one page summary of the article

China Economic Review 27 (2013) 227–237

Contents lists available at ScienceDirect

China Economic Review

The political economy of “currency manipulation” bashing☆

Carlos D. RAMIREZ Department of Economics, George Mason University, Fairfax, VA 22030, United States

a r t i c l e i n f o

☆ This paper has benefited from comments and sugg well as participants at the 2011 Columbia–Tsinghua C from an anonymous referee as well as suggestions fro

1043-951X/$ – see front matter © 2012 Elsevier Inc. A http://dx.doi.org/10.1016/j.chieco.2012.10.005

a b s t r a c t

Article history: Received 29 September 2011 Received in revised form 22 October 2012 Accepted 25 October 2012 Available online 31 October 2012

In recent years, one of the most frequently debated issues in Congress has been the value of the Chinese renminbi (RMB) relative to the U.S. dollar. Many members of Congress often accuse China of being a “currency manipulator.” This paper has two objectives. First, it investigates the extent to which PAC contributions from key interest groups as well as constituent interests influence the frequency with which members of Congress criticize China's exchange rate policy, controlling for other factors. The results indicate that the odds that a congressman will call China a “currency manipulator” are 1.35 times higher for every $5000 in PAC contributions from groups that favor legislation against China. In addition, the results show that a one percentage point increase in the share of the congressional district labor force in manufacturing is associated with a 19.6% increase in the likelihood that the district's legislator will label China a “currency manipulator.” Second, this paper investigates the consequences that “currency manipulation” bashing may have on the rate at which the RMB appreciates against the U.S. dollar. The results for a VAR model indicate that an increase in the incidence of “currency manipulation” bashing appears to temporarily slow down, rather than accelerate, the rate at which the renminbi appreciates against the dollar. This result suggests that bashing China may actually be counterproductive.

© 2012 Elsevier Inc. All rights reserved.

JEL classification: F59 D72

Keywords: Currency manipulation China bashing PAC contributions Renminbi

1. Introduction

The value of the renminbi (RMB) relative to the U.S. dollar and how it influences the US–China trade balance has been one of the most salient and debated issues in U.S. Congress and the administration over the last few years. Many members of Congress lay blame on China's exchange rate policy for explaining the decline in U.S. competitiveness as well as the loss of “American jobs” to places overseas, and in particular, to China. Since 2003, this dispute has resulted in a rise congressional activity manifested in the introduction of several bills presumably aiming at rectifying the apparent inequities caused by China's currency policy stance (Xie, 2010). On September 29, 2010, the House of Representatives passed by a wide margin the “Currency Reform for Fair Trade Act” (H.R. 2378), a bill that would permit the imposition of countervailing duties on countries found to have an “undervalued currency.”

Concurrent with the rise in China-related congressional legislation there has been a significant increase in verbal criticisms of China's exchange rate policy. When expressing these criticisms, legislators typically voice the particular grievance keywords “currency manipulation.” For example, in March 2010, 130 congressmen signed a petition to Timothy Geithner, Secretary of the Treasury, urging him to officially classify China as a “currency manipulator.” Such classification would then permit Congress to consider adopting punitive action against China (Staiger & Sykes, 2010).

estions from participants at the June 2011 Chinese Economists Society (CES) Conference in Beijing, China as onference on International Economics at Tsinghua University. I would like acknowledge helpful comments m Jiandong Ju and Shang-Jin Wei. Of course, they are not responsible for any errors or omissions.

ll rights reserved.

228 C.D. Ramirez / China Economic Review 27 (2013) 227–237

This paper is not about questioning whether the RMB is or is not undervalued relative to the dollar. A plethora of papers have already been written on this issue.1 But China is not the only country in the world that may have had an undervalued currency. If fact, using IMF figures for the implied Purchasing Power Parity and actual exchange rates for 2009, the degree of the RMB undervaluation relative to the U.S. dollar (approximately 43%) was within one standard deviation away from the mean undervaluation among a random sample of 45 currencies (see Fig. 1). The fact that China tends to get singled out in Congress for its currency stance raises suspicion that perhaps “currency manipulation” bashing has less to do with the degree of undervaluation and more to do with the influence interest groups exert in Congress. Thus, one of the two objectives of this paper is to answer the question: To what extent do interest groups influence legislators to bash China for its exchange rate management policy?

