Can Financial Frictions Explain China s Current Account Puzzle: A Firm Level Analysis (Preliminary)

Can Financial Frictions Explain China s Current Account Puzzle: A Firm Level Analysis (Preliminary) Yan Bai University of Rochester NBER Dan Lu University of Rochester Xu Tian University of Rochester February 16, 2015 Abstract China has been running current account surplus and accumulating large foreign exchange reserve despite the high growth rate. One potential explanation in the literature is the domestic financial frictions. This paper uses Chinese firm level data to quantify the financial frictions in China and asks to what extent China s accumulation of foreign reserve can be explained by these frictions. Keywords: China foreign reserve, current account surplus, financial frictions, firms debt financing JEL classification: E-mails: yan.bai@rochester.edu; danlu@rochester.edu; xu.tian@rochester.edu

1 Introduction China has been growing fast for decades. Despite the high growth rate, China runs current account surplus and accumulates a large foreign exchange reserves. This is puzzling under the standard economic theories, which predict that capital should flow into countries with a fast growing productivity and thus high returns. One explanation in the literature for this puzzle is the underdeveloped financial markets and distorted financial allocations in China 1. State-owned firms in China receive favorable lending terms from state-own banks, while privately-owned firms, which are more productive than the state-owned firms, rely heavily on internal financing. In spite of large literature on the effect of financial frictions on current account surplus and foreign reserve, few works use micro-level Chinese data to quantify these frictions. This paper fills the gap. We ask to what extent financial frictions identified with firm-level data can explain the current account surplus and large foreign reserves in China. We first document features of Chinese firm level data in terms of firm dynamics and financing. In particular, we are interested in how debt financing and growth vary with stateowned (SOE) and non-state-owned firms. Compared to the state-owned firms, non-stateowned firms have a significantly lower leverage, pay higher interest rate, and grow faster. Among non-soes, smaller firms have a lower leverage, a higher interest rate, and a much higher growth rate than large firms. Despite that China has been going through significant changes like SOE reforms and financial open up, these patterns of debt financing and growth are consistently observed over time. To identify the magnitude of financial frictions in China, we build a heterogenous-firm model with two sectors: a sector with state-owned firms and a sector with non-state-owned firms. All firms produce with a decreasing-return-to-scale technology and face stochastic productivity shocks. They finance their investment and dividend payouts with their own profits and loans from banks. Non-SOEs may default on their loans, while SOEs never default as long as they have the ability to repay. Banks provide debt schedules taking into account default risks of firms and fixed cost of issuing loans. Firm size distribution, leverage, and growth in the model are dictated by financial frictions: default risk and fixed cost of issuing loans, which are reflected in the endogenous debt schedule. On the one hand, borrowings from banks involve fixed cost. Small borrowings thus have high effective interest rate. On the other hand, large borrowing is associated with high default risk and thus also faces high effective interest rate. In equilibrium, small firms 1 Many papers have addressed this puzzle, for example, Buera and Shin (2009), Song, Storesletten, and Zilibotti (2011), Caballero, Farhi, and Gourinchas (2008), and Quadrini, Mendoza, and Rios-Rull (2009). 2

