting provides an ideal laboratory to address these issues, given its rapid growth despite a malfunctioning financial system (Allen et al. 2005). Moreover, China s regional financial development is strongly unbalanced. Firms in regions with 1 Among these are Carpenter et al. (1994; 1998), Guariglia (1999, 2000), Guariglia and Mateut (2006), Guariglia and Schiantarelli (1998), and Daripa and Nilsen (2011) who try and explain inventory investment as a function of a range of financial variables such as cash flow, the coverage ratio, trade credit, the debt to total assets ratio, liquidity, and so on. 2

3 different level of financial development may experience different costs of financing. By constructing a unique set of city-level financial development indicators, we are able to investigate the extent to which trade credit can act as a substitute for bank credit in those cities where access to formal finance is difficult due to a poor level of financial development 2. The National Bureau of Statistics (NBS) of China provides us a sizable dataset, which enables us to test the extent to which Chinese manufacturing firms inventory investment is affected by the availability of formal and informal credit, differentiating firms according to ownership and location. Specifically, our dataset is a large panel of 224,604 Chinese mostly unlisted firms operating in 286 cities covering the entire Chinese territory, over the period Our paper contributes to existing literature in the following aspects. Firstly, building on Carpenter et al. (1994; 1998) and Guariglia and Mateut (2006), who look at the role of financial variables in determining inventory investment in the US and the UK, respectively, we analyse, for the first time, the role played by bank loans and trade credit in explaining Chinese firms inventory investment. Secondly, we extend Fisman & Love s (2003) country-industry level analysis by investigating, for the first time, the extent to which city-level financial development influences firms choice of financing. Third, our study provides micro economic evidence on the debate of the finance-growth nexus in China (e.g. Allen et al. 2005; Guariglia & Poncet, 2008; Zhang et al. 2012), focusing on inventory investment, which makes up a large part of GDP growth. Finally, we provide a powerful and comprehensive analysis on how the mix between trade credit and bank credit differs across firms with different characteristics such as ownership, locations, and financial conditions. We find that Chinese firms make use of both formal and informal credit to finance their investment in inventories. Introducing financial development indicators in our analysis, we then find that firms located in cities with a low level of financial 2 One of the key obstacles of conducting research on the relationship between bank credit and trade credit across countries is that different accounting standards and institutional settings characterize different countries, making them not strictly comparable. Because we focus on a single country, our study is not affected by this problem. 3

4 development tend to use trade credit to substitute for bank credit. This substitution effect appears to be more significant for privately-owned enterprises and firms located in southeastern regions. The substitution effect is also affected by the level of financial inclusion where firms operate. We further provide a more direct test on this substitution effect. Our results are robust to using different specifications and different estimation methods. The remainder of the paper is organized as follows. Section 2 summarizes the literature related to inventory investment, trade credit, and financial development. We develop our hypotheses in Section 3. Section 4 describes the specification of our models and the estimation methodology. Section 5 presents the dataset and the summary statistics. In Section 6, we discuss our main results, as well as the results of a variety of robustness tests. Section 7 concludes. 2. Related literature 2.1. Theories and evidence of inventory investment Firms inventory investment decisions have been theoretically and empirically studied as a major economic factor due to its importance role in explaining the business cycle (Metzler 1941; Carpenter et al. 1994; Benito 2005). There is some evidence that considers inventory investment to account for a dramatic amount of decline in output during the recession and to play a key role in the diffusion of recessions (Blinder & Maccini 1991; Carpenter et al. 1994). 3 At the micro level, inventory investment, as a stabilizing factor, is adjusted to smooth production in response to different shocks. It appears, however that inventories are considered as an inversely stabilizing factor of GNP since production is generally more volatile than sales (Blinder & Maccini 1991). 4 This could happen because firms tend to adjust inventories first in the presence of monetary policy shocks due to high liquidity and low adjustment costs of inventories compared to changing in employment and fixed investment. The study of the paradox 3 For example, 87% of the drop in GNP during the average US post-war recession can be explained by the decline in inventory investment (Blinder & Maccini 1991). 4 Sales are equal to GNP (production) minus inventories. 4

5 and firms inventory adjustments offer insights on the channels of monetary policy transmission. Previous literature has tested different hypotheses of financing channels based on inventories. Firstly, Kashyap et al. (1994) find that bank-dependent firms without access to public bond markets have to reduce their inventories following tight monetary policy compared to nonbank-dependent firms. The explanation supports the view of bank lending channel, in which monetary policy has an essential impact on formal finance available to bank-dependent borrowers due to legal reserve requirements on bank deposits. If the supply of bank loans decreases, the high external financial premium may force bank-dependent firms to decrease their investment especially on low-adjustment-costs inventories (Kashyap et al. 1996). Trade credit is an important source of short-term external finance that a supplier provides to its customer (Petersen & Rajan 1997). The aggregate value of trade credit (accounts payable) for US non-financial firms is three times of bank loans and 15 times of commercial papers (Barrot 2015). Nilson (2002), Mateut et al. (2006) and Love et al. (2007) theoretically prove that there is a trade credit channel in addition to the bank lending channel of transmitting monetary policy. Based on an error-correction inventory investment model, Guariglia & Mateut (2006) provide empirical evidence of the coexistence of the trade credit channel and the traditional credit channel in monetary policy transmission in the U.K. The former has a significant substitution effect on the latter, especially for firms with financial constraints. Using a theoretical model, Bougheas et al. (2009) explore the trade-off between inventories and trade credit. Similarly, Mateut et al. (2015) show that inventory composition influence the extent to which firms uptake trade credit. Yang and Birge (2013) find that trade credit can extend part of inventory risk to the suppliers, which is supposed be borne by the retailers Financing channels and the finance-growth nexus in China 5