The second objective of this paper is to understand the consequences of such bashing on the rate at which the RMB appreciates against the U.S. dollar. Both objectives reinforce each other, which is why they are analyzed in one paper. Knowing why bashing occurs is important because of the consequences that it may have.

To answer the first question posed above I use a standard political economy model that explains the frequency with which members of Congress bash China through “currency manipulation” criticisms in the 111th Congress (2009–2010). Specifically, I count the number of times each congressmen voiced the particular keywords “currency manipulation” in the context of China. I then fit a regression of this count on the amount of money congressmen received from interest groups that were in favor of enacting the “Currency Reform for Fair Trade” bill (H.R. 2378), controlling for other factors known to influence legislative voting behavior, such as ideology, party affiliation, constituent interests, and how they voted on H.R. 2378.

The results of this first test are consistent with expectations. I find that PAC money from groups that supported the enactment of the September 2010 bill is one of the most important determinants of congressional China bashing. A $5000 payment from these groups is associated with legislators being 1.35 times more likely to criticize China for being a “currency manipulator.”

As indicated, the second objective of this paper is to evaluate whether currency manipulation bashing has worked. Presumably, its ultimate objective is to make China abandon its current exchange rate management policy, or at least let it appreciate against the dollar some 20 to 40%, if not more. Therefore, it seems pertinent to investigate how this bashing influences the Chinese renminbi, if at all. To accomplish this objective, I create a bashing index based on the total count of “currency manipulation” news reported in major U.S. newspapers from January 2000 through December 2010. I then test whether the recent nominal appreciation of the Chinese RMB against the U.S. dollar can be linked to the currency manipulation bashing index. If bashing works, one would expect to observe a positive relationship between this index and the rate at which the RMB has appreciated.

The results of the second test suggest that bashing appears to influence the nominal exchange rate for about 3 months, but perhaps not in the direction the U.S. would favor. The relationship between bashing and the rate at which the RMB appreciates, although temporary, is overall negative. This result implies that bashing may actually be counterproductive: the more of it there is, the less likely that it will work.

The rest of the paper is summarized as follows: Section 2 describes in more detail the construction of the currency manipulation count for each legislator. It also provides information regarding the data sources used in the first test. Section 3 presents the political economy model and the regression specification. Section 4 discusses the empirical findings of the first test. The second question addressed in this paper is discussed in Section 5. Part A addresses the data underlying the construction of the China “currency manipulation” bashing index, which is used in the VAR model (discussed in Part B). That section also discusses the empirical results of the VAR model. Section 6 offers some concluding remarks.

2. First test: data

As indicated in the introduction, the first objective of this paper is to understand the effect that PAC money may exert on a legislator's incentive to bash China for its exchange rate management policy. This test requires a quantifiable measure of legislator-specific currency manipulation bashing. To develop such measurement, I use the Factiva search engine to count the number of media articles that contain the following keywords: “China” and “currency manipulation” or “currency manipulator” or “exchange manipulation” or “exchange rate manipulator” and that also included the legislator's name.2 The coverage period for this count is from January 3, 2009 (the first day of the 111th Congress) to August 31, 2010 (the last day of the last month before the House voted on H.R. 2378, the “Currency Reform for Fair Trade Act”). This procedure resulted in a discrete count of articles for each legislator in the 111th Congress.

As one would expect, the distribution for this count variable is skewed. Nearly 45% of the legislators have a count of 0 articles, 13% of them have a count of 1, 7.6% a count of 2, and 7.9% a count of 3. The remaining 26.6% are scattered among cells of 4 or more articles, with the highest one having a count of 154.3

The next set of variables that need to be included in the model are, of course, contributions from PACs to individual legislators, as well as other variables known to influence congressional voting behavior, such as ideology, party affiliation, and constituent interests. The motivation for including these variables is based on previous Congressional voting behavior research. This literature

1 The vast majority of these papers argue that the RMB is undervalued by some 20 to 40%. Whether a revaluation of the RMB or even adopting a flexible exchange rate regime will help to correct the U.S.–China trade imbalance is still in dispute. For more on this issue see Cheung, Chinn, and Fujii (2010) and Chinn and Wei (forthcoming).