with low assets tend to be more financially constrained and end up having low leverage and inefficient size. When hit by a good productivity shock, these small firms grow faster from their inefficient size. Compared to SOEs, non-soes tend to default more and thus face higher borrowing rate, have lower leverage, and are more inefficient. Non-SOEs also grow faster. These predictions are consistent with the data. In aggregate, existence of financial frictions could contribute to current account surplus, and lead to accumulation of foreign reserves. In contrast with state-owned firms who have access to credit markets, non-soes are more financially constrained and finance investment mainly through internal savings. Under capital control over private foreign savings as in China, excess savings from household in state-owned banks are not invested in non-soes, and become foreign reserves as a result. Overtime, as the state-own sector shrinks, more domestic savings are invested in foreign reserves. To quantify the effects of domestic financial frictions on the net savings, we calibrate the friction parameters in our model to match the firm level features, and calculate the net savings generated by the financial friction in our model. Although financial frictions may explain the level of large foreign surplus, and the shrink of state-own sector may explain the early increasing of foreign surplus, it is still puzzling that we observed an ever growing foreign surplus. After year 2002, there are dramatic increases in foreign direct investment (FDI), and we would expect that firms are less financially constraint. Combining with the facts that SOEs share is relative stable later on, it is not straightforward that financial friction would be able to account for the fast growing foreign surplus. We include foreign direct investment in the model, and calibrate to the data patterns over time to address this issue. Our paper is related to the literature on the international capital flow puzzle, and more specifically on Chinese foreign surplus puzzle. Buera and Shin (2009) uses underdeveloped domestic financial markets to explain the joint dynamics of total factor productivity (TFP) and capital flows. When an economic reform eliminates pre-existing financial distortions, the TFP of a small open economy rises with more efficient reallocation of economic resources. At the same time, because of the financial frictions, saving rates surge, but investment rates respond with a lag, resulting in capital outflows. Using a growth model, Song, Storesletten, and Zilibotti (2011) shows that during economic transition high-productivity firms (non-state own firms) outgrow low-productivity firms (state own firms), if entrepreneurs have enough high savings. At the same time, the more financially integrated SOE sector shrinks and forces domestic savings to be invested abroad, which generates a foreign surplus. Wang, Wen, and Xu (2012) uses two types of capital, financial capital and fixed capital, to explain the two-way 3

flows of capitals. In their model, underdeveloped financial market in China lead to high rate of returns to fixed capital but low rate of returns to financial capital relative to the U.S.. As a result, households save abroad and FDI flows in. Our paper uses the debt financing features observed in the Chinese firm-level data to identify financial frictions, and quantifies the magnitude of net savings that can be generated by these frictions. Our paper also relates to the literature on the effect of misallocation from financial frictions on aggregate TFP. Hsieh and Klenow (2009) uses firm level data to quantify the potential extent of misallocation in China and India versus the United States. They document sizable gaps in marginal products of labor and capital across plants in China and India compared with the United States. Our paper further identifies the misallocation in the Chinese firm-level data that due to domestic financial frictions, by examining firms debt financing features and growth. Midrigan and Xu (2013) parameterizes a financial friction model to match salient features of plant-level data, and show that the model didn t predict large aggregate TFP losses from misallocation, and misallocation from financial constraint cannot explain TFP gap between countries with little external finance and the U.S. Arellano, Bai, and Zhang (2012) uses cross-country variation in financial market development to evaluate empirically and quantitatively the impact of financial frictions on firms financing choices and growth rates with firm-level datasets. Our goal is to identify the domestic financial frictions in China and quantify its effects on China s foreign surplus. The rest of the paper is organized as follows. Section 2 presents empirical findings on the features in Chinese firm level data in terms of debt financing, interest rate, and growth. Section 3 introduces the model. Section 4 presents the quantitative analysis. Section 5 concludes. 2 Data Empirical findings in the paper are based on a rich firm-level dataset of annual census of manufacturing enterprises by Chinese National Bureau of Statistics from 1998 to 2007. The dataset includes all SOEs and non-soes with sales over 5 million RMB (about 600,000 US dollars). It contains all information in balance sheet, profit and loss statement, and cash flow statement, which include more than 100 financial variables. This section describes overall patterns of firms asset, leverage, interest rate and growth, and compares patterns between SOEs and non-soes. Firm asset is measured by the book value of the firms total asset. To measure a firm s debt financing, we use both leverage and 4

