Trade Integration and the Trade Balance in China June 2017 George Alessandria University of Rochester and NBER Horag Choi Monash University Dan Lu University of Rochester First Draft: March 2016 Abstract We study China’s growth and integration (trade and …nancial) in a two country DSGE model with a dynamic exporting decision, pricing-to-market, incomplete …nancial markets, and aggregate shocks to trade barriers, productivity, and preferences. We estimate the changes in technology, trade costs, and preferences accounting for the dynamics of China’s gross and net trade ‡ows, export participation, real exchange rate, and growth from 1990 to 2014. We …nd a large unanticipated decline in bilateral trade barriers with persistent, but not permanent, innovations that include an important gradual, phased-in component. Since the Great Recession, average bilateral barriers have stabilized at low levels even as barriers on Chinese imports have risen substantially relative to exports. Trade stagnation since 2011 largely re‡ects the completed transition to past trade reforms rather than an increase in trade barriers or reversal in the expected pace of future integration. Trade is forecast to decline almost 1 percent per year starting in 2017. Changes in trade barriers are an important determinant of China’s trade balance and its accumulation of foreign assets, accounting for as much as 70 percent of the foreign assets accumulated by 2014. Shocks to trade barriers and in China are found to have increased ROW consumption by 11.9 percent and employment by 0.6 percent but lowered ROW output by less than 1 percent relative to 1990. JEL classi…cations: E31, F12. Keywords: Trade Integration, Trade Balance, Real Exchange Rate, International Business Cycles, Net Foreign Assets. [email protected]; [email protected]; [email protected]. We thank Raphael Auer, Oleg Itskhoki, Joe Steinberg, Przemyslaw Wozniak and participants at the Atlanta Fed/Emory International Economics Workshop, Bank of Canada-Toronto Workshop on Chinese Economy, IMF, and the NBER ITI, IFM and MWATS meetings for helpful comments. Carter Mix provided excellent research assistance.

1. Introduction China’s economic growth and integration with the world economy has been an important and de…ning economic event of the last twenty-…ve years. It has been characterized by a substantial rise in income, a large increase in international trade, persistent and large trade surpluses, the accumulation of substantial net foreign assets, and important swings in the real exchange rate.1 It has also been politically contentious with politicians, particularly in the US, linking China’s sustained trade surpluses to its trade and commercial policy. The political rhetoric surrounding China’s integration has been dismissed by most academics explicitly or implicitly. Indeed, most studies of the Chinese economy are undertaken in a one-good model that abstracts from the large increase in gross trade ‡ows or swings in the real exchange rate. Our goal is to study the joint determination of trade integration and trade imbalances and evaluating whether these common abstractions matter, as politicians suggest. We focus on the contribution of changes in trade barriers in China’s trade balance and foreign asset accumulation for three main reason. First, this is a period with substantial changes in policy and non-policy trade barriers as China has gone from a real trade share of GDP of about 25 percent to 75 percent even has it share of the world economy has quintupled. These changes in trade barriers have di¤ered on exports and imports and over time, and have had substantial forward looking components, providing a rationale for intertemporal trade. Second, the surge in China’s trade surplus as a share of GDP from 2004 to 2009 coincided precisely with it joining the WTO and the expansion in international trade. We show that once one accounts mechanically for the change in trade this surge looks quite similar to the earlier surge following the Asian crisis. Third, some recent papers have shown that trade policies do a¤ect the trade balance.

2

In particular, Alessandria and Choi (2015) show that

1

Two striking features of China’s economic growth over the last twenty years have been a sustained trade integration, with imports and exports expanding at roughly double the pace of economic activity, and large trade surpluses. For instance, China’s share of US trade (exports and imports) has grown from 2.5 percent in 1992 to 10.6 percent in 2013. This growth has been unbalanced with Chinese exports to the US on average 3 times Chinese imports from the US. Indeed, China has grown from being a slight debtor to the rest of the world of about 5 percent of its GDP in 1992 to being creditor of about 21 percent of its GDP in 2011 (Lane and Milesi-Ferreti 2007). 2 See for example, Ju, Shi, and Wei (2012), Barratieri (2014), Alessandria, Choi, and Ruhl (2014), Alessandria and Choi (2015), and Reyes-Heroles (2016)

1

a substantial share of the large expansion in the US trade de…cit as a share of GDP in the 2000s can be attributed to changes in trade barriers. We emphasize three mechanisms through which changes in trade barriers a¤ect the trade balance. First, transitory di¤erences in inward versus outward trade barriers provide a motive for borrowing and lending by changing the terms of trade in a predictable way. If barriers on Chinese exports decline before its import barriers, perhaps because trade policy vis-a-vis China has some additional objectives, then China will have an incentive to save for the future. Second, even when import and export barriers change by the same amount, they also a¤ect the trade balance as the scale of borrowing and lending in response to non-trade related shocks is expanded. Quite simply, a closed economy can not borrow and lend while an open economy can. Third, in an asymmetric world, a transitory decline in worldwide trade costs has a larger e¤ect on consumption in the smaller, more open country that it would like to smooth out.3 We study China through the lens of a two-country DSGE model with incomplete asset markets, heterogeneous producers, and endogenous trade participation. In our model trade barriers make exporting an explicitly forward looking decision.4 These types of dynamic models have been shown to best explain producer level export participation as well as the dynamics of trade integration. Indeed, Pierce and Schott (2016) and Handley and Limao (2017) emphasize China’s trade growth is best explained by a dynamic trade model since it allows for trade to respond to a variety of contemporaneous and future changes in trade barriers. The model has a di¤erent short-run and long-run trade elasticity which allows us to more accurately assess the sources of trade integration. We discipline the model with data on export participation from the Chinese Census of Manufactures. We allow for persistent changes in trade barriers, technology, and preferences. Technology and preference shocks capture the usual mechanisms emphasized as a¤ecting intertemporal trade. There are persistent shocks to common trade barriers and country-speci…c …xed and variable costs of trade. Common trade barriers are also hit by persistent shocks to 3

The …rst two channels are emphasized in Alessandria and Choi (2015) while the third is absent as they consider two symmetric countries. 4 The model here follows Dixit (1989), Baldwin and Krugman (1988), Das, Roberts, and Tybout (2007), Alessandria and Choi (2007, 14a,b) and Alessandria, Choi, and Ruhl (2014).

2

their growth rate. These growth rate shocks capture the phase-in period of trade pacts. We estimate the model using data on gross and net trade ‡ows, the real exchange rate and output in China and the rest of the world, and the dynamics of producer-level export participation from 1990 to 2014. A key output of our analysis is the estimated processes for trade barriers and shocks, making our model well-suited to evaluate the current and future dynamics of trade intensity in China and the rest of the world. Three things stand out. First, much of China’s trade growth was unexpected. Trade integration in the 2000s re‡ected a persistent negative shock to the growth rate of worldwide trade barriers starting in 1999 and a series of reductions to Chinese and worldwide trade barriers. Second, the Great Trade Collapse re‡ects a transitory rise in worldwide trade barriers but no persistent increase in the growth rate of trade barriers. Since the Great Recession, average barriers have stabilized at low levels while China has erected large barriers on inward trade ‡ows. Third, most of the slow-down in trade since the Great Trade Recovery re‡ects the waning in‡uence of past trade agreements, and the lack of new innovations, rather than an outright reversal. Going forward trade is forecast to decline by almost 1 percent per year as trade barriers are expected to rise. We …nd that changes in trade barriers do matter for the evolution of China’s trade balance, particularly the post-WTO surge in trade surpluses. In total, without these changes in trade barriers, China’s net foreign assets in 2014 would have been about 30 percent of their current level. We also …nd that changes in trade barriers and shocks to preferences and technology in China are found to have increased ROW consumption by 11.9 percent and employment by 0.6 percent but lowered ROW output by less than 1 percent relative to 1990. We …nd a larger role for changes in trade barriers in our benchmark dynamic trade model than a model with a static exporting decision. As this alternative model lacks a mechanism for trade to grow gradually it requires more persistent shocks to trade barriers to explain the same data. With close to permanent shocks there is less incentive to borrow and save. Additionally, because trade shocks are more persistent in this alternative model, the outlook for trade is less pessimistic than in our benchmark. The next section describes some related literature. Section 3 describes some salient fea3

tures of Chinese integration. Section 4 builds a dynamic general equilibrium model. Section 5 describes the solution and estimation of the model. Section 6 describes the properties of the model. Section 7 summarizes the source of various changes in our estimated economy. Section 8 reports the sensitivity of our results to our modelling assumptions. Section 9 concludes.

2. Related Literature A key contribution of this paper is to study China’s trade integration and imbalances in a uni…ed manner. Previous analyses of the determination of Chinese imbalances abstract from gross trade ‡ows and relative prices. Some examples of papers motivated by the China experience are Caballero, Farhi, and Gourinchas, (2008), Choi, Mark, and Sul (2008), Quadrini, Mendoza, and Rios-Rull (2009), Buera and Shin (2009), Song, Storesletten and Zillibotti (2011), Chang, Chen, Waggoner, and Zha (2015), Coeurdacier, Guibaud, and Jin (2015). In general, these papers propose various mechanisms to explain why China with its persistently fast growth rate does not borrow. We abstract from these mechanisms, although they will be captured in a reduced form way with our non-trade related shocks.5 Studies of Chinese integration have generally been undertaken in static trade models with exogenous imbalances (di Giovanni, Levchenko and Zhang, 2012 Tombe and Zhu 2013, Autor, Dorn and Hanson 2013, 2015). A key contribution of our paper is to estimate a dynamic GE trade model in a large trade liberalization episode. Our approach to measuring changes in trade barriers is consistent with the larger Gravity literature (see a survey by Anderson 2011) which uses changes in bilateral trade ‡ows and a model to infer changes in trade costs. Examples of recent work in this spirit that examines the role of changes in trade costs in aggregate ‡uctuations (Levchenko, Lewis, and Tesar, 2010, Jacks, Meissner, and Novy, 2011, Eaton, Kortum, Neiman and Romalis, 2014). Unlike this work, we use a two-country dynamic trade model to infer the changes in trade barriers as in Alessandria and Choi (2014b, 2015) and Alessandria, Kaboski and Midrigan, (2010, 2011, 2013). By using a dynamic trade model, we are able to capture the well-known lagged e¤ect of relative 5

Chari et al. (2007) emphasize that gaps in …rst order conditions, or wedges, in simple GE models can be used to capture the e¤ects of policies or shocks in richer models.

4

prices and trade barriers on trade ‡ows6 through the internal propagation of the model rather than an exogenous shock process. It also allows us to separate out the anticipated and unanticipated changes in trade barriers and also use the model to forecast trade growth.

3. Evidence We begin by summarizing some salient features of China’s integration into world markets from 1990 to 2014. We move from the more aggregate features to disaggregate aspects. While most of the evidence is well-known, it serves as a useful foundation for the model and quanti…cation. Rapid catchup: Over the last twenty …ve years, China’s real GDP increased by 900 percent, with an annual growth rate 9.6 percent, while the average annual growth rate was 2.4 percent for the US during this period. The sustained high economic growth makes China the world’s second largest economy after the United States. Panel (a) of …gure 1 shows the real GDP of China relative to the US, and relative to the rest of the world. China’s growth also seemed to accelerate relative to the ROW after 2005. Rapid trade integration: Meanwhile, trade growth has been phenomenal. Average annual growth in China’s real exports and imports has been about 14 percent (14.2 percent for real exports and 13.7 percent for real imports). As a result, China trade’s volume are 30 times higher than 25 years ago. Panel (b) of …gure 1 shows total trade (measured as exports plus imports) relative to GDP both in real and nominal terms. In terms of the real trade share, there is relatively stable trade from 1990 to 1996, a rapid increase between 1996 to 2007, followed by a stable trade share from 2007 to 2014. In the period of rapid integration, the real trade share rose from 31 to 76 percent. Since 2007 it has held roughly steady around 75 percent. Substantial trade surpluses and de…cits: The growth of trade has been unbalanced as China ran substantial trade surpluses. Panel (c (d)) of …gure 1 shows the real (nominal) trade balance and as share of GDP. There are substantial ‡uctuations with peak surpluses 6

There is a long tradition of considering the dynamic response of aggegrate trade ‡ows and relative prices (Magee, 1973, Junz and Rhomberg, 1973, Meade, 1988, Backus, Kehoe and Kydland, 1994). Examples of papers that estimate the dynamic response to relative prices include Hooper, Johnson and Marquez, 2000, and Gallaway, McDaniel, and Rivera, 2003.