6 The financial system in China remains mainly bank-based. The majority of Chinese banks, including the Big Five 5, are controlled by the government. The influence and intervention of the government play a significant role in banks decisions. For instance, the central bank explicitly sets primary deposit and lending interest rates and target levels for loan volumes. Furthermore, due to political reasons, in many circumstances, the government controls lending by pushing a large amount of loans to particular firms, sectors, and regions (Elliott & Yan 2013). The dominance of state-owned banks also causes a massive misallocation of financial resources in China, as these banks have a preferential policy of lending to stated-owned enterprises (SOEs), which crowds out the access to credit for SMEs (small- and medium-sized enterprises) and the private sector (Allen et al. 2005; Guariglia & Yang 2016b). Yet, China s underdeveloped and inefficient banking system hinders to some extent the fast progress of economic growth (Guariglia & Poncet 2008). Trade credit, which can provide funds through inter-firm transactions, plays an important role in China s rapid growth. Using a set of survey data, Ge & Qiu (2007) investigate how high growth of the non-state sector can be sustained by trade credit financing. They argue that high usage of trade credit helps non-soes bypass the limited access to formal banks and meet their financing needs. Further, according to Cull et al. (2009), SOEs with poor performance tend to redistribute bank loans via trade credit to prop up their faltering customers. On the contrary, they find a positive relation between trade credit and bank loans for profitable private firms, suggesting that profitable private firms are more likely to extend their trade credit. Degryse et al. (2016) find that the use of informal finance, including trade credit, can promote the high sales growth of small firms. In addition, a recent study of Guariglia & Mateut (2016) shows that Chinese firms with the strong or median level of political affiliations are easier to obtain a short-term external source of finance, and thus extend more trade credit than the non- 5 China s banking sector is dominated by the Big Five stated-owned commercial banks, which are the Bank of China (BOC), the People s Construction Bank of China (PCBC), the Agriculture Bank of China (ABC) the Industrial and Commercial Bank of China (ICBC), and Bank of Communications (BoCom). 6

7 affiliated counterparts. The evidence above suggests that trade credit is an important extension to the availability of the funds to Chinese firms. A large number of studies find that there is no consensus on the role of financial development in China s economic success. China seems to be a counter example to the traditional view on the finance-growth literature that financial development facilitates economic growth (Boyreau-Debray 2003; Allen et al. 2005; World Bank 2006; Guariglia & Poncet 2008). However, Cull & Xu (2005) show that access to bank loans is positively connected with China s profit reinvestment. Further research from Liang (2006) shows that financial development only positively affects economic growth in coastal areas, but not in interior regions. World Bank (2006) find that the capital market depth is positively associated with growth, but the impact of bank depth is sometimes negative. Ayyagari et al. (2010) find there is evidence that private firms benefit from utilizing bank loans. Formal finance through banks can facilitate Chinese firms efficient growth, which is confirmed by the positive relationship observed between bank financing and firms growth rates and reinvestment rates. At the city level, Zhang et al. (2012) find that the city s economic growth is positively correlated with its financial development in China. 3. Hypothesis development According to (Petersen & Rajan 1997), firms use trade credit as a source of finance mostly because they are unable to raise funds from the traditional bank finance channel. However, trade credit is typically more expensive than bank credit. This means that the growth of those firms which rely on this type of informal financing is constrained. Rajan & Zingales (1998) suggest that the development of financial markets can reduce the costs of formal external finance, and consequently enhance growth. Fisman & Love (2003) also emphasize the importance of financial development in explaining the substitution between bank credit and trade credit. They argue that firms in countries with less (more) developed financial markets rely more on trade credit (bank loans). In 7

8 line with their argument, and focusing on China, which contains very heterogeneous regions in terms of financial development, we hypothesize that: H1: Trade credit is likely to provide a substitute for formal bank loans in regions characterized by low financial development. In principle, firms can choose to finance their activities using either formal bank credit or informal trade credit. Yet, when they are financially constrained, they might not be able to obtain credit from formal financial institutions due to the cost premium to use external finance. These firms will, therefore, need to rely on trade credit. Suppliers may in fact still lend to these firms, as they are in a better position than banks to gather information on them. In line with this argument, and focusing on China, Guariglia and Mateut (2016) argue that more financially constrained firms indeed have a higher reliance on the trade credit financing(guariglia & Mateut 2006). This leads to our second hypothesis: H2: Firms with lower access to formal finance (e.g. private firms, firms located in southeastern regions, and financially constrained firms) will make more use of trade credit than bank credit. 4. Model specifications 4.1. Baseline specification: trade credit channel and bank lending channel The baseline specification in this paper is an extension of Lovell s stock adjustment model (1961), which has been widely used in the literature to explain the dynamic adjustment of inventory investment. 6 Specifically, denoting with I, the logarithms of firm s inventories; with S, the logarithms of sales; with Loans, the ratio of the sum of 6 The rationale behind the stock adjustment model is that when a firm s actual level of inventories is different from a desired target level that is proportional to sales, the firm will only try to adjust inventories partially towards the target level in any one period due to the adjustment costs and/or heterogeneous characteristic of stocks and/or the temporal uncertainty of ordered goods. Thus, the error-correction inventory investment model is less likely to be biased by the misspecification compared to those of fixed investment due to more appropriately accounting for investment opportunities (Guariglia & Mateut 2010). 8