2 To minimize the retrieval of irrelevant articles, I performed multiple permutations of the legislator's name (e.g. last name only, last name and first name, last name and nickname, last name, first name, and nickname, last name and first name initial) and also included the district and state that he or she was representing. I then selected those that correctly identified the legislator.

3 This legislator is Sandy Levin (D-MI, 12th Congressional District), who served as Chair of the Trade Subcommittee during the 111th Congress, and has been an outspoken critic of China's currency policy. See: http://thehill.com/business-a-lobbying/112967-dems-ready-to-push-china-on-currency-manipulation?page=2.

-200%

-150%

-100%

-50%

0%

50%

100%

N e w

Z e a la

n d

A u st

ra lia

S w

itz e rl a n d

D e n m

a rk

B o liv

ia

N o rw

a y

Ja p a n

U n ite

d A

ra b E

m ir a te

s

S w

e d e n

U n ite

d S

ta te

s

Ic e la

n d

C a n a d a

Is ra

e l

T ri n id

a d a

n d T

o b a g o

Q a ta

r

K u w

a it

S in

g a p o re

K a za

kh st

a n

C ze

ch R

e p u b lic

H u n g a ry

B ru

n e i D

a ru

ss a la

m

B ra

zi l

B a h ra

in

S a u d i A

ra b ia

O m

a n

P o la

n d

M e xi

co

R u ss

ia

K o re

a

C h in

a

M a la

ys ia

C o lo

m b ia

C h ile

M a u ri tiu

s

In d o n e si

a

S o u th

A fr

ic a

T h a ila

n d

S ri L

a n ka

U n ite

d K

in g d o m

L ib

ya

Ir a n , Is

la m

ic R

e p u b lic

o f

P a ki

st a n

N e p a l

In d ia

Fig. 1. Degree of undervaluation relative to the U.S. Dollar, 2009. Notes: Figures are in percent. China is identified in red. Negative numbers indicate overvaluation. Undervaluation is calculated as 1 minus the ratio of the implied purchasing power parity conversion rate (source: www.imf.org/external/pubs/ft/weo/2009/01/…/ WEOApr2009all.xls WEO subject, year: PPPEX, 2009) to the actual exchange rate (as of August 2009).

229C.D. Ramirez / China Economic Review 27 (2013) 227–237

has highlighted the role of such variables in helping to explain how legislators vote. For example, party affiliation may be important as representatives often influence by party discipline. The role of ideology in explaining legislative voting behavior has also been frequently highlighted in the literature (Kau & Rubin, 1979; Levitt, 1996; Poole & Rosenthal, 2007).

A commonly used ideology index is Poole and Rosenthal's (1997) comprehensive DW-Nominate scores. The more negative the score, the more “liberal” the legislator is deemed to be. Analogously, the higher (and the more positive) the score, the more “conservative” is the legislator. DW-Nominate scores are highly correlated with party affiliation. Therefore, studies that include this measure of ideology tend to not include party affiliation as an additional covariate. After all, the information contained in party affiliation is also contained in the DW-Nominate score. Nonetheless, it is possible that party affiliation exerts influence on legislative behavior, even after controlling for ideology. For that reason, this variable is kept in the regression specifications.

The role of PAC contributions in shaping legislative voting behavior has also been extensively emphasized in the political economy literature, although the degree to which it matters is still not fully resolved (Ansolabehere, de Figueiredo, & Snyder, 2003). In the context of this paper, there are well-defined interest groups with opposing views regarding China's exchange rate management policy: those who may be adversely affected by a low value of the RMB, and those who are positively affected by it. In order to identify which group falls into which camp, I rely on the PAC contribution data that individual groups made to different legislators at the onset of the passage of H.R. 2378. Clearly, businesses that have lost the most from the rise in the bilateral trade deficit would be most willing to influence congressional legislative action against China. Contributions from these groups can also be interpreted as buying Congressional bashing against China for its currency management policy.