5

in 1997/1998 and again in 2007/2008. In real terms, China ran a small de…cit in 1990s that turned into a surplus around the time of the Asian Crisis. Starting in 2004 the real trade surplus increased dramatically, reaching 10 percent of GDP in 2008. It then fell with the Great Recession. To partly control for the change in trade as a factor in the trade balance, we decompose the movements in the trade balance to GDP ratio into the movements in the m ) and trade to GDP ( x+m ) and then plot a counterfactual ratio of trade balance to trade ( xx+m y

trade balance holding trade constant at its pre-WTO level tb=yjpre

WTO

=

x mx+m jpre x+m y

WTO

This alternative measure of the trade balance suggests a much smaller surge in trade surpluses around 2004. Indeed, the peak trade surplus in 2008 now looks quite similar to the trade surplus in 1998. Another way to isolate the role of trade in the ‡uctuations in the trade balance is to compare the variance of the trade balance to trade ratio to the variance of the trade balance to gdp. Table 1 reports that the trade balance to trade ratio at pre-WTO trade levels accounts for only 18.9 to 24.5 percent of the ‡uctuations in the trade balance to GDP since 2000. Important role for the extensive margin: China has been expanding its export in several dimensions. China exports much more varieties and to more destinations over time. Panel (a) of …gure 2 shows that among the HS10 products the US import, China exports 45 percent of those product categories in 1992 and 80 percent in 2014. China’s value share of US imports had grown from 5 percent in 1992 to 19 percent in 2014. Panel (b) of …gure 2 shows the overall number of countries and HS 6 digit product pair in Chinese exports. It increased to 374324 in 2014 and is 4 times larger than in 1992. Producer level export participation and intensity display similar dynamics. The Chinese Annual Survey of Enterprises 1998-2012 conducted by the Chinese National Bureau of Statistics includes all the manufacturing (and very few mining and service …rms, we excluded them from the sample) State-Owned and non-SOEs …rms with sales over 5 million RMB (about 600,000 US dollars), yielding 133,426 …rms in 1998. This number rises to 381,739 in 2008 and drops to 320,391 in 2010.

6

The expanding of the sample could be due to …rms’entry, or more …rms have nominal value of sales higher than the survey threshold over time, or the survey has been expanding its coverage. To deal with the sampling problem, we run a linear probability model of …rms export participation on …rms size, controlling for industry and year …xed e¤ects, then we calculate the unconditional export participation for year 1998-2010. From the Chinese survey and data from the US Exporter Pro…le, we see a rapid expansion in export participation. For Chinese …rms, panel (c) of …gure 2 shows the dynamics of export participation for the period 1998 to 2010. Export participation grew consistently from 1998 to 2005 from 24.4 percent of …rms to 32.7 percent. From 2005 to 2009 it contracted to 27.3 percent of …rms and in 2010 it rose to 28.8 percent. To measure the increase in exporters selling to China we use data from the US. Here we see that only about 0.8 percent of US …rms exported to China in 1992 and that this has grown to about 6 percent in 2014, with expansion leveling o¤ since 2010 (panel (d) of …gure 2). Finally, we turn to a brief discussion of changes in trade barriers. An important source of the rising integration between China and the Rest of the World has been a reduction in policy and non-policy barriers to trade. An example of the reduction on inward barriers is a gradual fall in tari¤s on imports to China from 36.4 percent in 1990 to 4.2 percent in 2014 (Panel a of …gure 3). Similarly, the US ad valorem tari¤ rate (weighted by trade) on Chinese goods decreased from 9.4 percent in 1990 to 2.8 percent in 2014 (Panel b). Moreover, the US granted Permanent Normal Trade Relations (PNTR) to China in October 2000, which became e¤ective upon China joining the WTO at the end of 2001. Before that, even though Chinese goods were subject to the relative low NTR tari¤ rates since 1980s, they required renewals every year. Without successful renewal, the import tari¤s would jump to the higher non-NTR tari¤ rates. Figure 3 (c) shows the distributions of NTR rates and non-NTR rates for HS 8 digit products. Non-NTR rates are in general much higher, and on average 7 times higher than NTR rates. Pierce & Schott (2016) and Handley and Limao (2017) shows that PNTR removed the uncertainty and was associated with the increases in US imports from China and the number of Chinese exporters. Of course, tari¤s are only one barrier, and there are many others such as transportation costs, quantitative restrictions, ownership constraints, licensing, among others. Many 7

of these non-tari¤ barriers have been removed over time as well. In 1994, the Multi…ber Agreement (MFA) was replaced by the Agreement on Textiles and Clothing (ATC), which implemented a gradual 10-year phase-out of the quota restrictions (see Brambilla, Khandelwal, and Schott, 2010). Figure 3 (d) shows the distribution of the …ll rate (US import from China/Quota) of Chinese Textiles and Apparel products.7 In 1999, most products have a …ll rate equal to 1, indicating the US imports from China were constrained by the quotas. The quotas increased over time, fewer Chinese Textiles and Apparel products were constrained by the quotas in 2004, and the MFA expired on Jan 2005. The end of the ATC lead to additional quotas and surge protection in the EU and US from mid 2005 to 2008. The Chinese government also made major reforms to its export license system (see Khandelwal, Schott, and Wei, 2013). By 2000, private …rms with registration capital less than 8.5 million Chinese Yuan were restricted to export. By 2003 this number decreased to 0.5 million and very few …rms are restricted. There has also been numerous dumping cases both against China and by China (Bown, 2010). These features suggest substantial, but uneven and asymmetric declines, in trade barriers along with a substantial forward looking aspect to trade reform. Moreover, there have been changes in entry barriers, tari¤s and per unit shipping costs. To capture these features requires a general model with various forms of trade barriers and trade policy dynamics.

4. Model We now develop a two country dynamic stochastic general equilibrium model with heterogenous producers entering and exiting the export market with aggregate shocks to productivity, trade costs, and the discount factor. The model combines features of the heterogenous producer trade models with features of standard international macro models. It is a variation of the equilibrium dynamic export participation model of Alessandria and Choi (2007, 2015) that includes shocks to trade costs and the discount factor, variable markups, and incomplete asset markets. We assume that trade cost shocks are exogenous rather than being the outcome of some trade agreement among nations. The process for these shocks is quite 7

Only WTO members were included in the phase-out of under the ATC and so China’s quotas remained quite constrained until 2001 when it joined the WTO, at which point it jumped to the new levels of quotas.

8

general as it allows for country-speci…c changes in barriers as well as changes to future trade barriers. In each country, consumers consume a non-tradeable good made by combining a di¤erent mix of tradable intermediates, make a labor-leisure choice, and trade a non-contingent bond. Home and foreign prices are normalized to 1: Pt = Pt = 1; and the real exchange rate is de…ned as the relative price of a basket of foreign to home goods (a depreciation is an increase). The home country is China and the foreign country is the rest of the world. The benchmark model abstracts from capital accumulation for two reasons. First, this allows us to focus directly on how changes in trade barriers generate intertemporal trade. Second, it allows us to abstract from why China is not running large trade de…cits given that it has been growing very fast for a long period in a very parsimonious manner. This has been the topic of much research but it is not clear that the mechanisms being proposed are relevant for the trade channel emphasized here. Nonetheless, we later consider a variation of our model with capital accumulation and show changes in trade policy will generate similar movements in the trade balance and real exchange rate as in our benchmark model. Household: There is a representative household that captures the behavior of consumers and government.8 It maximize the discounted sum of utility, max E0

Ct ;Lt ;Bt

subject to:

where U (C; L) = C (1

1 X

Ct + Vt 1 + L)1

1

tU

(Ct ; Lt ) ;

t=0

= (1

Vt Bt 2 YtN b

);

Bt = Wt Lt + Bt

t

1

+

t

+ Tt ;

is the dividend payments from home …rms

and Tt is a lump-sum rebate of revenue from trade barriers and any portfolio holding cost, and Vt is the price of a noncontingent bond. There is also a small bond holding cost of b Vt B 2 YtN t

for home with YtN being nominal home GDP and

b Vt B t 2 qt YtN

for foreign. We also allow

8 See Chang (2015) for a discussion of the di¤erent saving motives of the government and household sector in this period.

9

the home discount factor to vary over time ln ( where

t+1 =

is the steady state

t)

= ln

and "

;t

t

= (1

b ) ln

+

b

ln

t 1

+" ;

is a shock. In a standard model

=

t

t

: These types of

shocks to the discount factor were introduced into international macro models by Stockman and Tesar (1995) and have been used extensively to explain the high savings rate of China (Kehoe, Ruhl and Steinberg, 2014) as well as crises episodes (Eggertson and Woodford, 2003, Christiano, et al. 2011). They are a parsimonious way to capture the rich array of explanations for China’s saving that are unrelated to trade. Aggregation Technology or Consumption Index: In each country, a competitive retail sector combines a continuum of domestic goods with a set of available imported goods to produce a …nal non-traded good.9 There is a unit mass of producers in each country. The aggregators are as follows: 1

1

1

YHt + a YF t

Dt =

1

; YHt =

Z

1

1

Yhit di

; and YF t =

Z

Yf it di

q

= qt

t

!

t t

1

:

and a is a measure of the taste for imported goods.10 The

elasticity of substitution for imported goods is allowed to be time varying, t

1

t

i2Et

0

The Armington elasticity is

1

t

= qt q yr;ty

yr;t y with qt being the real exchange rate in terms of home aggregate (a rise in q

means real depreciation of home) and yr = YH =YF is relative real income. The parameters q;

y

allow the markup on imported goods to vary systematically with the real exchange

rate and relative income. An appreciation of the Chinese real exchange rate (q falls) will lead to a decline in the demand elasticity and rise in the markup on goods exported to China if

q

> 0: Likewise as China gets richer so that yr rises, markups on exports to

China will rise if

y

< 0: This is a parsimonious way of embedding pricing-to-market that

9

Trade here is driven by product di¤erentiation, trade costs, and producer heterogeneity. We abstract from other comparative advantage based sources of heterogeneity. 10 The taste for the imported good, a; is a normalization. It is straightforward to increase the trade cost and lower the taste parameter without changing the properties of the model.

10

allows for incomplete exchange rate pass-through and systematic deviations from the law of one price with income. It can be microfounded using search frictions as in Alessandria (2009), Alessandria and Kaboski (2011), or Drozd and Nosal (2013). The price indices for the aggregates are

PHt =

Z

1

1 Phit di

1

; PF t =

Z

Pf1it t di

i2Et

!1 1

t

1 ; and Pt = PHt + aPF1 t

1 1

= 1:

In equilibrium …nal absorption equals consumption, Dt = Ct : Producers: Each country has a unit mass of producers each specialized in producing a di¤erentiated variety using just labor. Producers are heterogenous and face idiosyncratic and aggregate shocks to productivity and trade costs. The production function of a producer i is given by Yit = ezt + it Lit ; where zt is the country-wide productivity and

it

is the producer-speci…c productivity. The

country component of productivity follows an AR1 process which depends on a global and country-speci…c component. To export a producer must incur both …xed and variable trade costs. To export in the current period, producers must hire some domestic labor. The amount of domestic labor depends on a producer’s export history. A new exporter will pay Wt f0t to start exporting while a continuing exporters will pay Wt f1t . These costs are sunk and cannot be recovered in future periods. When an exporter stops exporting for a period to re-enter requires paying the cost Wt f0t : Producers also face a (gross) marginal trade cost given by and

;t

t

for China exporters,

for ROW exporters. The …xed export costs is allowed to di¤er across countries and

potentially change over time. The resource constraint for each good is Yit = Yhit + mit t Yhit ; where mit is the exporting decision of producer i in period t. The marginal trade cost is stochastic. 11

The dynamic programming problem of a …rm is then Vt ( ; m) = max pct (p) + m0 p ct ( p ) 0

m0 W fmt + Qt EVt+1 ( 0 ; m0 )

Wl

m ;p;p

where m = f0; 1g is an indicator that summarizes past export status and thus the current cost of exporting, fmt . It is well known that when Wt f0t > Wt f1t ; the decision to export is forward looking. It is also well-known that there is a threshold technology for exporters to continue exporting ( and a second threshold technology for non-exporters to start exporting (

0t ).