9 long-term and short-term debt to total assets, which measures a firm s bank finance; 7 with TC (trade credit), the ratio of accounts payable to total assets, which measures a firm s use of trade credit, we initially estimate the Equation below: I j,t = α 0 + β 0 I j,t 1 + β 1 S j,t +β 2 S j,t 1 + β 3 (I j,t 1 S j,t 1 ) + β 4 Loans j,t + β 5 TC j,t + V j + V t + V k + V p + V o + e jt (1) where the subscript j indexes firms; k, industries; p, provinces; o, ownership; and t, time (where t = ). The lagged inventory growth and sale growth are included in the regression to capture short-run dynamics. The coefficients β 0 and β 2 are therefore expected to be positive. The error-correction term (I j,t 1 S j,t 1 ) captures the cost of inventories being far from a target level expressed in terms of sales. We expect the coefficient β 3 to be negative, consistent with error-correction behavior. In other words, future inventory investment will increase (drop) if inventories are lower (higher) than the target. Furthermore, we expect both β 4 and β 5 to be positive, showing the presence of a bank lending channel and a trade credit channel. 8 The error term in Eq. (1) consists of three components: V j, a firm-specific component, embracing all time-invariant firm characteristics likely to influence inventories and any enduring additive measurement errors; V t, a time-specific component, accounting for possible effects of business cycles, which we control for by including time dummies; V k and V P represent industry- and province- specific effects, respectively, which we control for by including both industry and province dummies. We also include ownership dummies to control for the heterogeneity in ownership structure in the Chinese context (Vo). Lastly, e jt is an idiosyncratic component. 9 7 We exclude accounts payables from Loans. 8 A large amount of literature has documented the impact of the financial variables (e.g. bank loans, trade credit, liquidity) on firms inventory investment, due to the low adjustment costs of inventories (Carpenter et al. 1994; Carpenter et al. 1998; Guariglia & Mateut 2006; Guariglia & Mateut 2010; Daripa & Nilsen 2011; Caglayan et al. 2012) 9 We find similar results using city dummies instead of province dummies. 9

10 4.2. Financial development and the choice of financing To shed light on the extent to which the level of financial development of the city where firms reside can influence their choice of financing, we introduce the interactions of bank credit and trade credit with City_FinDev, which denotes the level of financial development of the city where the firm is located. We also introduce the City_FinDev not interacted. This leads to the following augmented model: I j,t = α 0 + β 0 I j,t 1 + β 1 S j,t +β 2 S j,t 1 + β 3 (I j,t 1 S j,t 1 ) + β 4 Loans j,t + β 5 TC j,t + β 6 City_FinDev c,t + β 7 Loans j,t City_FinDev c,t + β 8 TC j,t City_FinDev j,t + V j + V t + V k + V p + V o + e jt (2) where the subscript c indexes cities. The coefficient associated with City_FinDev is expected to be positive, as firms in more financially developed cities will find it easier to fund their inventory investment. Furthermore, based on H1, we expect the coefficient on Loans*City_FinDev to be positive, while the coefficient on the interaction term TC*City_FinDev should be negative. The rationale is that financial development tends to reduce the costs of external finance to firms (Rajan & Zingales 1998), hence high financial development promotes the bank credit channel. However, for firms in cities with a low level of financial development, trade credit may provide an alternative source of funds (Fisman & Love 2003), suggesting low financial development encourages firms to use more trade credit. 5. Data and summary statistics 5.1. The data We utilize firm-level data drawn from the annual accounting reports of industrial firms conducted by the National Bureau of Statistics (NBS) of China from 2004 to Our sample starts from 2004 because the information of account payables in the NBS dataset is not available until The NBS data contains accounting variables and firm-specific information for enterprises in manufacturing and mining sectors with 10

11 annual sales above 5 million RMB. These firms come from 31 provincial-level administrative units, which can be further decomposed into 286 prefecture-level cities or municipalities. To minimize the potential influence of outliers, we firstly drop observations with negative values for inventories, sales, total assets, total fixed assets, bank loans, accounts payable, current assets, current liabilities, total equity, total assets minus total fixed assets, and total assets minus liquid assets. Secondly, we exclude the firms that did not have complete records on the key variables used for main regressions. Furthermore, the top and bottom one percent of the observations for each main regression variable are excluded. This leads to a final unbalanced panel made up of 224,604 mostly unlisted firms, which corresponds to 579,250 firm-year observation. 10 Table A1 in the Appendix shows that the unbalanced panel is ranging from a minimum of 47,146 observations in 2004 and a maximum of 130,088 firms in All variables are deflated using the gross domestic product (GDP) deflator, which is provided by Federal Reserve Bank of Atlanta. 11 We next merge the data with the city- and districtlevel financial development data, which is collected from the China City Statistical Yearbook. 12 [Insert Table 1 here] Numerous literature shows that the ownership structure has a significant impact on how Chinese firms make use of different funds (Allen et al. 2005; Ge & Qiu 2007; Ayyagari et al. 2010; Guariglia et al. 2011; Guariglia & Yang 2016b). 13 In our study, the ownership status of Chinese firms is classified into four categories: state-owned, 10 We cannot separate public listed firms from these unlisted ones as the NBS dataset does not have an identifier for public listed companies. There are around 1000 listed firms in manufacturing and mining industries covered by the NBS dataset, which accounts about 0.4% of the total observations. 11 We use this GDP deflator instead of the data from Chinese statistical and government agencies because it adjusts seasonality more appropriately and is comparable to those commonly used in the studies on OECD (Organisation for Economic Co-operation and Development) countries. For the details about how to construct this GDP deflator, see Chang et al. (2015) and Higgins & Zha (2015). 12 The yearbook provides the aggregate loans, deposits, savings and other financial related variables at the city level. 13 There is a large imbalance in the allocation of financial resources in China: although firms from the private sector have been expanding very fast and contribute most to China s growth, the majority of domestic bank credit goes to the less efficient state-owned sectors, hence depriving most of the private firms from the access to bank credit (Allen et al. 2005; Cull et al. 2009; Ayyagari et al. 2010; Guariglia et al. 2011; Guariglia & Yang 2016b). 11