interest rate. Firm leverage is defined as the ratio of total debt and total asset, and total debt includes short-term and long-term debt as well as short-term credit from suppliers. Firm interest rate is defined as the ratio of interest payment and total debt. Firm growth is measured by the growth rate of value added. We restrict our sample to the firms with positive assets, non-negative total debt and positive sales, yielding 149675 firms in 1998, and 251018 firms in 2005. We use the registration type information to classify SOE and non-soe. 2 In our sample, 32.5 percent of the firms in 1998, and 5 percent in 2005 are SOEs. Table 1 reports descriptive statistics, the mean and median level of assets, value added, leverage, interest rate and growth rate for firms in 2005. 3 Table 1: SOE vs Non-SOE, Year 2005 Overall SOE Non-SOE Average Mean Median Mean Median Assets 37220 70022 21399 35826 12959 Value Added 12119 12442 3632 12106 5127 Leverage 0.58 0.77 0.72 0.56 0.58 Interest Rate 0.023 0.012 0.002 0.024 0.008 Growth Rate 0.21 0.10 0.08 0.21 0.18 SOEs have more assets than non-soes on average. Both asset distributions are highly skewed, as the mean asset levels are much larger than the median level, and the distribution is more dispersed for SOEs than non-soes. Although SOEs have much more assets, the mean value added is similar to the one of non-soes. In terms of debt financing, SOEs have higher leverage than non-soes, and they pay much lower interest rate than non-soes. Non-SOEs grow much faster than SOEs. Does firms in different size finance their project differently? Figure 1 plots the mean leverage for SOEs and non-soes under different asset levels. The x-axes in the figure is the asset percentiles for SOEs and non-soes, and on the y-axes, we plot the mean leverage for firms between each percentile. The plot for SOEs is noisy due to fewer observations in each asset percentile. Clearly, among non-soes, leverage increases with firms asset. 2 SOE includes those with ownership codes 110, 141, 143, 149, 151. In the appendix, we show results with the least restrictive definition of SOE, which includes all the firms with positive state asset as well as collective enterprises. 3 We drop the top 1 percent of firms for each variable to exclude outliers. Assets and value added are in terms of thousand RMB. 5

leverage.4.6.8 1 1.2 SOE POE 6 8 10 12 14 lnasset Figure 1: Leverage vs Asset, by Sector, Year 2005 To study the relation between firm asset and leverage systematically, we run a regression of leverage on firms asset level and interaction of ownership and firms asset, controlling for industry fixed effects. Table 2 reports the regression results. The first column of the table shows that leverage ratio are significantly higher for SOEs. Among non-soes, small firms have smaller leverage ratio, while among SOEs, larger firms have smaller leverage ratio. We also run a similar regression for interest rate, see the second column of Table 2. Interest rates are significantly higher for non-soes than SOEs. Among non-soes, small firms paid high interest rate, while among SOEs, large firms have higher interest rate. Table 2: Regression of Leverage and Interest Rate on Firms Asset, Year 2005 Regressions Leverage Interest Rate lnasset.0069*** -.0018*** (13.02 ) (-23.63 ) SOE.384*** -.0458*** (17.75) (-15.18 ) SOE*lnasset -.017***.003*** ( -8.25) ( 11.43) Observations 218,798 218,798 Industry FE Yes Yes Robust t-statistics in parentheses *** p<0.01, ** p<0.05, * p<0.1 How does firms growth rate relate to their asset level? Figure 2 plots the mean growth rate for SOEs and non-soes with different asset levels. As is well documented in the literature, 6

Growth Rate.2 0.2.4.6 SOE POE 6 8 10 12 14 lnasset Figure 2: Growth Rate vs Asset, by Sector, Year 2005 small firms grow faster. Moreover, non-soes grow faster on average than SOEs. In summary, SOEs are more dispersed in terms of assets and value added, they have a significantly higher leverage ratio, pay much lower interest rate, and their growth rates are much lower. Among non-soes, smaller firms have a lower leverage, pay higher interest rate, and have a much higher growth rate. Although there are significant changes over time, for example dramatic shrink of SOEs from 1998 to 2005 due to the SOE reform in China, the patterns we described above are consistently observed. These patterns are consistent with theories of financial frictions, specially the credit restriction in our model. In the model, non-soe can default, and banks provide debt schedules taking into account firms default risk. On average, the interest spread is higher for non-soe. The debt schedules provided affect non-soes decisions of financing and investment: borrowing more is helpful financing investment, but would also lead to higher default risk. And this leads to an inefficiently small firms size. Because getting loans are costly, and small loans are more costly due to the fixed issuing cost, smaller firms are more financially constrained and end up having lower leverage ratio. Small firms are more likely to be inefficient in scale because they face more strict loan schedules (higher spread rate), then they would grow faster when they get good shocks as they can use the output to increase their scale. The financial friction can generate a trade surplus. Due to financial imperfections non- SOE must finance investments through internal savings and state-owned firms have better access to credit markets. As the state-own sector shrinks, excess savings are invested in terms of foreign reserves. Then the question is, to what extend the financial friction explain the large international reserve of China? We calibrate the model to match the patterns of the 7