1t )

Producers will

move in and out of the export market in response to idiosyncratic and aggregrate shocks. These thresholds satisfy the following equations Wt f0t

t

(

0t )

= Et Qt+1 Vt+1 ( 0 j

0t )

Wt f1t

t

(

1t )

= Et Qt+1 Vt+1 ( 0 j

1t )

Vt ( ) = Vt ( ; 1)

Vt ( ; 0)

The left hand side of the …rst two equations measures the net current cost of exporting and is equal to …xed export cost minus the current pro…t. The right hand side measures the discounted expected gain in producer value from starting the next period as an exporter rather than the non-exporter, where Qt+1 is the stochastic discount factor. The advantage of starting out as an exporter is related to savings by not having to pay the high entry export cost. When Wt f0t > Wt f1t ;we have that

0t

>

1t

so that new exporters are relatively

productive compared to continuing exporters and there is what Baldwin and Krugman call exporter hysteresis in that exporter’s enter when their costs are relatively low and remain in the market even when their costs are relatively high. When Wt f0t = Wt f1t the export thresholds are identical,

0t

=

1t ,

and exporting is a static decision.

The measure of exporters and non-exporters over idiosyncratic productivity in each country is a state variable. It is potentially an in…nitely dimensional object. To simplify the state 12

space we follow Alessandria and Choi (2007) by assuming idiosyncratic productivity shocks are iid. This implies that we don’t need to keep track of the distribution of productivity of last period’s exporters and non-exporters, only the stock of past exporters. The stock of exporters evolves as Nt = Nt Nt

= Nt

1

Pr (

1t )

+ (1

Nt 1 ) Pr (

0t )

1

Pr (

1t )

+ 1

Nt

0t )

1

Pr (

Aggregate Variables: To take the model to the data requires de…ning some aggregate variables. Nominal output (GDP) is given by YtN

=

Z

Phit YHit +

Phit Y di: qt Hit

Real GDP is given by YtR =

YtN ; PHt

where we de‡ate nominal revenue by the domestic price of goods. Nominal exports are given by EXtN

=

Z

qt Phit Yhit di =

a 1 P Dt : qt Ht

The export price index is given by PXt =

PHt : qt t

Real exports are given by EXtR =

EXtN =a PXt

1 t

qt PXt Dt :

Nominal imports are given by IMtN

=

Z

Pf it Yf it di = aPF1 t Dt :

13

The nominal trade balance to nominal GDP ratio is given by N XYt =

EXtN IMtN : YtN

The import price index is given by PM t =

PF t

:

t

Real imports are given by IMtR =

IMtN =a PM t

1 t

PM t Dt :

Finally, we de…ne the nominal export-import ratio as EXM N = EXtN =IMtN and note that the trade balance as a share of trade is proportional to the log of the nominal export-import ratio EXtN IMtN EXtN + IMtN

0:5 ln EXM N :

Taken together, the nominal trade balance as a share of GDP can be decomposed as the product of the nominal export-import ratio and trade share N XYt

0:5 ln EXM N

EXtN + IMtN YtN

In all trade models, the …rst term is primarily determined by asymmetries across countries from shocks while the second term is primarily determined by the average level of trade costs and the relative size of countries.

5. Solution and Estimation We solve the model by linearizing it around the steady state. Given that we are dealing with a large model, we …x several parameters to conventional values and estimate the rest using Bayesian techniques. Table 1 reports calibrated and estimated parameters. 14

The time period is a year and so we set

= 0:96: The weight on leisure is set so that hours

worked in the rest of the world is equal to 1/4. The bond adjustment cost is set to 0.0001 to ensure stationarity but otherwise not a¤ect our results. The elasticity of substitution across varieties, ( = 5), is chosen to yield a 20 percent markup (Chang et al, 2015). The …xed trade costs (f0 ; f1 ; f0 ; f1 ), standard deviation of idiosyncratic productivity shocks

;

, variable trade costs, weight in the aggregator (a), and the elasticity of

substitution between imported and domestic goods ( ) determine trade ‡ows. We assume steady state …xed costs are the same in China and the ROW.11 Of course, given that …xed costs are based in units of local labor there will be large di¤erences across countries in the cost of exporting related to the di¤erences in wages. The standard deviation of idiosyncratic shocks is also assumed to be the same across countries and constant.12 To determine the parameters related to trade and heterogeneity we proceed in two steps. First, we choose the trade costs parameters so that in a symmetric version of our model with no iceberg costs we would have the following characteristics of trade and producers involved in trade: 1. 15 percent of producers export, 2. 12.5 percent trade share of gdp, 3. 2 percent annual exit rate, 4. Exporters that are 2.5 times larger than non-exporters. This yields an export entry cost that is almost 9 times the cost of staying in the market and a standard deviation of shocks of 23.5 percent. We use these estimates as our prior in our estimation in the second stage. To clarify the rest of the estimated parameters, we now describe the shocks. Productivity, trade cost, and discount factor shocks are independent. There is a global 11

We also estimated a version with di¤erent …xed costs but found these di¤erences to not be sign…cant when there are also shocks to …xed costs. 12 The idiosyncratic productivity are assumed to be the same since we would need more data on the characteristics of ROW exporters and there as some important changes in the China survey sample that would need to be considered. Additionally, many of these di¤erences show up as a¤ecting the amount of cross-country substitution from a country-speci…c shock and get absorbed into the estimated Armington elasticity.

15

and China-speci…c productivity shock. The global productivity follows an AR1 process, c z

ln zct =

ln zf t

+ "czt ; "czt

1

iid

c z)

N (0;

while productivity in China depends on the global component and its country-speci…c component, ln zht = zf t + zd;t d z

ln zdt =

ln zdt

z 1

+ "dzt ; "dzt

iid

N 0;

d z

where z is the productivity disadvantage of China relative to the rest of the world. There are 6 trade costs split between two variable costs and four …xed costs. The variable trade costs and China’s …xed costs vary over time. For the iceberg trade costs, the countryspeci…c iceberg cost includes a common cost and a di¤erential cost, ln

t

= ln

ct

+ ln

dt ;

ln

t

= ln

ct

ln

dt :

To capture the e¤ect of phasing in of trade barrier reductions the common iceberg trade cost is assumed to have both a transitory and more persistent growth rate component,

gt 1 .

It

is assumed that the shock to the growth of the common iceberg trade cost is known 1 year in advance of implementation while the transitory common shock occurs contemporaneously. Allowing for a gradual decline in global trade costs will allow for trade to grow persistently without relying on a sequence of unanticipated shocks to trade costs. The process for the di¤erential trade cost is also assumed to be persistent, ln

ct

=

ln

gt

=

ln

dt

=

1 g

1

c

ln

ln

gt 1

d

c

+

c

ln

ct 1

+ ln

ln

dt 1

+ " dt:

gt 1

+ " ct;

+ " g t;

ln

d

+

d

Shocks to variable trade costs are assumed to be independent and have standard devia16

tions (

c;

d;

g) :

We also allow for persistent shocks to China’s …xed costs, ln f0t =

1

f0

ln f0 +

f0

ln f0t

1

+ "f 0;t ;

ln f1t =

1

f1

ln f1 +

f1

ln f1t

1

+ "f 1;t :

In total, there are seven exogenous variables, (zc ; zd ; ; c ;

d ; f0 ; f1 )

driven by eight shocks

"fzt ; "dzt ; " c ; " g ; " d ; "f 0 ; "f 1 ; "b . The shocks are assumed to be independent. The persistence of the …xed export costs are constrained to be the same

f0

=

f1

=

f

. We also estimate

the level of variable trade costs and the productivity gap between countries pricing-to-market parameters

q;

y

c; d; z

, the

; and the Armington elasticity and risk aversion para-

meters ( ; ). These parameters are estimated using the following six time series from China and the rest of the world for the period 1990 to 2014: 1. China’s relative size measured using constant 2005$, 2. A weighted average of ROW GDP with a linear trend removed, 3. The real exchange rate (OECD), 4. The ratio of nominal exports to nominal imports13 , 5. The share of trade in real gdp (X+M)/GDP, 6. The share of Chinese exporters to the rest of the world (1998 to 2010). For these parameters we have relatively ‡at priors (table 1). Figure 4 depicts the series used to estimate the model and the results from the estimation. Even though we have missing data on Chinese export participation in the 1990s and since 2010, the estimation yields quite reasonable movements in export participation. In particular, it suggests gradual expansion prior to 1998 with some reasonable swings. It also predicts a robust expansion since 2010.14 Figure 5 depicts the estimated innovations to productivity, tastes, and trade costs. 13

We chose to use the real trade share rather than the nominal trade share for three reasons. First, it is consistent with matching real GDP. Second, we would like to evaluate the impact of trade and thus would like to capture how real expenditures have shifted to traded goods. Finally, some of the fall in nominal trade since the Great Recession re‡ects a decline in commodity prices and our model is not well suited to capture these types of price changes. Results matching the nominal trade share are available. 14 These movements seem somewhat consistent with the alternative measures of the extensive margin documented.

17

Table 1 reports some moments related to our estimated parameters. We discuss the posterior mode of estimated parameters. The Armington elasticity is estimated to be 1.82 and the risk aversion parameter is 4.42 which are consistent with values commonly used in the literature on international business cycles (see Chari, Kehoe and McGrattan, 2002). The Armington elasticity is lower than commonly employed in trade integration studies, where the tendency is to set

= . It is substantially lower than employed in studies of

the China’s trade balance where home and foreign composites are assumed to be perfect substitutes (Song et al. 2011 and Chang et al. 2015). We also …nd that the cost of starting to export is about 9 times the cost of continuing in the export market which again is consistent with the literature (see for example Das, Roberts, and Tybout, 2007, or Alessandria and Choi, 2014). We estimate that the dispersion in idiosyncratic shocks is 0.185. These three parameters govern the characteristic of exporters as well as the strength of the extensive margin response. We also estimate that there is about two times the pricing-to-market with the real exchange rate (0.053) as with relative income (-0.028). The shocks are estimated to be quite persistent with annual autocorrelations that range between 0.73 and 0.999. The least persistent shock is the detrended global productivity shock with an autocorrelation of about 0.738 while China’s productivity is very close to a unit root (0.999). The discount factor shock has an autocorrelation of nearly 0.974. Di¤erential trade costs are also quite persistent at over 0.987. The transitory common trade cost has an autocorrelation of 0.962 while the persistence of the growth shock is 0.975. Shocks to export …xed costs are also persistent with an autocorrelation of 0.75. Owing to the sustained growth of China and the persistent swings in the real exchange rate, the persistence of the China’s productivity is most precisely followed by the di¤erential trade costs. There is more uncertainty about the persistence of the common trade cost15 and the discount factor shock, although less than with the trend shock or the …xed cost shocks given we only have 25 years of data. Transitory common and di¤erential trade costs are roughly equally volatile and 15

Without the trend trade cost shock we estimate nearly unit root common trade costs. We do not follow this approach as introspection and the estimates of the persistence and variance of trend shock suggests trends matter.