12 foreign, collective, and private firms, on the basis of the majority share capital paid-in by each type of investors in each year. 14 For instance, a firm is categorized as stateowned in a given year if the proportion of its paid-in-capital owned by the state in that year is greater than 50% Measures of Financial Development To investigate the extent to which financial development affects the use of different financial sources, we construct a set of financial indicators to proxy the level of financial development in the locality where the firm resides. Typically, financial development should proxy the overall depth and availability of financial intermediaries and markets across regions. Thus, financial development should measure how easily borrowers and savers can be brought together. As in Zhang et al. (2012), we measure financial development at the city level 15, and construct the following three indicators. City_FinDev is the ratio of total loans in the city s financial system (from both banking and non-banking institutions) to the city s gross regional product (GRP). City_FinDev2 is the ratio of total deposits in the city s financial system to the city s gross regional product (GRP). These two indicators serve as proxies for the overall depth of financial intermediation. City_FinDev3 is the ratio of total household savings in the city s financial system to the city s gross regional 14 The NBS dataset classifies the investors of the firms into six categories: the state investors; foreign investors (excluding those from Hong Kong, Macao, and Taiwan); Hong Kong, Macao, and Taiwan investors; legal entities; individuals; and collective investors. As in Guariglia et al. (2011), based on their economic characteristics, we group foreign investors and Hong Kong, Macao, and Taiwan investors into a new category named foreign; legal entities owned and individuals owned firms are grouped as private firms. In addition, our results are robust to only considering firms owned by individuals as the private category. 15 There are three levels of cities in China: municipalities (Beijing, Shanghai, Tianjin and Chongqing), which are directly governed by the central government and are administratively equivalent to provinces; prefecture-level cities, which are directly governed by the provincial government and are ranked below provinces and above counties in China's administrative structure; and county-level cities, which are governed by the prefecture-level government. In this paper, to measure financial development, we use 286 municipality- or prefecture-level cities, including both urban and rural regions. We find similar results when we exclude the municipalities in our study. 12

13 product (GRP). It measures the city s financial development in terms of mobilizing household savings. Next, we measure financial development at the district level 16. To this end, we construct the indicator City_FinDev4, which denotes the ratio of total loans in the city s main district to the gross regional product (GRP). Furthermore, following Firth et al. (2011) and Guariglia & Yang (2016a), we also measure financial development (City_FinDev5) using the National Economic Research Institute (NERI) index of marketization (Fan et al. 2007) of the province or major municipality where a firm resides. 17 Finally, we design a composite index of financial development, City_FinDev6, which is the weighted geometric mean of City_FinDev, City_FinDev2, and City_FinDev3 18. Figures 1 and 2 show maps of the level of financial development measured as the ratio of total loans to GRP (City_FinDev) across different Chinese cities in 2004 and There are a total of 286 municipality- or prefecture-level cities in our maps. The figures suggest that there is a substantial imbalance in the level of financial development across the different regions of China, which can have a significant impact on how difficult firms find it to raise funds, as well as their choice of financing. Not surprisingly, coastal regions, major municipalities, and capitals of provinces enjoy the highest financial development. As a consequence of the recent financial crisis, the level of financial development in 2009 is slightly lower than that in Specifically, 5 out of 12 coastal provinces show a drop of their ratio of total loans to GRP between 2004 to The corresponding numbers for central and western regions are, respectively, 7 out of 8 and 7 out of 10. The imbalanced nature of financial development across different regions of China provides us with a unique opportunity to analyze whether 16 A district refers to a subdivision of a prefecture level city or a municipality in China. A district of a municipality is generally prefecture-level; and a district of a prefecture-level city is county-level. The main districts of a city are a densely populated area. 17 See Appendix A for the detailed definitions of our five measures of financial development. 18 The composite index of financial development is constructed based on factor analysis. We use the properties of our factor analysis to derive the weight of the each dimensional indicator. 13

14 informal finance can act as a substitute for bank finance in less financially developed areas. [Insert Figure 1 here] 5.3. Summary statistics Table 1 reports the means and medians for a number of key variables used in this study. Column 1 refers to the full sample covering all firms; columns 2-5 correspond to the sub-samples across four different ownership types, i.e. state-owned, collective, private and foreign firms. We observe that firms experienced positive inventory and sales growth in general; however, private firms are the major contributor to the growth while the other three types of firms experience low or negative rates of growth. With regards to different uses of funds, SOEs exhibit the highest bank loans to assets ratios (0.421) and the lowest trade credit to assets ratio (0.127), compared to the rest of the sample. This is consistent with Ding et al. (2013) and Guariglia et al. (2011). Due to the soft budget constraints from which they benefit, SOEs are able to obtain more bank credit than other firms, despite experiencing negative sales growth. Consequently, SOEs do not need to rely too much on informal finance such as trade credit. In addition, we observe that the size (total assets) of state-owned and foreign firms is generally larger than that of collective and private firms. Furthermore, state-owned and collective firms have a significantly longer history than private and foreign firms, because the latter were only allowed to start their business after China s reform and opening up. [Insert Table 1 here] In terms of financial development, Table 2 shows that the average ratios of total loans, total deposits and total household savings to GRP are 100.1%, 144.3%, and 67.6%, respectively, suggesting that the financial system in China remains mainly bank-based. 19 Next, our descriptive statistics show a significant imbalance regarding 19 According to Elliott & Yan (2013), the ratio of total bank credit to GDP reached 128% in 2012, which was much larger than the corresponding ratio in the US in the same year (48%). 14