firm level data, and using those parameters we predict the international reserve level that caused by the financial frictions. The moments we use are the moments that characterize the plant-level facts, such as leverage, interest spread, growth rate differences in Table 1, relation between leverage and asset by sectors in Table 3, size distribution by sectors in Table 4, and the distribution of growth rate, persistence and volatility of firms sales. Table 3: Leverage ratio by asset quantiles Leverage Ratio Asset (%) SOE non-soe 0 to 10.81.54 10 to 20.75.56 20 to 30.75.56 30 to 40.75.57 40 to 50.77.57 50 to 60.82.56 60 to 70.75.57 70 to 80.78.57 80 to 90.80.57 90 to 100.74.58 Table 4: Size(VA) Distribution, Year 2005 Fraction of VA Produced TOP Percentiles SOE non-soe 5 0.76 0.53 10 0.86 0.66 20 0.93 0.78 3 Model We consider a small open economy with two sectors, one with a continuum of state-owned firms (SOE), and one with a continuum of privately owned firms (POE). The measure of state-owned firms is µ and that of privately owned firms is 1 µ. Financial markets are imperfect in that firms can only borrow state-uncontingent bond. SOEs are not allowed to default as long as they are able to repay their debts. POEs, however, can default. Banks offer firm-specific debt contracts that compensate for default risk and fixed cost of lending. 8

Firms produce with a decreasing return to scale technology using capital as input, y = zk α where z is the stochastic productivity following a Markov process given by f(z ; z). Firms use its internal return or external borrowings to finance their investment. A firm is labelled with its state variable (z, k, b) with b denoting its initial debt holding. If b > 0, the leverage of the firm is given by b/k, otherwise the leverage is zero. We assume that firms face limited liabilities. Upon observing its productivity shock, a firm with (z, k, b) in the POE sector decides whether to default by comparing the default value V d with the repayment value V c, V (z, k, b) = max d {0,1} (1 d) V c (z, k, b) + dv d (z, k). Let the optimal decision rule on default be d(z, k, b) with d(z, k, b) = 1 denoting default. If repays, the firm chooses new investment and borrowings. In particular, it makes decision on next period s capital k, dividend x, and a loan b with price q(z, k, b ) incorporating the default probability. The repaying value is given by V c (z, k, b) = max {x,k,b } x + βev (z, k, b ) st x = zk α + (1 δ)k b + q(z, k, b )b k φ(k, k ) 0 where φ(k, k ) is the capital adjustment cost. If it defaults, the firm gets its debt written off but it will be penalized with some sanction and will be excluded from the credit market for a random period. After default, the firm chooses dividend and new investment to maximizes its value, V d (z, k) = max x,k x + βe [ (1 λ) V d (z, k ) + λv c (z, k, 0)] ] st x = γzk α + (1 δ)k k φ(k, k ) 0 where λ is the probability that the firm regains access to the credit market in the next period and γ is the fraction of productivity losses after default. Banks and SOE firms have good relationship and thus SOE firms never default as long as it is feasible for them to repay. SOE firms therefore face a natural borrowing limits B(z, k ), which guarantees that tomorrow even under the worst scenario firms are still able to repay 9