18

about 10 times as volatile as the trend trade cost shocks. Figure 6 plots productivity, discount factor, and trade costs as deviations from the steady state. China productivity grows quite substantially while ROW productivity ‡uctuates around zero and there is a substantial decline with the Great Recession. The Chinese discount factor shows some substantial ‡uctuations over this period with relatively high patience at the beginning of the sample, in the late 90s around the time of the Asian Crisis, and in the period prior to the Great Recession. The relatively high initial patience provides an important reason that China is accumulating assets at the start of the sample. Trade costs from China fall substantially more than trade costs to China but there are some interesting variations. In particular, in the early period trade costs to China are falling while trade costs from China are rising. These two trade costs start falling in sync in the mid-to-late 1990s and then the trade cost from China starts falling quite quickly in the early 2000s until the Great Recession. Since the Great Recession trade costs from China have continue to fall while trade cost to China have actually risen. This may perhaps re‡ect the stimulus programs in China favoring expenditures on domestic goods. Panel (e) shows that at the beginning of the sample trade costs were expected to fall almost 3 percent year. The expected pace of liberalization accelerated through 2004, peaking at about 7 percent, since then expectations have reverted to about 5.5 percent. We …nd that the …xed entry costs to be stable throughout the period while the continuation cost fell and then rose temporarily around 2005. These costs have mean reverted by 2014, although one should be cautious in interpreting these since we lack evidence on Chinese export participation from 2011 onwards, although some of our variety based measures show an extensive margin expansion in this period (Figure 2). Stationarity This is a period in which China’s output and trade grows substantially and this makes the trade share and China’s relative size to appear non-stationary. For output this is captured by a sequence of nearly permanent shocks that take China from about 60 percent below its steady state to 60 percent above. One might be concerned that such large shocks might be inconsistent with our linearized solution method; however, we have solved a version of our

19

model with shocks half this size16 using global methods and …nd the impulse response of the nominal export-import ratio and real exchange rate are almost identical. We also have compared the impulse responses with di¤erent trade shares and China productivity and found them quite similar since China remains a small part of the world. Third, we explicitly account for some of the e¤ects of changes in trade barriers on trade balance as a share of GDP and the accumulation of foreign assets. Finally, we also estimate a variation of our model in which output shocks are permanent but there is a cointegrating relationship with the ROW as in Rabanal et al. (2011). While this structure alters the estimated innovations to China’s productivity and discount factor, trade shocks remain important factors for foreign asset accumulation. Trade integration is captured by a sequence of persistent negative innovations to current and future trade costs that takes our economy far from the steady state trade. These shocks are estimated to be persistent but not permanent, although we generally can not reject a unit root, particularly for the gap in trade costs. Obviously, with unit root shocks to trade costs there is less incentive to borrow and lend. We discuss this issue further when we compare our benchmark dynamic model to a static trade model.

6. Model Properties To illustrate the mechanics of our model and shed some light on the identi…cation in our estimation, we consider the impulse response from the steady state to each shock. Specifically, we consider the impact of a one standard deviation increase in each of our shocks on the variables used in the estimation (we abstract from …xed cost shocks as the impacts are primarily on export participation) evaluated using our estimated parameters. Figure 7 summarizes the dynamics of some key macro variables while …gure 8 depicts changes in export participation in China and the ROW. Panels (a) and (b) of …gure 7 shows the response of China and ROW real GDP respectively to each shock. Productivity shocks are the main drivers of real gdp. China’s real gdp rises almost permanently with China’s productivity. The global productivity shock leads to 16

Solving the model for larger shocks is computationally challenging given the persistence of the productivity process.

20

a transitory rise in China and ROW real GDP. An increase in China’s patience has a small positive e¤ect on real GDP in China and negative e¤ect on the ROW. These shocks have a larger e¤ect on China since it is smaller. Panel (c) shows that the nominal export-import ratio responds more to a larger range of shocks. Increases in patience ( ), future trade costs

g

, di¤erential trade costs ( d ) ; and

China’s productivity lead to surpluses. A rise in the transitory common trade costs leads to a de…cit. A transitory increase in the relative costs of getting goods into China relative to the ROW raises the price of Chinese consumption leading it to save and thus run a surplus. The e¤ect of changes in global trade barriers on the export-import ratio is a result of having an asymmetric model with China consumption more reliant on trade. A transitory increase in global trade costs has a larger e¤ect on the more open economy leading it to borrow to smooth consumption. A future increase in trade barriers leads China to run a surplus today to bring more consumption into those periods where consumption is lower. The real exchange rate (panel d) depreciates (rises) with an increase in Chinese productivity and patience. It appreciates substantially with an increase in the di¤erential trade cost. The contrasting e¤ects are key to identifying the shocks in the estimation since China has had a large appreciation at the same time it has grown fast and run persistent surpluses. Transitory and persistent growth changes in global trade costs have relatively mild e¤ects on the real exchange and work in opposite directions. Panel (e) show the impact of the shocks on the real trade as a share of GDP. Trade declines most in response to increases in current and future trade barriers. The impact of the transitory shock is immediate and persistent, while the changes in future trade barriers only has a gradual impact. The two shocks have the same impact after 12 years. Thus, to get a sustained increase in China’s trade share will either require a growth shock or a sustained sequence of declines in trade costs. Increases in China’s productivity lowers its trade share as it becomes a larger share of world output. The increase in patience has a small transitory negative e¤ect on trade. Finally, we consider the impact of each shock (including …xed export costs) on Chinese and ROW export participation. Given the relative size of the countries the shocks have a larger e¤ect on China’s export participation. China’s export participation increases in 21

response to more patience and decreases with global trade costs (transitory and growth), di¤erential trade costs, and China’s …xed export costs. The dynamics of export participation are hump shape in response to transitory shocks because of the need to build export capacity through the dynamic export decision. Not surprisingly, increases in global trade costs (transitory or growth shocks) lead to a decline in export participation in both countries. Increases in patience, di¤erential trade costs, and China’s productivity move export participation in China and the ROW in opposite directions. A transitory rise in global productivity increases export participation slightly as this is the only accumulable asset that can be used to smooth consumption. Finally, an increase in China’s …xed export costs leads to a decline in export participation in both countries as this e¤ectively raises the cost of accumulating exporters and leads world economy to sacri…ce these investments to smooth consumption.

7. Results We now use the estimated model to answer a range of question which essentially amount to evaluating the contribution of di¤erent shocks to the dynamics in some key variables. Our discussion moves from the long-run changes in output, consumption and trade, to the determination relative prices, of net ‡ows, and China’s net foreign asset accumulation. Figure 9 plots a shock decomposition of some key aggregate variables. Figure 10 plots a shock decomposition of trade related variables. Figure 11 plots the change in China’s trade balance and net foreign assets as a share of GDP and the contribution of di¤erent shocks allowing for the interaction e¤ects of di¤erent shocks. How much of the growth in output and consumption in China was from a decline in trade barriers? For relative output (Yh/Yf), the model attributes 99.6 percent of its growth relative to the world to China’s productivity improvements. The reduction in common trade barriers had almost no impact on China’s relative output while the rise in inward barriers accounted for 2.3 percent of the rise in China’s output share. The relatively minor impact of trade on output echoes the result in Waugh (2010) about di¤erences in trade costs having a minimal impact on di¤erences in income per capita. Trade costs have a more signi…cant impact on relative consumption though, with productivity accounting for only 88.4 percent of China’s consumption growth relative to the ROW, global trade integration shocks about 10.9

22

percent, while the di¤erential trade costs reduced the gap by 7.2 percent. Mean reversion accounted for 6.9 percent of the change in relative consumption. How have shocks in China a¤ected output, consumption, and employment in the rest of the world? As we have detrended world output, rest of world output is e¤ectively unchanged from 1990 to 2014. This re‡ects a 1.8 percentage point decline from shocks to global productivity and a 0.8 percentage point decline from the change in di¤erential trade costs that is o¤set by the waning in‡uence of initial conditions. While there is no long-run e¤ect on output, most of the ‡uctuations in output over this period are a result of common shocks to productivity. In contrast to output, world consumption grew 14.3 percent in this period, again relative to the trend in output. Chinese productivity growth more than o¤set the decline in global productivity while trade costs combined accounted for 66 percent of the growth in consumption. In total, China’s productivity growth and changes in trade barriers boosted the rest of world real consumption by about 11.9 percent. Finally, we can also consider the impact of China on employment in the ROW. Here we …nd that employment has grown 1.7 percent compared to 1990. About 62 percent of this was a result of initial conditions. The decline in common barriers increased employment by about 0.7 percent. China’s productivity growth and recent impatience boosted employment in a way that was almost completely o¤set by the rise in inward trade costs. Even though employment is higher in 2014, employment in production is 1 percent lower as substantially more workers are allocated to …xed cost activities related to trade. The combined movements in real output and employment yield a decline in measured labor productivity of 1.7 percent compared to an increase in underlying productivity of 0.7 percent. What explains the growth in trade in China and the ROW? From the top panel of Table 5, the main source of the 47.9 percentage point rise in China’s trade share of GDP (measured in real terms) was the decline in worldwide trade barriers. The gradual reduction of trade costs accounted for 43 percent of the rise while the common shocks accounted for 55 percent. Changes in …xed export costs had a minimal e¤ect on the change in trade. The 22.1 percentage point rise rest of the world trade (measured in nominal terms) is roughly equally split between common (27.8 percent) and growth rate shocks (21.4 percent), China productivity growth (18 percent), and the di¤erential trade cost shocks (12.6 percent). 23

We also …nd that much of China’s trade integration was a surprise. That is the waning e¤ects of the initial conditions contributed to about 1/3 of the growth in trade in China and 1/5 of the growth in ROW trade. In terms of the changes in variable costs, we …nd that only about 28 percent of the decline in common trade costs was expected as of 1990 while almost all of the long-run changes in the gap in trade costs was unexpected. What explains the weak trade growth since 2011? The bottom panel of Table 5 reports the average annual contribution of shocks to trade growth in China and the ROW in the period 2011 to 2014 relative to the expansion period from 1997 to 2007. Overall, we see that trade relative to GDP has grown 0.8 percentage points slower in the ROW and 4.6 percentage points slower in China. The common trade cost is the main source of the slow-down accounting for a reduction in the growth of trade to gdp in the ROW of 124 percent and 80 percent in China. This primarily re‡ects the large negative innovations in the earlier period and a lack of any additional negative innovations and mean reversion in the latter period. The trend shocks have also contributed to the slow-down in the pace of trade integration in the ROW. While future reductions in trade barriers are expected from the trend shock (…gure 12 plots the dynamics of the trend shock to trade costs), we can see clearly the trend accelerated from 1998 to 2004 but has since retrenched. These future declines from the trend are not enough to o¤set the mean reversion in trade barriers and so the trade share is expected to fall gradually over the next 15 years, with an almost 17 percent decline in China’s trade share and an almost 16 percent decline in the ROW trade share. How did changes in preferences that encouraged Chinese savings a¤ect Chinese trade integration? Changes in preferences have almost no impact on trade integration as they primarily lead to substitution between imports and exports. Indeed it is an important driver of export participation but works in opposite directions at home and abroad. Overall, we …nd that changes in the discount factor accounted for only 2.5 percent of the increase in real trade in China and less than 1 percent in the ROW. What explains the dynamics of the Chinese real exchange rate. The real exchange rate in 2014 is roughly equal to its level in 1990. Over the period though, there are some substantial swings. In particular there is a 60 percent depreciation from 1990 to 1994 that is partially reversed by 1998. The real exchange rate then appreciates by about 20 percent through 2005 24

and then depreciated by about 30 percent. The model primarily attributes the movements in the real exchange rate to the di¤erential trade cost and di¤erential productivity. Indeed, we see that the appreciation since 2004 is primarily a result of the higher trade barriers on Chinese imports more than o¤setting the expected depreciation from Chinese productivity growth and a decline in the discount factor. What accounts for the movements in the Chinese trade surplus? Here we focus on the trade balance as a share of trade (the nominal exports-imports ratio) since this removes the e¤ect of changes in the trade share of GDP. Figure 9 (e) shows that China has maintained a surplus for most years except 1994. It also shows peak surpluses of about 30 percent in 1998 and 1999 and again in 2007 and 2008. Our shock decomposition suggests that initial conditions account for the early surpluses but die out by 2005. Shocks to preferences ( ) are an important high frequency contributor that explain the surpluses of the late 1990s but had only a minor impact on the post-WTO trade surplus. Since the Great Recession, China has become less patient and this has substantially reduced the trade surplus although these e¤ects were partly o¤set by the changes in trade barriers, particularly the rise inward barriers.17 What explains the dynamics of the trade balance as a share of gdp? It is more common to study the trade balance as a share of GDP. Measured this way, the Chinese trade surplus as a share of GDP (in nominal terms) ‡uctuates substantially with surpluses of almost 5 percent of GDP in 1997 and 8 percent in 2007. From our initial decomposition in …gure 1d we have shown that a substantial portion of the relatively large post-WTO surpluses relate to trade being a large share of nominal GDP so that aggregate shocks in the latter period will end up leading to larger imbalances. To capture this interaction e¤ect, recall that the trade balance as a share of GDP is well approximated by TB GDP

0:5 ln (EXN =MN )

EXN + MN : YN

We then use the behavior of each of these variables in our model to construct the trade bal17

Shocks to the discount factor account for about 43 percent of the unconditional variance of the nominal export-import ratio.