15 financial development. Specifically, the majority of the firms (55.5%) in our sample locates in southeastern regions. In addition, southeastern regions enjoy the highest level of financial development. In contrast, central regions are the least financially developed, which is true regardless of what measures of financial development are used. [Insert Table 2 here] 6. Empirical results 6.1. Main results We estimate Eq. (1) and Eq. (2) for the whole sample using the fixed effects estimator, and the results are shown in Table 3. Fixed effects allow us to control for unobserved firm-specific heterogeneity. Column 1 reports the estimation results of the baseline model (Eq. (1)). The coefficients associated with the bank loans and the trade credit in column 1 are both positive and significant, which is the evidence in favour of the presence of bank lending and trade credit channels operating in the Chinese manufacturing sectors. The elasticities of inventory growth with respect to a change in bank loans and trade credit, evaluated at sample means are respectively 6.26 and This suggests that a 1 % increase in the bank loans and trade credit to total assets ratios leads to a 6.26% and 2.90% rise in inventory investment, respectively. Focusing on the other regressors, as expected, the significant and positive coefficients on the lagged dependent variable suggest that there is persistence in firms inventory investment. In addition, the current sales growth is positively and significantly related to inventory accumulation, whilst the lagged sales growth is significantly negatively related to it. The sum of the coefficients on the change in sales is positive, suggesting that the stock of inventories move together with sales growth, as there is a large cost of being out-of-stocks when firms face high demand for their goods. The coefficient on the error-correction terms is significant and negative as predicted by the theory, suggesting that inventories move towards their long-run targets and tend to 20 For example, this elasticity is defined as the ratio of the change in inventory growth for a relative change in the bank loans. Considering that the mean of inventory growth is and the mean of the bank loans is 0.374, it is given by 6.26=0.519*(0.374/0.031). 15

16 close the gap with their desired levels. [Insert Table 3 here] Columns 2-4 of Table 3 present the estimates which look at the role of financial development. Using the ratio of total loans to GRP (City_FinDev) as an indicator of financial development, in column 2, we include the level of financial development and the interaction of bank loans and financial development in the regression. It is found that a higher level of financial development significantly increases firms inventory investment. The interaction coefficient is positive and significant, implying that the bank lending channel prevails for firms in well financially developed cities. The reason could be that financial development reduces the firms costs of formal external finance (Rajan & Zingales 1998). In column 3, we instead include the interaction of trade credit and financial development. The coefficient on the interaction term is negative and significant, suggesting that firms in less financially developed cities are more likely to use trade credit to fund their inventory investment. Finally, in column 4, we include both the interactions of bank loans and trade credit with the financial development indicator at the same time. The coefficient on the former is still positive and significant, while the coefficient on the latter still displays a significantly negative coefficient. In line with our first hypothesis, these findings suggest that high level of financial development promotes the use bank credit channel, whilst a low level of financial development encourages firms to use more trade credit. In Table 4, we re-estimate Eq. (2) differentiating firms by ownership types. Specifically, we partition our firms into state-owned (column 1), collective (column 2), private (column 3), and foreign (column 4), according to the shares of paid-in-capital contributed by the four types of investors in each year. Similar coefficients as in Table 2 hold for the explanatory variables in the regressions 21, apart from that the interaction 21 We observe that the coefficients on both bank loans and trade credit positive and significant across firms with different ownership, however the coefficient on financial development is only significantly positive for private and foreign firms, which suggest these firms can benift from financial development in terms of inventory investment. 16

17 effects of financial development both with trade credit and bank loans, which are only significant for private firms. When the level of a city s financial development is high, private firms are more likely to make use of bank loans to accumulate inventories. By contrast, when a city s financial development level is low, discriminations in bank lending become severe, and private firms have to rely more on trade credit to accumulate inventories. This suggests that well financially developed markets enable private firms, which are more likely to face financial constraints, to use bank loans rather than trade credit. Informal credit finance can substitute for private firms access to bank loans in less financially developed regions. Our results are in line with the findings in Ge & Qiu (2007), who, using survey data argue that high usage of trade credit helps non-soes bypass the limited access to formal banks and meet their financing needs. For state-owned firms, the interaction terms are not significant. This suggests that the level of financial development has no impact on the choice of trade credit or bank loans to finance investment. SOEs have in fact the privilege to access funds from state banks, regardless of the level of financial development of the region where they operate, due to strong connections with local government and their role in maintaining social stability and keeping low unemployment rates. Similar results can be found for collective firms expect for the negative and significant coefficient on the interaction between loans and the financial development indicator. This may be due to the relatively small size of collective firms and to the fact that they are mainly located in rural areas, which are less financially developed. Collective firms are more likely to obtain bank loans due to the relative privilege in rural areas. For foreign firms, the interaction between financial development and trade credit has a significantly negative coefficient, whilst the interaction with bank loans is no longer significant. This finding indicates that a high level of financial development tends to reduce the use of trade credit, but not to promote bank lending for foreign firms. Our descriptive statistics reported in Table 1 showed that, compared to firms owned by other agents, foreign 17

18 firms have the highest liquidity ratio 22 and trade credit, and the lowest bank loans to assets ratio. This suggests that foreign firms prefer to use internal finance and trade credit rather than bank loans, even when the level of financial development is high. Moreover, they tend to keep a relatively low level of bank loans as they are able to obtain an offshore source of funds from their parent companies. [Insert Table 4 here] Next, as in World Bank s (2006) China investment climate report, we group the 31 Chinese provincial-level units into six regions: Bohai regions, central regions, northeastern regions, southeastern regions, southwestern regions and northwestern regions. 23 The rationale for this classification is that China not only has a very large territory, but its regional economy is also far beyond full integration. Geographical divergence in China may have significant implications on regional resource allocations and economic development. Based on the descriptive statistics in Table 2, we find that southeastern regions are the most financial sound. By contrast, central regions are the least financially developed. Thus, it is expected that regional variation may affect firms use of funds. [Insert Table 5 here] We re-estimate Eq. (2) separately for each of the above mentioned geographical locations. Columns 1 to 6 of Table 5 correspond to firms in the Bohai, central, northeast, southeast, southwest and northwest regions, respectively. In all types of specifications, the explanatory variables are always precisely determined. However, the coefficients on the interaction terms of the financial variables with financial development indicate that only in firms located in southeast regions does financial development affect both the use of bank loans and trade credit. Specifically, the use of bank loan is enhanced and the use of trade credit discouraged in cities characterized by more financial development. For firms in other regions, the coefficient associated with the interaction 22 The liquidity ratio for state-owned, collective, private and foreign firms are 0.089, 0.146, and 0.220, respectively. 23 For the distribution of these provincial level units and regions, see Appendix A, Chinese Provincial Units. 18