There are also interest groups that benefit from a high value of the dollar vis-à-vis the RMB. Such groups tend to represent importers of Chinese intermediate and final products, whose profitability clearly depends on the exchange rate. For these businesses, a high value of the dollar relative to the RMB keeps them competitive.4

The political economy literature also highlights the importance of constituent interests in influencing legislative behavior (e.g. Kalt & Zupan, 1984, 1990; Krehbiel, 1993). According to these models, legislators, in their quest for maximizing their chances for reelection, react to constituent interests and behave according to their preferences. In the context of this paper, congressional districts that have

4 www.opencongress.org identifies specific organizations that supported the passage of H.R. 2378. These include: AFL-CIO, International Association of Machinist and Aerospace Workers, Alliance for American Manufacturing, American Iron & Steel Institute, Aluminum Extruders Council, and United Steelworkers. Specific organizations that opposed this bill include: Coalition of Service Industries, National Cattleman's Beef Association, National Retail Federation, American Soybean Association, International Dairy Foods Association, National Council for Farmer Cooperatives, National Fisheries Institute, U.S. Chamber of Commerce, TechAmerica, American Apparel & Footwear Association, Securities Industry & Financial Markets Association, USA Poultry & Egg Council, Sporting Goods Manufacturers Association, Pacific Coast Council of Customs Brokers and Freight Forwarders, American Meat Institute, Financial Services Roundtable. Source: www.opencongress.org/bill/111-h2378/money. Financial data for these groups was obtained from www.opensecrets.org and www.maplight.org.

230 C.D. Ramirez / China Economic Review 27 (2013) 227–237

been disproportionally affected by trade with China, such as districts that are more oriented towards manufacturing industries, are more likely to be in favor of legislation against China. They are also more likely to be critical of China's exchange rate policy.

Of course, there may be congressional districts that benefit from trade with China. Districts with a significant amount of employment in trade industries, such as wholesalers, are unlikely to favor legislation against China. Therefore, it is important to control for the influence of constituent interests in the regressions.

3. First test: model specification

The model adopted in this analysis is grounded on the political economy of legislative behavior literature.5 Formally, the regression specification is:

cmbi ¼

5 The applicat

6 Con 7 This 8 It is

α0 þ α1 PAC�Favori � �

þ α2 PAC�Againsti � �

þα3 DW�Nomi � �

þ α4 Pol�Af f ili � �

þ α5 Manuf acturingið Þ þ α6 Tradeið Þ þ α7 Voteið Þ þ εi: ð1Þ

In Eq. (1), the dependent variable is cmb, which stands for “currency manipulation bashing” from Congress member i. This variable is the legislator-specific currency manipulation count described above.

The independent variables PAC_Favor and PAC_Against capture the influence of PAC contributions on the frequency of currency manipulation bashing. A priori, one would expect α1 to be positive. As the legislator receives more contributions from interest groups that stand to lose from trade with China, the incidence of bashing should rise.

By contrast, PAC_Against should not have any significant influence on cmb. Why not? cmb tracks legislator-specific statements that are critical of China for its exchange rate regime stance. In fact, the specific keywords “currency manipulation” or “currency manipulator” convey a message with a disapproving judgment or tone against China. Therefore, it is unlikely that legislators who do not agree that China has an exchange rate problem would use it. Even if they did—say, to deny that there is a “currency manipulation” issue—the incidence of its usage would likely be very limited.

The variable DW_Nom (DW Nominate scores) controls for the influence of ideology on the legislator's motivation to criticize China's exchange rate policy. Poole and Rosenthal (2007)'s DW scores contain two dimensions for each legislator. To be as comprehensive as possible, the model includes both of them. Ideology is an important variable to include because it is possible that ideological differences among congressmen can explain differences on how they feel about China. As a robustness check, political party affiliation (Pol_Affil) is also included in the model.

To capture the influence of constituent interests Eq. (1) includes the log of the share of the congressional district's labor force involved in manufacturing (manufacturing), as well as the log of the labor force share in wholesale trade, (trade). The coefficient for the manufacturing sector variable is expected to be positive since those districts are likely to be adversely affected by the bilateral trade deficit. By contrast, the trade variable coefficient is expected to exert a negative effect on cmb as these districts benefit from trade with China.6

PAC contributions, ideology, party affiliation, and constituent interests comprise the most important determinants of legislative voting behavior according the literature. Nonetheless, it is possible that there are other factors that may explain the frequency with which legislators express critical opinions about China and that may also influence the legislator's voting behavior. After all, there is a high correlation between those legislators that criticized China for its currency stance and how they voted in H.R. 2378. Therefore, another useful variable to include as a robustness check is the actual vote on H.R. 2378. The inclusion of this variable may be quite powerful if it captures other controls missing from the model. The variable “vote” is an indicator variable which takes the value of 1 if the legislator voted in favor of H.R. 2378, 0 otherwise. If this variable's coefficient is estimated to be positive and significant, it would indicate that there may be other factors influencing cmb that are not captured by the controls already included.