25

ance as a share of GDP. Given that we have important interaction e¤ects and many shocks, we present the e¤ects by summarizing over three groups of variables: Initial, Productivity and Preferences (Zc ; Zd ; ); and Trade ( c ;

d ; g ; f0 ; f1 )

tively. We report 9 cases, Data; All; I; P; T; (I (P

T = PT

P

T ) ; (I

P

T = All

summarized as I; P; and T respec-

P = IP I

P

T

I

P ) ; (I

I

P

I

T = IT T

P

I

T);

T ) : From

…gure 11 (a) we clearly see that changes in trade barriers have been the main contributing factor in the expansion in the trade surplus from 2004 to 2007. We also see that the reversal of the surplus from 2007 to 2011 can be attributed to the aggregate shocks and the interaction of the aggregate shocks with the change in trade barriers. The model does not exactly match the data as it has been estimated to match the real trade share rather than the nominal trade share. As these two series move fairly closely together for most of the sample this is creates relatively minor gaps until the end of the sample. What explains the large increase in Chinese net foreign assets as a share of GDP? 11 (b) plots the change in China’s net foreign assets as a share of gdp in the model and data (from the World Development Indicators18 ). Over the period, China increases its foreign assets by 36.2 percent of GDP. To construct the model equivalent, we follow the same approach as for the trade balance. Speci…cally, the model’s initial assets to GDP (V1990 B1990 =Y1990 ) are updated using the law of motion Vt Bt = YN t

Vt 1 Bt YN t 1

1

1 Vt

YN t 1 YN t

1

+

0:5 ln (XN t =MN t ) ; YN t

and our measure of each variable from the model. As our model captures both the nominal trade balance and real trade share of gdp, the main di¤erence between our measure and the data is most likely due to di¤erences in returns on foreign and domestic assets and real and nominal gdp. Overall, the model does quite well in capturing the dynamics of net foreign assets. As before we consider the impact of three di¤erent groups of shocks. Table 4 summarizes the contribution of these shocks to the change in NFA from 1990 to 2014 (measured as percentage points of GDP). We see that initial conditions (40.2 percent), 18

We focus on this measure rather than the Lane-Milesi-Ferretti (2007) measure since this measure makes fewer adjustments for valuation e¤ects.

26

Trade costs (28.3 percent), and the interaction of Trade costs and initial conditions (25.3 percent) are the main reasons China accumulated assets while the interaction of ProductivityPreference shocks with Initial conditions (-28.4 percent) and with all shocks were forces for reduced asset accumulation (-13.9 percent). By themselves, the Productivity-Preference shocks only increased net foreign assets by 3.5 percent of gdp. The bottom three lines of Table 4 report the contribution of pairwise combinations of the shocks and initial conditions. We …nd that with just the preference and productivity shocks plus the initial conditions, China would have accumulated assets of 15.2 percent of GDP, thus changes in trade barriers accounted for nearly 70 percent of the increase in Chinese net foreign assets.

8. Sensitivity We now consider the sensitivity of our results to our modelling assumptions along four dimension. First, we consider the role of pricing-to-market and the dynamic exporting decision. Next, we consider raising the elasticity of substitution to a more common level for analyses of China’s trade and growth. Third, we evaluate how the number and structure of shocks in‡uence our …ndings. We primarily focus on understanding how these alternative assumptions in‡uence the …t of the model, the estimated parameters, and our accounting for trade and asset accumulation growth. Finally, we discuss how including capital might in‡uence our …ndings. Pricing-to-Market and Static vs Dynamic Models The benchmark model with pricingto-market is compared to an alternative model without pricing-to-market and another model without pricing-to-market and a static export decision. The benchmark model with no pricing-to-market,

q

=

y

= 0; is denoted No PTM, and then a version of this model with a

static export decision, f0 = f1 ; is denoted Static. These alternative models generate qualitatively similar results about the importance of changes in trade integration for the dynamics of the Chinese trade balance, but understate their quantitative signi…cance. Table 6 compares the posterior mode of each of the estimated parameters in each of the models. For the most part, the parameters are quite similar. The main di¤erences come in our estimated parameters related to producer heterogeneity. In the Static model, only the most productive producers export and so there must be much less producer heterogeneity to 27

generate the same amount of trade as in the benchmark model (4 percent compared to 18 percent in the benchmark). The process for aggregate shocks is also similar across models with shocks generally slightly more persistent and larger in the static exporting model. The biggest di¤erence comes in the process for the shock to …xed export costs. In the static model, the export cost shock has an autocorrelation of 0.986 compared to 0.75 in the benchmark model. Table 4 reports the contribution of each shock in these alternative models to the change in net foreign assets to GDP from 1990 to 2014 in China. The models generate slightly di¤erent increases in assets even though they each generate the same dynamics of nominalexport import ratio and real GDP because of some di¤erences in initial asset positions and the movements in nominal and real GDP. Speci…cally, eliminating pricing-to-market from our benchmark model, leads to a larger accumulation of foreign assets in China between 1990 and 2014 (66.6 percent vs 54.9 percent). Without changes in trade barriers, the model would have generated an increase in assets of 24.1 percentage points, thus trade shocks increase asset accumulation by 42.5 percentage points compared to 39.7 percentage points in the benchmark. With just the static exporting decision, the model predicts an even larger rise in foreign assets from 1990 to 2014 of 71.6 percentage points. Without changes in trade barriers, the model would have generated an increase in assets of 33.3 percentage points, thus trade shocks increase asset accumulation by 38.3 percentage points. We next consider how the source of trade integration depends on the structure of our model. Table 7 reports the contribution of each shock to the growth in the nominal trade share in the ROW and China in our three models. The top panel reports changes for the whole sample while the bottom panel focuses on the di¤erence in the annualized growth rate from 2011 to 2014 compared to 1997 to 2007. In terms of the growth in the trade share of nominal GDP over the whole sample, we …nd that eliminating pricing-to-market from the benchmark model primarily makes growth shocks less important (35.0 vs 43.0 percent in the benchmark). The reduced importance of the trend shock makes the common (59.5 vs 55.5 percent) and di¤erential costs more important (14.2 percent vs 2.7 percent) and initial conditions less important (36.9 percent vs 43.9 percent). Eliminating the dynamic export decision along with PTM yields more 28

similar results as with our benchmark model. The main di¤erence is that changes from …xed costs provide a small drag on integration in the ROW and China in our benchmark models, but are a small positive contributor in the Static model. For the trade slowdown in the ROW, all three models primarily point to the common shocks, but the static model attributes relatively more (88.0 percent vs 79.9 percent in the benchmark for China trade). The static model suggests that the shocks to trend trade cost growth have actually been a slight positive compared to the small drag we …nd in the benchmark model. Figure 12 compares the dynamics of the trend trade growth across three models. Expectations of trade costs declines are much lower in the benchmark model than the static model, although expectations for future integration remain at levels similar to those in the late 1990s in all three models. The benchmark model also suggests a much larger slow-down in the pace of future global trade cost declines than our alternative models. Figure 13 depicts the point estimate of a forecast of the change in the trade share in China and in these three models (top panels). All three models predict a decline in trade, with the static model providing the most optimistic projection suggesting only about a 10 percent decline in the trade share over the next 15 years (in China and the ROW) while the benchmark model predicts closer to a 16 percent decline. The gap between the models is very small in the …rst three years but then grows persistently. The static model predicts a more optimistic path for trade and less important role for trade shocks in net trade ‡ows than the benchmark model for the same reason. As the static model lacks any endogenous trade growth from a decline in trade barriers, the exogenous movements have to be more persistent making providing less incentive for borrowing and lending and less mean reversion in trade barriers. Elasticity of Substitution Most studies of trade integration or China’s growth are done in models with a higher Armington elasticity than those estimated, which is about 1.82 in our benchmark with a 90 percent con…dence interval of 1.52 and 2.08. We thus consider a version in which we raise the Armington elasticity to 3.5. The posterior mode of this variation is reported in the Column Benchmark High Armington. This version is a worse …t with a log Data Density of 203.76 compared to 220.81 in our Benchmark. Moreover, we …nd

29

that the trade cost shocks are considerably more persistent with much less of a phased-in component while the discount factor shock is considerably less persistent. Consequently, trade shocks become relatively less important in the accumulation of foreign assets. Shocks Eliminating the gap in trade barriers (No Gap) worsens the …t substantially with the log data density dropping to 176.3. It also substantially overstates the accumulation of assets in China and reduces the contribution of trade to 32 percent of the increase in foreign assets. This model predicts no noticeable changes in trade over the next 15 years. Eliminating the …xed cost shocks (No Fixed) lowers the log data density to 207.4. It primarily reduces the persistence of the trade shocks, suggesting a stronger role for intertemporal trade. This model generates the worst …t in terms of the change in China’s net foreign assets as it actually predicts a small decline. With less persistent trade shocks, trade is now expected to decline by 3 to 4 times as much as in the benchmark model. We also introduce a process whereby China catches up to the rest of the world. In particular, in a version we denote Di¤usion Growth, we allow a fraction

of the productivity

gap between China and the ROW to close each year zd0 = zd +

(zgap

zd ) + "zd

This model is a slightly worse …t with a log data density of 214.3. We estimate a di¤usion rate of 0.6 percent19 With permanent shocks, we …nd that trade shocks now account for only about half of the growth in assets. The forecast for trade growth is quite similar to our benchmark model. Capital Accumulation We have considered the role of anticipated and unanticipated changes in trade barriers on the dynamics of the trade balance. These shocks lead to borrowing and saving using a non-contingent bond. An alternative view is that most movements in the trade balance re‡ect ‡uctuations in capital investment, a margin that has not been 19

For comparison Rabanal (2011) estimate a quarterly di¤usion rate of 0.7 percent for advanced economies. The much smaller di¤usion rate here may arise because we do not distinguish between a size di¤erence from employment and productivity.

30

considered here explicitly20 . We now brie‡y consider the role of investment in capital for the trade balance empirically and then theoretically in a variation of the model with capital accumulation and a time varying investment subsidy. Figure 14 plots the nominal export-import ratio (i.e. the trade balance as a share of trade) and the investment rate as a share of GDP. The correlation between these two variables21 is only -0.09 for the whole period22 which suggests investment is not the full story. While the decline in the trade balance since the Great Recession does coincide with an increase in the investment rate, the large post-WTO trade surpluses leading to the Great Recession arises with no substantial change in the investment rate. Moreover, the large trade surpluses of the late 1990s, which are slightly larger than the pre-Great Recession surpluses, occur with a substantially smaller investment share. Next we consider the dynamics of the trade balance and real exchange rate in a variation of our model with capital accumulation and an investment subsidy that is rebated lump sum. We also eliminate pricing-to-market. The budget constraint now equals Pt (Ct + (1 where

x

x ) It )

+ Vt 1 +

Vt Bt 2 YtN b

Bt = Wt Lt + Bt

1

+

t

+ Tt

is a subsidy on investment …nanced by lump sum taxes, Tt : The law of motion for

capital is Kt = (1 where

) Kt

1

+ [1

(Kt ; Kt 1 )] It

(Kt ; Kt 1 ) is a quadratic adjustment cost on capital. The shock to the investment

subsidy is assumed to follow an AR(1) with persistence of 0.9. We also follow Alessandria and Choi (2007) and assume that each …rm chooses its capital stock in advance so that there is a distribution of capital stocks. Figure 15 plots the impulse responses of the real exchange rate and nominal trade balance as a share of gdp to each shock in the model with and without capital.23 For the model with 20

We say explicitly, since domestic absorption in the current model includes investment and consumption. If we use the nominal trade balance as a share of GDP the correlation is actually positive at 0.12. 22 Using the Chang (2015) data from 1980 to 2013 we …nd a positive correlation between the trade balance and investment rate of 0.36. In their data the persistently high trade surplues since 2000 are associated with a higher investment rate. 23 As before China’s trade share decreases with common trade costs and China productivity and is hardly 21

31

capital, we use the estimated parameters from the benchmark model.24 The shocks yield qualitatively similar dynamics in the model with and without capital. The top panel reports the response to an decrease in

. In both models, there is a

desire to borrow using the international bond and this leads to a persistent trade de…cit and appreciation. We also plot the response to an increase in the investment subsidy in the model with capital. We chose this shock to reduce the investment rate by 3 percentage points. On impact this shock has a similar impact on the trade balance and real exchange rate as the shock to beta. The investment subsidy shock though has a less persistent e¤ect on the trade balance and real exchange rate. The second row of …gures shows that an increase in global productivity has no noticeable impact on the real exchange rate or nominal trade balance in either model. The third row shows that an increase in China-speci…c productivity leads to a depreciation in both models, although with capital the initial impact is muted and the long-term impact is stronger. The di¤erent real exchange rate dynamics re‡ect the desire to increase Chinese investment initially and thus run a trade de…cit that turns into a longer run surplus. As Backus, Kehoe, and Kydland (1994) emphasize, capital accumulation substantially alters the e¤ect of a productivity shock on the trade balance. Much of the literature on China’s trade balance have proposed various mechanisms to undo this e¤ect (Song et al 2011). Here, we see that abstracting from capital accumulation is a parsimonious way to mute this channel. To the extent that these channels are orthogonal to our trade policy channel, this would not a¤ect our accounting of the contribution of trade policy to imbalances. Finally, changes in trade costs generate quite similar dynamics in the trade balance and real exchange rate with and without capital. In response to a di¤erential trade cost shock, both models generate an appreciation and a surplus. In response to a common increase in trade costs both models generate an appreciation but the model with capital generates a temporary surpluses followed by a sustained trade de…cit. a¤ected by the other shocks. 24 Aside from the parameters in the benchmark model, the capital share is = 0:36, the annual depreciation rate is = 0; and the adjustment cost is set to yield investment that is 2 times as volatil as output.