19 between bank loans and financial development is no longer significant, whilst the coefficient associated with the interaction between trade credit and financial development is negative and significant in the Southeastern and Southwestern regions, and marginally significant in the central and northwestern regions. For firms in Bohai and northeast regions, the insignificant coefficients of TC*City_FinDev suggest the level of financial development has no impact on and the use of trade credit. Our findings can be explained considering that firms operating in the most fastgrowing southeastern regions face high competition for a limited amount of funds 24. Therefore, they are likely to suffer from severe financial constraints and have to pursue alternative ways of finance, such as trade credit when they operate. In line with this argument, a World Bank s survey (2006) documents that firms in Southeastern cities, and especially SMEs indeed find it easier to meet their financial needs from informal sources of finance such as trade credit. Thus, the financial development of the city they operate in could help these firms gain more access to formal funds so that they can rely less on trade credit to accumulate inventories. On the contrary, the Chinese government has policies aiming at developing north, central and western regions, lowering the costs and increasing the availability of finance in these regions (Goodman 2004; Guariglia et al. 2011). 25,26 Thus, firms operating in these regions benefit from financial incentives and financial development is relatively unimportant to their choice of financing. In columns 7 and 8 of Table 5, we present the estimation results for firms 24 In our sample, there are around 56.3% of the total firms are operating in coastal regions. 25 During the transition period from a planned to an open market economy, southeastern regions enjoyed the fastest growth rate in China (more than 50% of the firms in our sample locate in southeastern regions). Southeast regions also benefited from the open-door policy and the coastal development strategy (the highest scores of financial development from Table 2), which notably increased inter-regional imbalances. In order to reduce the imbalances, after the late 1990s, regional development policies such as the western development strategy, the northeast revival strategy, and the rise of central China strategy have been implemented by the Chinese government in order to speed up the development of central and western regions and reduce regional imbalance. Through such regional coordination, a substantial amount of state funds has been invested in these areas especially in infrastructure, energy, and natural resources projects (Goodman 2004; Chen 2008). 26 The decrease in the use of trade credit with the development of local financial market cannot be offset by the increase use of bank finance in western regions may be because that the inventory growth of western firms have already heavily depended on bank finance, indicated by the highest coefficient of loans among the three regions. It might be difficult to increase the degree of dependency of inventory investment on bank loans. 19

20 operating in provinces with a high and low level of financial inclusion. 27 The results show the coefficient associated with the interaction effect between bank loans and the financial development indicator is only significantly positive for firms in high financial inclusion regions, while, the coefficient on the interaction between trade credit and financial development indicator is only significantly negative for firms in low financial inclusion regions. This is because that formal bank loans (informal trade credit) plays a relatively more important role in high (low) financial-inclusion areas Direct test on the substitution effect The substitution effect varies significantly with the level of financial development To provide a direct test of the substitution effect of trade credit on bank credit, we augment Eq. (1) by interacting trade credit with the dummy High_Loans, which is equal to 1 in a given year if a firm s loans is greater than the median value of the ratio calculated in the industry the firm belongs to in that year, and 0 otherwise. 28 The rationale is that because high loans are associated with high access to bank finance, we should expect a negative sign on the interaction term due to a lower need for informal finance, i.e. trade credit. The results are presented in column 1 of Table 6. The coefficient on the interaction term is negative and significant, implying that the trade credit channel is less important when firms have more bank loans, which can be interpreted as evidence in favor of the substitution hypothesis, on the ground that firms with a high level of loans are more likely to access external finance and therefore have a lower need to make use of trade credit. [Insert Table 6 here] To test whether the substitution effect is stronger in weak financially developed cities, in column 2, we add the additional interaction term 27 See Appendix B for the detailed definition of financial inclusion index. 28 We find similar result when we interact trade credit with leverage ratio. Due to the higher correlation between leverage ratio and the dummy High_Loans, we do not include both of them in the regressions. 20

21 (TC*High_Loans*Low_FD) 29. We observe that the coefficient on this interaction term is negative and statistically significant, suggesting that the substitution effect is more pronounced when the level of financial development is low. In columns 3-6 of Table 6, we present results for firms owned by different agents. We observe that the triple interactions remain significantly negative for private firms, while they are poorly determined for state-owned, collective and foreign enterprises. The findings are consistent with the results shown in Table 3. This implies that due to the privileged bank relationship, the substitution effect of trade credit for SOEs is not as important as for private firms The substitution effect varies significantly with the degree of financial constraints A large literature has shown that financially constrained firms are more likely to be influenced by financial variables (Fazzari et al. 1988; Guariglia & Mateut 2006). Previous literature has also shown that the offsetting effect of the trade credit channel is more significant for financial constrained firms (e.g. Guariglia & Mateut, 2006). Using conventional criteria, i.e. firms size and age (Gertler & Gilchrist 1994; Beck et al. 2005; Clementi & Hopenhayn 2006; Guariglia & Yang 2016b), we then augment our analysis in Eq. (1) by including financially constrained indicators. It is expected that small and young firms tend to face more serious financial constraints than large and mature ones. In China, legal protections for creditors is still poor and SME borrows sometimes might run away to avoid paying back their loans (World Bank 2006). Given the fact it is more difficult for financial institutions to collect the information from small and young firms, 30 these firms are prone to be affected by asymmetry information in financial markets, leading to higher financial premiums. We define a firm as facing more (less) financing constraints in a given year if its total assets/age falls in the bottom (top) half of the distribution of the total assets/age of 29 Low_FD is dummy variables, equal to 1 in a given year if the city s level of financial development (City_FinDev) where a firm is located lies in the bottom half of the distribution of City_FinDev of all cities in the same province, and 0 otherwise. 30 The financial statements and audits of SMEs are relatively vague. 21