with the most external resources they can receive. Let the value of SOE firms be W (z, k, B). We define the problem as, W (z, k, b) = max x,k,b x + βew (z, k, b ) st x = zk α + (1 δ)k b + 1 1 + r b k φ(k, k ) 0, b B(z, k ), where B(z, k ) is recursively defined as B(z, k ) = { { }} 1 min z k α + (1 δ)k + max z f(z ;z)>0 k 1 + r B(z, k ) k φ(k, k ). Banks are competitive and risk neutral. They have to pay a fixed credit cost ξ for every loan they offer. The fixed cost captures banks overhead cost and also the cost for obtaining information for each loan. Given that SOEs never default, we assume their borrowing rate is the risk free rate, q(z, k, b ) = 1 1 + r. It is easy to see that with fixed cost, the effective interest rate of large borrowings is lower. The bond price schedule for POEs incorporates both the fixed cost and the future default probabilities, [ q(z, k, b )b + ξ = b 1 1 + r d (z, k, b ) f (z ; z) dz ]. For savings b 0, banks charge bond prices reflect the risk free rate, q = 1/(1 + r). Definition: A recursive equilibrium consists of decision rules and value functions of firms, and bond price schedule q(z, k, b) such that 1. Given the bond price schedule, the decision rules and the value functions solve each firm s problem. 2. Given interest rate and the decision rules, the bond price schedule makes banks break even in expected value. 10

4 Quantitative Analysis In this section, we present the quantitative analysis over the model. We choose parameters to generate firm distribution in China in 1998. We assume that firm s productivity has two components: a permanent component a and an idiosyncratic component ν. In particular, z it = (1 + g) t a i + ν it where g is the aggregate productivity growth in the economy. Following Midrigan and Xu (2012), we assume that the permanent component follows a Pareto distribution with a upper bound A and a shape parameter µ, i.e. P r(exp(a i ) x) = 1 x µ 1 A µ. The idiosyncratic component follows an AR(1) process, log(ν it ) = ρ log(ν it 1 ) + σε it, ε it N(0, 1). Let the capital adjustment cost be ( ) k 2 (1 δ)k. Φ(k, k ) = k φ 2 k The annual risk free rate r is chosen to match the deposit rate. The capital depreciation rate δ is chosen to be 10% annually. The capital adjustment cost φ is chosen to match the relative volatility of investment to GDP. The other parameters are chosen jointly β, ξ, δ, λ, γ, φ, ρ, σ, ν, A to match the following moments in the data: mean leverage, mean interest rate of POEs, average year of returning to market after default, persistence and volatility of firms sales, IQR of leverage, and distribution of value added. Our model predictions are consistent with the data observation in many dimensions. In both the model and the data, privately owned firms face a higher borrowing rate than stateowned firms. The leverage of state-owned firms is, however, higher than that of privately owned firms. After the reform and financial open up, the leverage of both types of firms decrease. We illustrate the model implications by presenting some quantitative simulations in this section. Figure 3 presents bond price schedules. The figure shows that small and large loans are the most expensive. Small loans have large effective interest rates due to the fixed cost of 11

Bond Price Schedule 1 0.8 0.6 0.4 0.2 0 0.5 0 0.5 1 1.5 Next period bond holding Figure 3: Bond Price Schedule vs Debt, Model Simulation issuing loans; large loans are expensive because of the higher default risk. For very small value of debt holding, the fixed cost effect dominates, and we observe a jump of interest spread. And for large values of debt holding, firms start to default with some probability, the interest rate increases, and for even larger value of debt holding, the firms default with probability 1. Figure 4 presents the mean leverage for SOEs and non-soes with different asset levels implied by the model. SOEs have much higher leverage ratio, and their leverage decrease with firms asset level, while POEs leverage increases with firms asset. Figure 5 presents the mean interest rate for SOEs and non-soes with different asset levels implied by the model. SOEs have much lower interest rate, and in this example, they all borrow at the risk free rate. Among POEs, the interest rate spread decreases with firms asset. Figure 6 presents the mean growth rate for SOEs and non-soes with different asset levels implied by the model. SOEs have much lower growth rate, the growth rate decreases with firms asset. 12