32

9. Conclusion China’s economic growth and integration with the world economy has been an important and de…ning economic event of the last twenty-…ve years. It has also been politically contentious, with much discussion of the link between the trade balance and trade policy and the e¤ect of China on employment and consumption outside of China. We study the source of this growth and integration through the lens of a two-country DSGE model with incomplete asset markets, heterogeneous producers, and endogenous trade participation featuring a dynamic decision and persistent changes in trade barriers, technology, and preferences. The model is used to identify shocks to trade, productivity, and preferences and then account for the contribution of these shocks on China and the Rest of the World. Our dynamic model of trade integration and growth allows for trade integration to in‡uence the trade balance and lead to changes in foreign assets. We …nd that trade integration has indeed been an important driver of ‡uctuations in China’s trade balance and accumulation of foreign assets, accounting for as much as 70 percent of its asset accumulation (as a share of GDP) through 2014. We also …nd that China’s growth and trade policy has substantially raised consumption outside of China and that the e¤ects on employment has actually been positive but relatively minor. Recent protectionist tendencies in China have been largely o¤set by its increase in spending. Our model can account for the dynamics of trade integration between China and the Rest of the World. By allowing for both a persistent shock to common trade costs as well as a persistent shock to the future growth rate of common trade costs, we can distinguish between the lagged e¤ects of past trade agreements, current shocks, and future shocks. We …nd that the growth in trade to GDP from the late 90s to the Great Recession re‡ected a combination of shocks with reductions in common and di¤erential trade barriers and a persistent decline in the rate of trade cost reductions. We …nd the Great Trade Collapse re‡ects a transitory rise in worldwide trade barriers but no persistent increase in the growth rate of trade barriers. Indeed, we …nd that most of the slow-down in trade since the Great Trade Recovery re‡ects the waning in‡uence of past trade agreements rather a substantial retrenchment in the pace of future trade agreements. Indeed, it is the period from the 90s to the Great Recession that is unusual and not the period since the Great Recession. Our benchmark model suggests 33

trade will continue to decline going forward. New trade agreements are needed to set the world economy o¤ on a new wave of trade integration and prosperity. We have remained agnostic about the source of di¤erent trade cost movements following much of the Gravity literature in measuring these as a residual. Having found these may have had important aggregate consequences, a useful next step would be to relate these to actual policy or technological changes. Finally, to focus on the role of trade barriers as a source of intertemporal trade we have largely abstracted from a number of key features studied elsewhere such as capital accumulation, sectoral heterogeneity, …scal policy, and …nancial constraints. Given that our main ideas about how trade policy matters for intertemporal trade relates to trade policy either having a transitory component or a¤ecting the ease in which a country can borrow and lend, it is likely that these insights will carry over to richer models. With respect to capital accumulation this certainly seems to be the case as preliminary work suggests changes in trade barriers have similar e¤ects in a model with capital. Of course, our quantitative results are likely to depend on the structure of the model and so should be read with caution - at least with as much caution as results based on one good models common in the literature. We are currently pursuing this sort of quanti…cation.

10. Data Time period 1990 to 2014. China’s relative size is measured using constant 2005$, World Development Indicators WDI The real exchange rate (OECD). The ratio of nominal exports to nominal imports (China Statistical Bureau). The share of trade in real gdp (X+M)/GDP (China Statistical Bureau). Chinese manufacturing export participation from Chinese Annual Survey of Enterprises 34

(China Statistical Bureau). China’s net foreign assets - 1990 - 2014 (World Development Indicators). US Export Participation to China - US Exporters from US Exporter Pro…le (1992, 1996 to 2014) scaled by US …rms with 20+ employees from Small Business Administration (1988 - 2011). Share of exporters interpolated for 1993 to 1995 period. Number of …rms from 2012-2014 is assumed to grow at a constant rate. China’s Import Tari¤s - World Development Indicators, Mean, manufactured products (%) and Weighted mean, manufactured products (%). Investment rate - Gross Fixed Capital Formation percent of GDP (China Statistical Bureau).

References Alessandria, G. and H. Choi, 2007. “Do Sunk Costs of Exporting Matter for Net Export Dynamics?”Quarterly Journal of Economics, 122(1), 289-336. — — , 2014a. “Establishment Heterogeneity, Exporter Dynamics, and the E¤ect of Trade Liberalization.”Journal of International Economics. — — , 2014b. “Do Falling Iceberg Costs Account for Recent US Export Growth?”Journal of International Economics. — — , 2015. “The Dynamics of the U.S. Trade Balance and the Real Exchange Rate: The J Curve and Trade Costs?”mimeo. — — , and Ruhl, 2014. “Trade Adjustment Dynamics and the Welfare Gains from Trade” NBER WP 20663. Alessandria, G., J. Kaboski, and V. Midrigan, 2010. “The Great Trade Collapse of 2008-09: An Inventory Adjustment?, IMF Economic Review. — — , 2011, “U.S. Trade and Inventory Dynamics,”American Economic Review. — — , 2013. “Trade Wedges, Inventories, and International Business Cycles,” Journal of Monetary Economics, 60(1), 1-20. Alessandria, G., S. Pratap and V. Yue. 2011. “Export Dynamics in Large Devaluations,” mimeo. Autor, D., D. Dorn, and G. Hanson, 2013, “China Syndrome: Local Labor Market E¤ects of Import Competition in the United States,”American Economic Review 103 (6), 2121-2168. — — , 2015, “Untangling Trade and Technology: Evidence from Local Labor Markets,”The Economic Journal. 35

Backus, David, Patrick Kehoe, and Finn Kydland. 1994. “Dynamics of the Trade Balance and the Terms of Trade: The J Curve?”American Economic Review, LXXXIV, 84–103. Baldwin, R and P Krugman, 1989. “Persistent Trade E¤ects of Large Exchange Rate Shocks,”Quarterly Journal of Economics, 104(4), 821-854. Barattieri, Alessandro, 2014. “Comparative Advantage, Service Trade, and Global Imbalances,”Journal of International Economics, Vol. 92, 2014, pages 1-13. Bown, Chad P. 2010. “China’s WTO Entry: Antidumping, Safeguards, and Dispute Settlement,” Chapter in NBER book China’s Growing Role in World Trade, Robert C. Feenstra and Shang-Jin Wei, editors (281-337). Buera, F. and Y. Shin, 2009, Productivity growth and capital ‡ows: The dynamics of reforms, Technical report, National Bureau of Economic Research. Brambilla, I, A. Khandelwal, P. K. Schott. 2010. “China’s Experience under the Multi-Fiber Arrangement (MFA) and the Agreement on Textiles and Clothing (ATC)”Chapter in NBER book China’s Growing Role in World Trade (2010), Robert C. Feenstra and Shang-Jin Wei, editors (p. 345-87). Caballero, R.J., E. Farhi, and P. Gourinchas. 2008. “An Equilibrium Model of “Global Imbalances”and Low Interest Rates.”American Economic Review 98:1, 358–393. Chang, C., K. Chen, D. Waggoner, and T. Zha. 2015. “Trends and Cycles in China’s Macroeconomy,”NBER Working Paper 21244. Chari, V.V., P.J. Kehoe, and E.R. McGrattan. 2002. “Can Sticky Price Models Generate Volatile and Persistent Real Exchange Rates?”Review of Economic Studies, 69(3): 533-63. Chari, V.V., P.J. Kehoe, and E.R. McGrattan. 2007. “Business Cells Accounting,” Econometrica 75 (3) 781-836. Choi, H., N. Mark, and D. Sul. 2008. “Endogenous Discounting, the World Saving Glut and the U.S. Current Account,”Journal of International Economics, 75, 2008, 30-53. Christiano, L. M. Eichenabum, and S. Rebelo. 2011. “When Is the Government Spending Multiplier Large?,”Journal of Political Economy, 119 (1), 78 - 121 Coeurdacier, N. Guibaud, S. and K. Jin. 2015. “Credit Constraints and Growth in a Global Economy,”American Economic Review, 105(9): 2838–2881 Das, Sanghamitra, Mark Roberts, and James Tybout, 2007. “Market Entry Costs, Producer Heterogeneity, and Export Dynamics,”Econometrica, 75(3), 837-873. Dekle, R., Eaton, J., S. Kortum, 2008. “Global Rebalancing with Gravity: Measuring the Burden of Adjustment,”IMF Sta¤ Papers, 55(3), 511-40. di Giovanni, J., A. Levchenko, and J. Zhang, “The Global Welfare Impact of China: Trade Integration and Technological Change,”American Economic Journal: Macroeconomics. 36

Dixit, Avinash K., “Hysteresis, Import Penetration, and Exchange Rate Pass-Through,” Quarterly Journal of Economics, CIV (1989a), 205–228. — — , “Entry and Exit Decisions Under Uncertainty,”Journal of Political Economy, XCVII (1989b), 620–638. Drozd, L. and J. Nosal, 2012. “Understanding International Prices: Customers as Capital,” American Economic Review, Vol. 102 (1). Eaton, Kortum, Neiman and Romalis, 2014. “Trade and the Global Recession.” Eggertsson, Gauti, and Michael Woodford. 2003. "The Zero Bound on Interest Rates and Optimal Monetary Policy." Brookings Papers on Economic Activity 1:139-211. Gallaway, M., C. McDaniel, S. Rivera. 2003. “Short-run and long-run industry-level estimates of U.S. Armington elasticities”North American Journal of Economics and Finance 14, 49– 68. Handley, Kyle and Nuno Limao, 2017. “Policy Uncertainty, Trade and Welfare: Theory and Evidence for China and the U.S.”American Economic Review. Hooper, P., K. Johnson and J. Marquez, 2000. “Trade Elasticities for G-7 Countries,”Princeton Studies in International Economics No. 87 (Princeton University Press: Princeton NJ). Ju, Jiandong, Kang Shi, Shang-Jin Wei. 2012. “Trade Reforms and Current Account Imbalances: When Does the General Equilibrium E¤ect Overturn a Partial Equilibrium Intuition?” NBER Working Paper No. 18653. Junz, H. and R. Rhomberg, 1973. “Price Competitiveness in Export Trade Among Industrial Countries.”American Economic Review, 63(2), 412-418. Jacks, D., C. Meissner, and D. Novy. 2011. Trade Booms, Busts and Trade Costs. Journal of International Economics. Khandelwal, Amit K., Peter K. Schott, and Shang-Jin Wei. 2013. "Trade Liberalization and Embedded Institutional Reform: Evidence from Chinese Exporters." American Economic Review, 103(6): 2169-95. Kehoe, T., K. Ruhl and J. Steinberg. 2014. “Global Imbalances and Structural Change in the United States”NBER Working paper 19339. Lane, P. and G. Milesi-Ferretti. 2007. "The external wealth of nations mark II: Revised and extended estimates of foreign assets and liabilities, 1970–2004", Journal of International Economics 73, November, 223-250. Magee, Steven, 1973. “Currency Contracts, Pass-through and Devaluations,”Brookings Papers on Economic Activity. 1973(1), 303-325. Meade, Ellen, 1988. “Exchange Rates, Adjustment, and the J-Curve.”Federal Reserve Bulletin, 74(10): 633-644. 37

Obstfeld, M. and K. Rogo¤. 2005. “Global Current Account Imbalances and Exchange Rate Adjustments.”Brookings Papers on Economic Activity 1: 67-146. Pierce, J. and P. Schott, 2016. “The Surprisingly Swift Decline of U.S. Manufacturing Employment,”American Economic Review 106(7): 1632–1662. Quadrini, V., E. Mendoza, and V. Rios-Rull, 2009, Financial integration, …nancial deepness and global imbalances, Journal of Political Economy 117. Rabanal, P., Rubio-Ramírez, J., Tuesta, V., 2011. Cointegrated TFP processes and international business cycles. Journal of Monetary Economics 58, 156–171. Reyes-Heroles, R., 2016. “The Role of Trade Costs in the Surge of Trade Imbalances,”mimeo. Roberts, M. and J. Tybout, 1997. “The Decision to Export in Colombia: An Empirical Model of Entry with Sunk Costs,”American Economic Review, 87(4), 545-564. Stockman, A. and L. Tesar. 1995. “Tastes and Technology in a Two-Country Model of the Business Cycle: Explaining International Comovements,”American Economic Review, Vol. 85, (1) pp. 168-185 Song, Z., K. Storesletten, and F. Zillibotti, 2011, “Growing Like China,”American Economic Review 101, 196-233. Tombe, T and X. Zhu, 2015, “Trade, Migration and Productivity: A Quantitative Analysis of China,”Working Paper. Waugh, M. 2010. “International Trade and Income Di¤erences,”American Economic Review, 100(5): 2093-2124.