22 all firms operating in the same industry in that year. 31 The estimation results are shown in Table 7. In columns 1 and 2, we measure the degree of financial constraints on the basis of the whole sample firms size and age, respectively. It is found that the coefficients on the interaction terms between trade credit and the High_Loans dummy are more pronounced for financially constrained firms. This finding is in line with Guariglia & Mateut (2006), who argue that the substitution effect of the bank lending channel by the trade credit channel should increase with the degree of financial constraints firms face. [Insert Table 7 here] Columns 3 to 10 repeat the same exercise of columns 1 and 2 for firms owned by different agents. In general, the triple interaction terms of the financially constrained indicators are negative and significant, regardless of what type of ownership. 32 In short, we find that financially constrained firms tend to make more use of trade credit when they have difficulties to access bank loans Robustness tests Using different measures of financial development In addition to financial development measure used in the main tables (total loans divided by the gross regional product (GRP) of a city), as in Zhang et al. (2012), we adopt two additional indicators at the city level to proxy for financial development. These are City_FinDev2, the ratio of total deposits to the city s GRP and City_FinDev3, the ratio of total household savings to the city s GRP. We also use City_FinDev4, which is defined as the ratio of total loans to the main district GRP in a city. In addition, following Firth et al. (2011) and Guariglia & Yang (2016a), we construct a fifth indicator of financial development, City_FinDev5, using the marketization index of 31 As a robust check, we also use 30% threshold to classify financially constrained firms. Besides, we also define a firm as facing a higher level of financing constraints in a given year if its total assets and age fall in the bottom 50% or 30% of the distribution of the total assets/age of all firms operating in the same industry as well as have the same ownership type in that year. The results are qualitatively the same and available upon request. 32 One exception is state-owned firms when firms age is used as a criterion. We do not observe significant differences between financially constrained and unconstrained firms for state-owned firms. 22

23 regional financial development for each Chinese province and major municipality, compiled by China s National Economic Research Institute (NERI) (Fan et al. 2007). Finally, using factor analysis, we construct a composite index of financial development, City_FinDev6 33, which is the weighted geometric mean of City_FinDev, City_FinDev2, and City_FinDev3. Table 8 presents the estimates of Equation (2) using these alternative financial development indicators. 34 Regardless of how we measure financial development, the estimates in column 1 suggest that for the full sample, the interaction terms of financial development indicators with both bank loans and trade credit are statistically significant at the 1% level. The positive coefficient on the former interaction and the negative coefficient on the latter are consistent with our prior findings that regional financial development has a significant impact on firms choice of between formal bank credit and informal trade credit. In particular, the substitution effect of trade credit is significantly greater in poorly financially developed regions. Also, all financial development indicators have a positive and significant direct impact on inventory growth. As for the other explanatory variables, the estimates are qualitatively the same as in Tables 2 and 3. [Insert Table 8 here] Columns 2-5 of Table 8 report corresponding results for firms owned by different agents. We observe significant coefficients on the interaction terms for private firms and part of foreign and collective firms. In contrast, all interaction terms are insignificant for SOEs 35. These results confirm once our hypotheses according to which the impact of financial development on the substitution effect of trade credit is more pronounced for private enterprises compared to SOEs. 33 We use factor analysis to identify the dimensions of financial inclusion and derive weights. 34 We do not report the full results for the sake of conserving space. 35 One exception is Loans*City _FinDe5 in column 2. 23

24 Alternative methods of addressing endogeneity Our main empirical regressions are estimated using the fixed effects estimator. A caveat to the estimator is that in a dynamic panel setting, the fixed effects estimator may suffer from endogeneity problems. To overcome the potential problems of endogeneity, we re-estimate Eq.(2) using the Generalized Method of Moments (GMM) estimator developed by Arellano & Bover (1995) and Blundell & Bond (1998). We treat all regressors including the interaction terms as being potentially endogenous. Levels of these variables lagged three times, and further are used as instruments in the first-differenced equations and first-differences of these same variables dated twice are used as additional instruments in the level equations. To ensure the credibility of our instruments and the specification of the models, we carry out the Hansen (J) test or the m(3) test. Table 9 presents the results of the system GMM estimation. Again, our results remain qualitatively unchanged. The substitution effects are significant for the whole sample and private and foreign firms, but not for state-owned and collective firms. Although the Hansen (J) test indicates some issues with the instruments and/or the specification 36, the m(3) test suggest that the instruments used are valid and our model is correctly specified. [Insert Table 9 here] Financial development is found to be endogenous in the finance-growth literature (Rajan & Zingales 1998; Becker 2007; Butler & Cornaggia 2011; Maskus et al. 2012). For instance, economic outcomes may have an impact on the demand for financial resources and instruments. In turn, our proxies for financial development (e.g. the volumes of total loans, total deposit, and total household saving) are likely to respond to both supply and demand for capital. Although it seems unlikely, in our case that 36 For the dynamic model, the J-tests might reject the null that the over-identifying restrictions are valid. This could be due to the fact that the presence of intra-cluster correlation or heteroskedasticity causes standard statistics to over-reject the null (Arellano & Bond 1991; Hall & Horowitz 1996; Hoxby & Paserman 1998). 24