As mentioned above, the dependent variable, cmb, is a count variable that is discrete in nature. Therefore, Eq. (1) is unlikely to satisfy the basic assumptions underlying ordinary least square regressions—continuity and normalcy of the error terms. Under such circumstances the more appropriate estimation method is either a Poisson regression, or a negative binomial regression (Hilbe, 2011). Eq. (1) is estimated using the negative binomial regression, as opposed to the Poisson regression, as one of the critical assumptions behind the use of the Poisson regression—that the mean and standard deviation of the dependent variable ought to be roughly similar—is violated.7 Nonetheless, it is worth pointing out that when the regressions were done using the Poisson technique, the results were very similar.8

literature on this subject is quite vast. Grossman and Helpman (2001, 2002) provide a comprehensive overview. See also Stratmann (2005). For a recent ion of the political economy model in the context of China-related legislation see Xie (2010). gressional district employment data is from the U.S. Census Bureau (www.census.gov). test is done via the “alpha” test, which is included in the regression results table. also worth mentioning that the results using standard OLS were qualitatively similar to those obtained with the negative binomial regression.

231C.D. Ramirez / China Economic Review 27 (2013) 227–237

4. First test: empirical findings

Table 1 presents the main results of the first test. The explanatory variables presented in Eq. (1) are introduced in a nested form in order to evaluate the effect that the controls may have on the point estimate of the PAC contributions variable. The results in regression (1) indicate that an additional $5000 in PAC contributions from groups that favored the passage of H.R. 2378 increases the odds of that a legislator would engage in “currency manipulation” bashing by about 1.35. This estimate is measured with enough statistical precision that it is rendered significant at the less than 0.001%. Of course, the results from regression (1) do not control for other factors. However, the remaining regressions suggest that the estimated coefficient survives the inclusion of the full set of controls. The reported estimate ranges from around 1.26 to 1.32 in regressions (2) through (5). Thus, even after including all control variables (in Regression 5), the results unambiguously indicate that PAC contributions are associated with higher levels of Congressional China bashing.9

Regressions (2) through (5) also indicate that there are other factors that affect the incidence of currency manipulation bashing. Regression (2), for example, includes measures of ideology as well as the political affiliation indicator variable. The 1st dimension of the DW-Nominate scores variable as well as the “Democrat” indicator variable (which is the political affiliation control variable discussed above) are both statistically significant at the 5% level in all but one instance. (In regression (5) the “Democrat” variable is significant at the 10% level.) Thus, we can conclude that both ideology and political affiliation influence the incidence of bashing. The “Democrat” indicator variable is less than one suggesting that, holding all else constant, Republican legislators criticize China's exchange rate policy more frequently than Democrat legislators do.

Regressions (3), (4), and (5) indicate that constituent interests also matter. The “Manufacturing” variable coefficient, for example, ranges from 1.78 to about 2.07. These magnitudes imply that a one percentage point increase in the manufacturing employment share of the labor force at the congressional district is associated with an increase of approximately 19.6% in the incidence of currency manipulation bashing by the legislator of that district, holding all else the same.10 The “Trade” sector variable (in Regressions (4) and (5)), measures the wholesale employment share of the labor force at the district level. Although the estimated coefficients are significant at the 10% level, they are in line with expectations. The fact that they are less than one suggests that a higher proportion of the congressional district's labor force involved in the trade industry is associated with a lower incidence of currency manipulation bashing by the district's legislator.

As a robustness check, the indicator variable “Yes Vote” is included in regression (5). This variable takes the value of 1 if the legislator voted in favor of the “Currency Reform for Fair Trade” bill. As explained above, the idea behind the inclusion of this variable is to control for any remaining factors not already captured by the other covariates, and to evaluate its effect on the PAC contribution variable. The estimated coefficient, although it is larger than one, is not statistically significant at standard levels. This finding suggests that the model appears to capture the relevant variables that explain the incidence of bashing among legislators.