4 3 2 1 0 1 2 0.7 Leverage across different asset 0.6 0.5 Leverage 0.4 0.3 0.2 0.1 0 Asset Figure 4: Leverage vs Asset, by Sector, Model Simulation 4.1 Estimation Procedure (TBA) 4.1.1 Least square Our estimation is based on the following moment condition E [Γ (Θ 0 )] = 0 where Θ 0 is the true value of the parameters. With an identity weighting matrix, we find Θ minimizes Γ (Θ 0 ) Γ (Θ 0 ) 4.1.2 Optimal Weighting With an optimal weighting matrix, (compute the (general) inverse of the variance-covariance matrix of the data moments, or use continuous-updating estimator of the weighting matrix)... 13

4 3 2 1 0 1 2 0.35 Interest rate across different asset 0.3 0.25 Interest Rate 0.2 0.15 0.1 0.05 0 Asset 5 Conclusion To be added. Figure 5: Interest Rate vs Asset, by Sector, Model Simulation References Arellano, C., Y. Bai, and J. Zhang (2012). Firm dynamics and financial development. Journal of Monetary Economics 59 (6), 533 549. Buera, F. J. and Y. Shin (2009). Productivity growth and capital flows: The dynamics of reforms. Technical report, National Bureau of Economic Research. Caballero, R. J., E. Farhi, and P.-O. Gourinchas (2008). An equilibrium model of global imbalances and low interest rates. American Economic Review 98 (1), 358 393. Hsieh, C.-T. and P. J. Klenow (2009). Misallocation and manufacturing tfp in china and india. The Quarterly Journal of Economics 124 (4), 1403 1448. Midrigan, V. and D. Y. Xu (2013). forthcoming.finance and misallocation: Evidence from plant-level data.. American Economic Review. Quadrini, V., E. Mendoza, and V. Rios-Rull (2009). Financial integration, financial deepness and global imbalances. Journal of Political Economy 117 (3). 14

0.25 Growth rate across different asset 0.2 0.15 Growth Rate 0.1 0.05 0 0.05 0.1 4 3 2 1 0 1 2 Asset Figure 6: Growth Rate vs Asset, by Sector, Model Simulation Song, Z., K. Storesletten, and F. Zilibotti (2011). Growing like china. The American Economic Review 101 (1), 196 233. Wang, P., Y. Wen, and Z. Xu (2012). Two-way capital flows and global imbalances: a neoclassical approach. Federal Reserve Bank of St. Louis Working Paper Series (2012-016). A Changes over Time (incomplete) Comparing over time, and taking into account the macroeconomic changes provide more regularity to our calibration. Table 7 and Table 8 show how firms leverage, interest rate and growth rate change in our sample. 15

Table 5: Parameters Parameters Parameter Value Risk free rate r.02 Capital depreciation rate δ.10 Span-of-control α.65 Fraction of SOE µ.30 Productivity growth g.05 Capital adjustment cost φ.7 Discount factor β. Persistence of shocks ρ. Standard deviation of shocks σ. Pareto shape ν. Pareto upper bound A. Fixed credit cost ξ. Probability regains access λ. Productivity losses after default γ. Table 6: Data Moments Sectors SOE Non-SOE Mean leverage 0.77 0.56 Interest rate 0.012 0.024 Growth rate 0.10 0.21 Variance of growth rate 0.40 0.39 IQR of leverage (75/25) 1.04 Persistence of firms sales.943.838 Variance of firms sales 2.27 1.36 Distribution of value added TOP Percentiles SOE non-soe 5 0.763244 0.5298738 10 0.8616406 0.6582288 20 0.9316424 0.7838495 16

Table 7: SOE vs Non-SOE, Year 1998 and 2005 1998 2005 Average SOE Non-SOE Average SOE Non-SOE Leverage.666.772.622.579 0.769.564 Interest Rate.056.042.062.023 0.012.024 Growth Rate.014 -.066.046 0.208.10.214 Table 8: Size(VA) Distribution: Fraction of VA Produced, Year 1998 and 2005 1998 2005 TOP Percentiles SOE non-soe SOE non-soe 5.6832755.4341073 0.763244 0.5298738 10.7916158.5725119 0.8616406 0.6582288 20.8887547.7162838 0.9316424 0.7838495 17