25 inventory growth is causal to either demand or supply for financial resources, we still take a cautious instrumental variables (IV) approach to reduce the concerns about the endogeneity and reverse causality. To this end, following Becker (2007) and Butler & Cornaggia (2011), we use an alternative instrument for the regional supply of finance: the proportion of seniors (those aged 64 and older) in a given province-year. 37 The intuition of utilizing the demographics is that seniors tend to participate less in the labor force and consume less, while they hold more bank deposits compared to other age groups. Thus, a large proportion of seniors in a region will have a positive effect on local capital supply rather than the demand for business finance. 38 In addition, we use the number of bank branches per 1000 km squares in each province as another instrument as the number of bank branches is likely to be correlated with the level of financial development, but is unlikely to be correlated with inventories. We re-estimate Eq. (2) using both instruments for the financial development indicator. We also instrument TC*City_FinDev and Leverage*City_FinDev in a similar way. [Insert Table 10 here] Table 10 presents the results of the IV estimation. Column 1 shows that interaction terms of both bank loans and trade credit are significant and exhibit all expected signs for the full sample. After we partition our firms based on their ownership type, the coefficients on the interaction terms are only significant for the sub-group of private firms. To evaluate the validity of the instrument, following the work of Stock & Yogo (2005), we show that the first-stage regressions pass the F-tests, suggesting the instruments used in our estimations are not subject to the weak instrument. Other tests, such as the Cragg-Donald F statistic, the Anderson statistic, the Sargan statistic also suggest that our model is identified and/or our instruments are valid. In short, our results are robust to account for the potential endogeneity of financial development. 37 Data of the fraction of seniors is drawn from the National Bureau of Statistics (NBS). 38 We find similar results when we use only one instrument: the proportion of seniors. 25

26 Other robustness tests A potential criticism is that firm s true ownership class might changes during the sample period, which could be endogenous because of the firm s strategical act to take advantage of or to avoid certain policies that affect the particular ownership class. In order to tackle this issue, we use the ownership classification made on the basis of ownership shares immediately before the sample period begins. We also define a firm s ownership based on the average shares of capital paid-in by our four types of investors during the sample period to minimize the endogenous nature of the ownership structure. Again, we find similar results to those of our baseline regressions. 39 Furthermore, it has been documented that there has been a high rate of entry and exit of firms during the sample period (Brandt et al. 2012), which might have a significant impact on how inventory investment is influenced by different financing channels in the Chinese manufacturing sectors. In order to control for the dynamic setting and to check if our results are driven by the massive entry of new firms, we reestimated our models based on a balanced sample, which allows us to focus on firms, which have been present throughout. We obtained qualitatively the same results to those reported in the paper. These results are not reported for brevity, but are available upon request. In unreported results, following Caglayan et al. (2012) and Chen & Guariglia (2013), we also include Liquidity in our baseline model to control for the role of internal finance. We find similar results to those of our baseline regressions and a negative coefficient associated with Liquidity as firms might increase their stocks of inventories by reducing their liquidity. In summary, our main results are robust to a variety of alternative variables and tests and the results in this section provide concrete evidence to our hypotheses. 39 The NBS dataset does not provide the information about firms ownership classification in year 2008 and To overcome this problem, in our study, we assume the ownership classes of firms do not change after The additional tests here also address the concern about the missing a part of F 26

27 7. Conclusion This paper presents concrete evidence on how financial development affects different sources of funds provided by bank loans and trade credit for Chinese firms by examining a panel of 224,604 Chinese firms from 286 cities over the period Specifically, firstly, we employ a set of unique city-level financial development data to test the impact of financial development on the substitution effect of trade credit on bank loans. We further split the firms based on different types of ownerships and the locations where they reside. In addition, we provide a direct test of the substitution effect across firm-years making low and high use of bank loans. Finally, we also test the substitution effect between firms facing high and low levels of financial constraints. Our results suggest that there exist the bank lending channel and the trade credit channel operating in China, which have significant impacts on inventory investments. We find that financial development promotes firms inventory investment, which is in line with Zhang et al. (2012) that the level of financial development is positively related to the city s economic growth. We also find that the substitution effects of trade credit on bank loans are more significant for firms in cities with weak financial development. The impact of financial development on the substitution effects is more pronounced for firms with lower access to formal finance, i.e. private firms, firms in southeast regions, and financially constrained firms. This adds city-firm-level evidence on Fisman & Love (2003) s theory about the relation between financial development and trade credit, which is based on the country-industry-level analysis. Our results are robust to a variety of controls, different measures of financial development, financial constraints, and ownership classification, as well as alternative estimation methods, i.e. GMM and IV estimations. Our study offers new insight into the finance-growth relationship in emerging markets. Our findings provide a portrait of the choice of financing used by different types of Chinese firms. This enables us to understand the roles of different channels of transmission of monetary policy operating in China. Both banking lending channel and trade credit channel should be considered by China s central bank when designing 27

28 monetary policy. In addition, the results also provide micro economic evidence on the relationship between financial development and economic growth in China. The importance of informal finance i.e. trade credit to private firms, firms from southeast regions, small and young firms, suggests that under-developed and inefficient financial markets might be an obstacle to restrict the fast growth of these firms. If these firms have difficulties in extending the trade credit, China s fast growth cannot be sustained. Given that private firms, firms in southeast region, SMEs constitute the engine of growth of the Chinese economy, the policy makers should think about to create more supportive legal and regulatory system to promote more both formal and informal sources of funds to these firms. A more effective financial system and better allocation of resources would therefore benefit the economy. 28

29 Figure 1: City-level financial development in China Note: This figure presents the city-level financial development in China in 2004 and The level of financial development of a city is measured by the ratios of total loans in the city s financial system (both banking and non-banking institutions) to the city s gross regional product (GRP). 29