In order to get a better sense of the magnitude of the PAC contributions variable vis-à-vis the effect of the other variables in the model, Table 2 presents the negative binomial regression results in elasticity terms. The implied elasticities for the variable that measures PAC contributions from groups that favored legislation against China are not trivially small. They range from 0.37 to nearly 0.48. Hence, a 1% increase in these contributions results in approximately 0.4% increase in the likelihood that a legislator will voice criticism against China. By contrast, the elasticities of PAC contributions from groups that were against passing anti-China legislation, while positive, are all statistically insignificant at standard levels.11 The remaining implied elasticities are very consistent with Table 1 regression results: the first dimension of the DW Nominate scores, political party affiliation, and constituent interests affect the incidence of currency manipulation bashing.

In terms of size (and statistical significance) the most important variables are the ones that capture constituent interests (manufacturing, trade), the political party affiliation effect (identified as “Democrat” in the table), and PAC contributions from anti-China groups. The fact that the implied elasticity for this last variable is comparable (in absolute value terms) from the ones estimated for constituent interests and party affiliation suggests that the effect of PAC contributions from these groups is indeed quite sizable.

Overall, the results of this first test deliver a clear and consistent message: Congressional bashing of China's exchange rate policy is intimately associated with PAC contributions from business interests as well as constituent interests directly affected by the bilateral trade deficit.

5. Second test: does “currency manipulation” bashing work?

As mentioned in the introduction, investigating the issue of whether “currency manipulation” bashing affects the nominal RMB/$ exchange rate complements the first test discussed above. It is a natural follow up question to address.

To address this second question, I investigate whether the recent nominal appreciation of the RMB against the U.S. dollar responds to the intensity of currency manipulation bashing. In order to execute this test, it is necessary to have an index that

9 The regressions also indicate that the PAC_Against coefficient, although statistically insignificant, is larger than 1, which seems somewhat counterintuitive. This result can be attributed to randomness in the estimation process. Indeed, when the regressions are estimated via OLS, these coefficients are, for the most part, negative (and statistically insignificant), while the other coefficients are qualitatively similar to the ones presented in Table 1. 10 The 19.6% figure is calculated as follows: since the mean manufacturing share is 10%, a one percentage point increase (i.e. from 0.10 to 0.11) is equivalent to a 9.53% increase in log terms (=ln(0.11)−ln(0.10)). With the estimated negative binomial factor of approximately 2, this translates to a 19.6% increase in currency manipulation bashing. 11 These elasticities were actually negative (and statistically insignificant) when the regressions were estimated via OLS.

Table 1 Explaining the incidence of congressional “currency manipulation” bashing.

Coefficient Reg. 1 Reg. 2 Reg. 3 Reg. 4 Reg. 5

PAC-Favor-5k 1.351 1.315 1.273 1.265 1.264 (0.058) (0.057) (0.056) (0.056) (0.056) 0.000 0.000 0.000 0.000 0.000

PAC-Against-5k 1.006 1.018 1.024 1.022 1.025 (0.016) (0.017) (0.017) (0.017) (0.017) 0.692 0.275 0.152 0.177 0.126

DW Nom 1 0.237 0.198 0.207 0.274 (0.121) (0.102) (0.105) (0.150) 0.005 0.002 0.002 0.018

DW Nom 2 0.859 0.755 0.732 0.697 (0.190) (0.170) (0.163) (0.157) 0.492 0.214 0.163 0.109

Democrat? 0.323 0.326 0.341 0.391 (0.178) (0.179) (0.184) (0.216) 0.041 0.041 0.046 0.089

Manufacturing Sector 1.783 2.065 2.006 (0.367) (0.454) (0.445) 0.005 0.001 0.002

Trade Sector 0.496 0.524 (0.179) (0.190) 0.052 0.076

Yes Vote 1.501 (0.447) 0.172

Alpha 2.986 2.844 2.768 2.734 2.717 (0.267) (0.258) (0.252) (0.250) (0.249) 0.000 0.000 0.000 0.000 0.000

Num obs 433 433 433 433 433 LR Chi-2 64.80 77.67 85.30 89.10 90.94 Prob>Chi-2 0.000 0.000 0.000 0.000 0.000 Pseudo R2 0.033 0.039 0.044 0.046 0.047