38

Table 1: Empirical Decomposition of trade balance to gdp var(TB/Yjtry)/var(TB/Y) % Real Nominal Whole Sample 45.2 34.9 Since 2000 24.5 18.9

39

Table 2: Parameters Fixed Parameters a1 0.96 0.0001 0.30 0.16 5 b

Estimated Parameters

zd zc c d

b g

f zd zc c d g

f0 f1 b

q Yw

z c d

&q &y f0 f1

prior mean 0.975 0.75 0.5 0.5 0.5 0.5

posterior 90% HPD mean mode 0.9977 0.9993 0.995 0.7807 0.7379 0.6058 0.9312 0.9617 0.8617 0.9846 0.9871 0.9684 0.9341 0.9739 0.8452 0.8417 0.9745 0.6373

0.5 0.7895 0.7547 0.5925 0.07 0.0692 0.0664 0.052 0.033 0.0361 0.0332 0.0279 0.2 0.1568 0.1567 0.1218 0.124 0.1399 0.1326 0.1051 0.016 0.0418 0.0113 0.0055 0.01 0.0076 0.0047 0.0025 0.22 0.2208 0.2188 0.205 0.005 0.0057 0.0033 0.0019 -1 -1.0459 -1.0518 -1.3639 -1.335 -1.3198 -1.3273 -1.3958 2 1.7975 1.8167 1.5194 5 4.7777 4.4192 3.3591 2.42 2.338 2.3449 2.1749 0.5 0.5031 0.5045 0.4251 0.1 0.1132 0.1001 -0.0379 0 0.0364 0.0526 -0.1515 0 -0.0203 -0.0275 -0.113 0.37 0.3786 0.3681 0.2969 0.04 0.0422 0.0428 0.0348 0.235 0.199 0.186 0.1508

Interval

prior

0.9999 0.9559 0.9996 0.9999 0.9986 0.9975

unif unif unif unif unif unif

prior std.dev. 0.0144 0.1443 0.2886 0.2886 0.2886 0.2886

0.9939 0.086 0.0432 0.1895 0.1742 0.0912 0.0136 0.2376 0.0109 -0.6777 -1.2479 2.0776 5.9876 2.4864 0.5742 0.2631 0.2441 0.0888 0.4545 0.0498 0.2434

unif invg invg invg invg invg invg invg invg norm norm invg invg norm norm norm norm norm invg invg invg

0.2886 0.025 0.025 0.05 0.05 0.02 0.05 0.01 0.01 0.25 0.2 1 1 0.1 0.05 0.1 0.15 0.15 0.05 0.005 0.05

Notes: Based on annual data from 1990 to 2014. 40

Table 3: Source of Changes from 1990 to 2014

Output Ratio * Cons. ratio China nom. Trade China real Trade ROW nom. Trade Common trade cost Di¤. Trade Cost Fixed cost (f1) ROW Consumption Labor Production Labor

Change (log)

Share of Change (%) Zc

Zd

c

d

0.139

0

1.00

0

0.02

1.519

0

0.88

0.06

0.301

0

-0.33

0.479

0

0.221

g

f1

Initial

-0.01

0

0

-0.01

-0.07

0.01

0.05

0

0.07

0.66

-0.29

-0.01

0.52

-0.01

0.45

-0.46

0.55

0.03

0.02

0.43

-0.01

0.44

0

0.18

0.28

0.12

0

0.22

-0.01

0.21

-1.872

0

0

0.40

0

0

0.31

0

0.28

1.118

0

0

0

0.98

0

0

0

0.02

0.002

0

0

0

0

0

0

0.64

0.36

0.143

-0.12

0.18

0.27

0.18

-0.02

0.21

-0.01

0.32

0.017

0

0.09

0.22

-0.37

0.24

0.19

0.01

0.62

-0.010

0

0.20

0.38

0.84

-0.34

0.30

-0.03

-0.36

Absolute Changes

Output Productivity (Zc)

0.000

-0.0175

-0.0023

-0.0029

-0.0077

0.0035

-0.0023

0.0002

0.029

0.007

-0.018

0

0

0

0

0

0

0.025

* absolute changes.

41

Table 4: Source of Change in China’s Assets-GDP (%)

Data All Initial (I) Prod.-Pref. (P) Trade (T) IxP IxT PxT IxPxT IP IT PT

Benchmark 36.2 54.9 40.2 3.5 28.3 -28.4 25.3 0.0 -13.9 15.2 93.8 31.7

Share from trade*

72.3

No High PTM Static Armington No Gap No Fixed Di¤usion 36.2 36.2 36.2 36.2 36 36.2 66.7 71.6 71.8 83.9 -4 76.6 1.0 1.4 39.2 22.9 190 -2.0 0.8 1.7 1.2 3.6 6 -0.8 37.4 17.6 13.4 -6.0 53 24.4 22.3 30.2 -13.9 30.7 -245 40.9 33.6 40.9 49.5 28.7 33 38.2 -8.4 3.1 9.9 12.3 -12 -7.1 -20.1 -23.4 -27.5 -8.3 -29 -17.2 24.1 33.3 26.5 57.2 -49 38.2 72.0 59.8 102.1 45.6 276 60.7 29.9 22.4 24.5 9.9 47 16.6 63.9

53.5

63.1

31.8

-1094

50.2

* Calculated as (1- IP/All). Data from WDI. P is shocks to (Zc ; Zd ; ); & T denotes shocks to ( c;

d; g ;

0;

1 ):

Cross (x) is interaction e¤ects while IP, IT, PT is impact of two together

42

Table 5: Source of Change in Trade-GDP (1990 to 2014, %) China China Real Nominal 0.1 0.0 -46.3 -32.9 2.5 -0.5

Productivity-Global Productivity-China Discount Factor Trade Cost Common Di¤erence Trend Fixed-enter Fixed-continue Initial Total

55.5 2.7 43.0 0.0 -1.3 43.9 47.9

66.6 -29.3 51.5 0.0 -0.7 45.4 30.2

ROW Nominal 0.0 17.6 0.5 27.8 12.6 21.4 0.0 -0.5 20.8 22.1

Each entry measures the share of the change in trade from 1990 to 2014 from that shock alone.

Change in the Source of the Change in Trade-GDP (2011-2014 vs 1997 to 2007) China China ROW Real Nominal Nominal Productivity-Global 0.7 0.2 0.0 Productivity-China 10.5 5.1 -12.7 Discount Factor 5.9 -2.8 6.3 Trade Cost Common 79.9 74.1 124.1 Di¤erence -5.3 15.5 -29.1 Trend 6.1 3.9 7.6 Fixed-enter 0.0 0.0 0.0 Fixed-continue -6.6 -2.8 -8.9 Initial 9.0 6.7 11.4 Total -4.6 -4.3 -0.8 Each entry measures the share of the change in trade growth accounted for by each shock.

43

Table 6: Posterior Mode of Parameters - Alternative Models

zd zc c d

b g

f zd zc c d g

f0 f1 b

q Yw

z c d

&q &y f0 f1

Log Data Density

Benchmark 0.9993 0.7379 0.9617 0.9871 0.9739 0.9745

No PTM 0.9995 0.7349 0.9581 0.9869 0.9556 0.9696

Static 0.9994 0.7289 0.9763 0.9883 0.9731 0.979

0.7547 0.8494 0.9836 0.0664 0.0662 0.0664 0.0332 0.0332 0.0333 0.1567 0.1628 0.1695 0.1326 0.1559 0.1437 0.0113 0.0107 0.0115 0.0047 0.0047 0.2188 0.2194 0.2184 0.0033 0.0038 0.003 -1.0518 -1.0045 -0.9968 -1.3273 -1.3287 -1.3899 1.8167 1.8536 1.7564 4.4192 4.4147 4.5284 2.3449 2.3431 2.3482 0.5045 0.5023 0.5019 0.1001 0.1 0.1177 0.0526 -0.0275 0.3681 0.3706 0.0428 0.0439 0.0848 0.186 0.1838 0.0398 220.8 219.5 219.3

44

High Armington Di¤usion 0.9999 0.7224 0.7334 0.9942 0.9809 0.9989 0.9949 0.232 0.9603 0.5022 0.9733 0.9882 0.0732 0.0331 0.135 0.1068 0.0099 0.0047 0.227 0.0031 -1.0309 -1.2209 3.5 4.6276 2.4579 0.4391 0.1849 0.006 0.0362 0.3425 0.0382 0.293 203.8

No …xed 0.9991 0.7561 0.8594 0.9881 0.9792 0.8802

No Gap 0.9996 0.7345 0.9827 0.9471 0.9807

0.9851 0.6194 0.9949 0.0669 0.0692 0.0332 0.0332 0.0665 0.1543 0.1677 0.0331 0.1261 0.1238 0.1515 0.01 0.104 0.0108 0.0047 0.3199 0.2197 0.2245 0.003 0.0027 0.0027 -0.9739 -1.2703 -0.9436 -1.2859 -1.3094 -1.2368 2.0809 1.7798 2.6258 4.4712 4.2385 4.4943 2.3841 2.3883 2.3604 0.5028 0.4917 0.4914 0.1155 0.1073 0.1041 0.065 0.1816 0.7065 -0.0435 0.0992 0.2174 0.352 0.3793 0.3711 0.0402 0.0388 0.0407 0.2343 0.2087 0.2797 0.0003 214.3 207.4 176.3

Table 7: Source of Change in Trade-GDP (1990 to 2014) China Real Trade share Benchmark No PTM Static Productivity-Global 0.1 0.0 0.0 Productivity-China -46.3 -45.8 -44.1 Discount Factor 2.5 3.4 2.9 Trade Cost Common 55.5 59.5 53.8 Di¤erence 2.7 14.2 5.0 Trend 43.0 35.0 39.3 Fixed-enter 0.0 0.0 0.0 Fixed-continue -1.3 -3.3 3.1 Initial 43.9 36.9 40.1 Total 47.9 47.9 47.9

ROW Nominal Trade Share Benchmark No PTM Static 0.0 0.0 0.0 17.6 18.0 15.0 0.5 0.6 0.5 27.8 12.6 21.4 0.0 -0.5 20.8 22.1

31.7 13.9 18.7 0.0 -1.4 18.5 20.7

29.3 10.8 21.3 0.0 1.3 21.7 21.9

Each entry measures the share of the change in trade from 1990 to 2014 from that shock alone.

Change in the Source of the Change in Trade-GDP (2011-2014 vs 1997 to 2007) China Real Trade share ROW Nominal Trade Share Benchmark No PTM Static Benchmark No PTM Static Productivity-Global 0.7 0.7 0.0 0.0 0.0 0.0 Productivity-China 10.5 10.9 11.4 -12.7 -10.7 -10.4 Discount Factor 5.9 8.8 7.4 6.3 4.8 3.9 Trade Cost Common 79.9 88.0 88.0 124.1 127.4 132.5 Di¤erence -5.3 -14.4 -5.7 -29.1 -28.6 -23.4 Trend 6.1 5.5 -0.2 7.6 6.0 -2.6 Fixed-enter 0.0 0.0 0.0 0.0 0.0 0.0 Fixed-continue -6.6 -7.4 -7.7 -8.9 -8.3 -9.1 Initial 9.0 8.1 6.8 11.4 9.5 10.4 Total -4.6 -4.6 -4.6 -0.79 -0.84 -0.77 Each entry measures the share of the change in trade growth accounted for by each shock.

45

b. Trade share of GDP .8

.4

a. China Relative Size (real GDP) ROW

Real Nominal

0

.2

.1

.4

.2

.6

.3

US

1990

1995

2000

2005

2010

2015

1990

TB/Y

2005

2010

2015

2010

2015

TB/Y

0 -.05

-.05

0

.05

TB/Y (pre-WTO trade)

.05

TB/Y (pre-WTO trade)

1990

2000

d. Nominal Trade Balance .1

.1

c. Real Trade Balance

1995

1995

2000

2005

2010

2015

1990

1995

2000

Figure 1: Aggregate Dynamics in China

46

2005

1990

1995

2000

2005

11 11.5 12 12.5 13 Log No.

Log No. of Cty 5.28 5.35.325.345.36

.2

Exports of China .05 .1 .15 Share of value

Share of varieties .4 .5 .6 .7 .8

China's share in US imports

2010

1990

1995

2000

2005

2010

Varieties share in US imports

Countries

Value share in US import

Cty *HS6

US Exporters to China

.24

.26

.28

.3

.32

.01 .02 .03 .04 .05 .06

Export Participation

2015

1995

2000

2005

2010

1990

1995

2000

2005

2010

2015

2010

2015

Figure 2: Disaggregate Trade Dynamics in China China import tariff

2

5

4

6

8

10 15 20 25 30

10

US tariff on Chinese imports

1990

1995

2000

2005

2010

2015

1995

2000

2005

China MFA quota to US fill rate

0

0

1

2

3

4

5 10 15 20

Tariff HS 8 digit

1990

0

.2

.4

.6

.8

1

0

.2

.4

NTR rate

y ear 1990

Non-NTR rate

y ear 2004

kernel = epanechnikov, bandwidth = 0.0040

.6

kernel = epanechnikov, bandwidth = 0.0580

Figure 3: Trade Barriers: China and US 47

.8

1

y ear 2000

0.2

Relative GDP

1.5

0.1

1

0

0.5

1

Real Trade Share

0.5

0.5

0

0

-0.5

0.1

World GDP

0.4

0

Nominal Export/Import

Real Exchange Rate

Chinese Exporters

0.2

-0.1 1990 1995 2000 2005 2010

0 1990 1995 2000 2005 2010

Figure 4: Historical and Smoothed Series

48

Z 2

Z

c

3 2 1 0 -1

0 -2

2

d

c

2 0 -2

d

1

g

2 0

0

0 -2 -2

0.02

f0

0 -0.02 1990

2000

2010

3 2 1 0 -1 1990

-1 1990

f1

2000

2010

Figure 5: Estimated Shocks

49

2000

2010

Figure 6: Deviations from Steady State of Exogenous State Variables 1

Z (China)

0.15

Z (ROW)

0.02

0.1

0.5

Beta (China)

0.015

0.05 0

0.01 0

-0.5 -1 1990

1

0.005

-0.05 2000

2010

Iceberg Costs

-0.1 1990

0

2000

2010

Trend-Iceberg

0.8 0.6

0

-0.02

2000

2010

Fixed-Costs Entry Continue

0.4

-1 -0.04

0.2

-2

0

-3 -4 1990

0 1990

To China FromChina

2000

2010

-0.06 -0.2 -0.08 1990

2000

:

50

2010

-0.4 1990

2000

2010

a. Real GDP

0.1

b. World GDP

0.04

0.05

0.02

0

0

-0.05

-0.02 0

5

10

15

20

25

0

c. Nominal Export-Import Ratio

5

10

15

20

25

15

20

25

d. RER

0.1

0.05

0.05

0

0

-0.05

-0.05

-0.1 0

5

10

15

20

25

0

5

10

e. China Trade Share of GDP 0.02 Z

0

Z

d c

-0.02 c

-0.04

d

-0.06

g

0

5

10

15

20

25

Figure 7: Impulse Response Benchmark

51

China Exporters (%) 0.01 Z Z

ROW Exporters (%)

10 -3

3 d

2

c

0.005 c

1

d g

f

0

f

0

0 1

-1 -0.005 -2

-0.01

-3

-4 -0.015 -5

-0.02

-6

-7 -0.025 0

5

10

15

20

25

0

5

10

15

Figure 8: Exporter Dynamics: Impulse Response

52

20

25

Figure 9: Decomposition of Aggregate Dynamics

53

Figure 10: Decomposition of Trade Dynamics

54

TB/Y (Nominal) 0.1

Data

0.05

All

0

-0.05 1990

1995

2000

2005

2010

Ne t Fore ign Asse ts/GDP 1

0.5

0

-0.5 1990

1995

2000

2005

2010

Figure 11: Decomposition of TB/Y and Chane in Net Foreign Assets

55

Iceberg cost - Trend growth 0

-0.01

-0.02

-0.03

-0.04

-0.05

-0.06

-0.07

-0.08 1990

Static Dynamic - No PTM Benchmark 1995

2000

2005

Figure 12: Trend Trade Cost

56

2010

2015

China trade share

ROW trade share

0

0

-5

-5

-10

-10

-15

-15

Benchmark Static No-PTM

-20 2015

2020

-20 2015

2020

2025

2030

0

2030

0 -10

-20

-20 -40 -30 -60 2015

2025

2020

2025

-40 2015

2030

Benchmark High Armington No Gap No fixed Diffusion 2020

Figure 13: Forecasted Trade Share

57

2025

2030

30

50 35

-10

Investment Rate 40 45

0 10 20 Nominal Export-Import Ratio

I/Y EX/M

1990

1995

2000

2005

2010

2015

Note: Correlation = 0.088

Figure 14: Investment Rate and Net Trade Flows Shock

0.02

Real Exchange Rate

0.05

0

Capital No capital Capital subsidy

0

/5

x

-0.02

0

-0.05 0

20

40

60

0.04 Z

0.02

c

0

-0.2 0

20

40

60

0.05

0.1

0

0

-0.05 0

20

40

0.05

60 Z

d

0 20

40

20

40

60

0.05

0.2

0

0

60

0.2 c

0.1 0 20

40

20

40

60

0.02

0.05

0

0

60

0 d

-0.1 -0.2 20

40

60

40

60

0

20

40

60

0

20

40

60

0

20

40

60

0

20

40

60

-0.05 0

20

40

60

0

0.2

-0.05

0

-0.1 0

20

-0.2 0

-0.02 0

0

-0.1 0

-0.05 0

Trade Balance

0.2

-0.2 0

20

40

60

Figure 15: Impulse Response with and without Capital 58

Trade Integration and the Trade Balance in China

changes in technology, trade costs, and preferences accounting for the dynamics of China's gross and net trade ... Keywords: Trade Integration, Trade Balance, Real Exchange Rate, International Business. Cycles, Net ... models have been shown to best explain producer level export participation as well as the dynamics of ...

542KB Sizes 3 Downloads 305 Views

Recommend Documents

Aid and Regional Trade Integration
4.1 Macroeconomic and trade data . ..... Thirdly, the above analysis assumes aid is an integral long-run part of a country's long-run ..... environment where the income elasticity of demand for agricultural goods is less than unity and with a.

Trade Integration and the Political Support for Labor ...
In this paper I will study the effect of trade integration on the political support for employment protection, focussing in particular on the applicability of the Ricardian argument outlined above. Employment protection is inherently a policy designe

Trade Booms, Trade Busts, and Trade Costs
measure of trade frictions from leading trade theories and use it to gauge the ... regardless of the motivation behind international trade, be it international product ...

Trade for Trade - NSE
Feb 1, 2018 - 20, 2012 – Sundaram Finance Limited (SUNDARMFIN) shall be transferred from trade for trade segment (series: BE) to rolling segment (series: ...

Trade flows and trade disputes
Nov 9, 2014 - Springer Science+Business Media New York 2014 ..... We also classify as export policy disputes the much smaller number of cases over ..... 800-. 850. 850-. 900. 900-. 950. 950-. 1000. >1000. Imports (millions of $2005).

Trade Flows and Trade Disputes - Semantic Scholar
10 Jul 2014 - Kara M. Reynolds. ‡. American University. This version: July 2014. Abstract. This paper introduces a new data set and establishes a set of basic facts and patterns regarding the. 'trade' that countries fight about under WTO dispute se

Gains from Financial Integration and Trade Openness ...
and Foreign countries can freely trade goods, but individuals are not able to invest abroad. ... Section four presents the quantitative results for the open economy: first under ... Moreover, capital account openness is associated with higher output

Announcement on China Lanzhou Investment and Trade Fair.pdf ...
Page 2 of 2. Page 2 of 2. Announcement on China Lanzhou Investment and Trade Fair.pdf. Announcement on China Lanzhou Investment and Trade Fair.pdf. Open. Extract. Open with. Sign In. Main menu. Displaying Announcement on China Lanzhou Investment and

Commodity Trade and the Carry Trade - University of Chicago
Jan 29, 2014 - Simon School of Business, University of Rochester. ‡. The Wharton ...... the forward premium puzzle, Working Paper Harvard University. Ferraro ...

Trade in Services and Trade in Goods: Differences and ...
trade in “Other Commercial Services”, which consists mainly of business-to- business .... 6There are notable exceptions to the non-storability criterion, e.g. computer softwares, .... This variable accounts for rigidities .... relative to small o

Trade in Services: Cross-Border Trade vs. Commercial ...
and cross border trade in services contradict this assumption. Not only is commercial ... investment. For the empirical analysis we make use of bilateral data .... large Japanese electronic and automobile assemblers, which are not verti-.

trade and productivity - CREI
Our empirical approach accounts for the endogeneity of trade and ... of openness as a measure of trade can be illustrated using a small open economies.

trade and productivity - CREI
institutional quality as well as geography and take into account the ... financial support from the CREI and CREA research institutes, the European Fund for .... Purchasing power parity (PPP) GDP differs from GDP in that the production of each good i

Trade for Trade - Lloyds Steels Industries Limited
Jul 18, 2016 - BE) to rolling segment (series: EQ) with effect from August 01, 2016 (Monday). For and on behalf of. National Stock Exchange of India Limited.

1499499676064-skyplay-untrammelled-forex-trade-skies-trade ...
... Trade Software. Without Doubt Identifies PatternAnd Trades Them. Page 2 of 2. 1499499676064-skyplay-untrammelled-forex-trade-skies-trade-system-ea.pdf.

Immigration, Trade and Wages in Germany
Editors welcome the submission of manuscripts both in electronic (E-mail ... Publisher: Adrian Bodnaru Cover Design: Dan Ursachi Layout: Dragoş Croitoru ...

Does trade integration alter monetary policy ...
Sep 13, 2010 - There is, however, a secular trend in trade integration, suggesting that ...... if somewhat high, is still consistent with evidence reported by micro ...

Immigration, Trade and Wages in Germany
manuscript should be under Microsoft Word. ... with postal code must be given at the bottom of the title page, together with Phone/Fax numbers and ... A New. Perspective on Status Inconsistency | 153. Monica ROMAN and Christina SUCIU, ...

The Atlantic Slave Trade
trade (inset map). Most slaves did not tell about their passage—many died, and those who survived faced lives of bondage in the Americas. But some. Europeans .... AMERICA. AFRICA. EUROPE. Atlantic. Ocean. Pacific Ocean. New York. Charleston. Bahia.

Does trade integration alter monetary policy ...
Sep 13, 2010 - University of Bonn and CEPR ... monetary policy transmission, open economy, trade integration, exchange rate .... are used to produce a unit of the wholesale good—thereby determining the degree of openness. ..... Letting St denote th