Notes: This table presents the negative binomial regressions of a congressman's “currency manipulation” count on the following independent variables: “PAC-Favor-5k”—contributions from Political Action Committees that favored the passage of the “Currency Reform for Fair Trade” bill in $5000 units (i.e. 1=$5000); “PAC-Against-5k”—contributions from Political Action Committees that were against the passage of the “Currency Reform for Fair Trade” bill, in $5000 units; “DW Nom 1” is the DW Nominate ideology score from Poole and Rosenthal, first dimension; “DW Nom 2” is the second dimension of the DW Nominate ideology score; “Democrat?” is an indicator variable equal to 1 if the legislator is a member of the Democratic party, 0 otherwise; “Manufacturing Sector” is the log of manufacturing employment relative to total employment in the Congressional District; “Trade Sector” is the log of wholesale trade employment relative to total employment in the Congressional District; “Yes Vote” is an indicator variable equal to 1 if the congressman voted in favor of the “Currency Reform for Fair Trade” bill, 0 otherwise; “Alpha” is the test of the hypothesis that the mean and standard deviation of the dependent variable is the same. The reported coefficients measure how the odds of the “currency manipulation” count change with a unit increase in the independent variable. Thus, for example, the coefficient of 1.351 for the PAC-Favor-5k says that an additional $5000 in PAC contributions to a particular congressman from groups that favored passage of this bill increased the odds of that congressman's “currency manipulation” count by a factor of 1.351, holding all else constant. Standard errors are included in parenthesis under the coefficients. P-values are reported in italics under the standard errors.

232 C.D. Ramirez / China Economic Review 27 (2013) 227–237

tracks “currency manipulation” bashing over time. I developed such an index following a procedure similar to the one described above for the construction of the China “currency manipulation” count for individual legislators. The following section describes in more detail the construction of this index.

5.1. Deriving the China “currency manipulation” bashing index (CCMBI)

The CCMBI index is defined as follows:

CCMBIt ¼ #CMt þ #CMrt þ #EMt þ #EMrt

#Chinat #CMt+#CMrt+#EMt+#EMrt is the number of newspaper articles printed at time t that contain the keywords “China”

where

and one or more of the following keywords: “currency manipulation” (CM) or “currency manipulator” (CMr) or “exchange rate manipulation” (EM) or “exchange rate manipulator” (EMr) at time t. The denominator of the CCMBI index is the total number of newspaper articles about China printed at time t. This index is computed on a monthly frequency from January 2000 through December 2010.

The newspaper articles used to construct the CCMBI were retrieved using the Factiva search engine. Factiva is an ideal source because it is very comprehensive, tapping all major U.S. newspapers including the New York Times, Wall Street Journal, and Washington Post. Moreover, it permits the imposition of filters to eliminate articles that are irrelevant for the purposes of the search. Obituaries, recurrent market news, advertisements, and the like were eliminated from the counts. To obtain the largest set of newspaper articles, I selected as the source “all major newspapers: U.S.”

Table 2 Explaining the incidence of congressional “currency manipulation” bashing: estimated elasticities.

Coefficient Reg. 1 Reg. 2 Reg. 3 Reg. 4 Reg. 5

PAC-Favor-5k 0.476 0.432 0.382 0.372 0.370 (0.068) (0.068) (0.069) (0.069) (0.069) 0.000 0.000 0.000 0.000 0.000

PAC-Against-5k 0.027 0.080 0.106 0.097 0.112 (0.068) (0.073) (0.075) (0.072) (0.074) 0.692 0.275 0.152 0.177 0.126

DW Nom 1 −0.060 −0.067 −0.065 −0.053 (0.021) (0.021) (0.021) (0.023) 0.005 0.002 0.002 0.018

DW Nom 2 −0.017 −0.031 −0.034 −0.040 (0.025) (0.025) (0.025) (0.025) 0.492 0.214 0.163 0.109

Democrat? −0.665 −0.658 −0.633 −0.552 (0.325) (0.322) (0.317) (0.325) 0.041 0.041 0.046 0.089

Manufacturing Sector 0.578 0.725 0.696 (0.205) (0.220) (0.222) 0.005 0.001 0.002

Trade Sector −0.701 −0.646 (0.361) (0.364) 0.052 0.002

Yes Vote 0.326 (0.239) 0.172

Notes: This table presents the implied elasticities from negative binomial regressions results presented in Table 1. The variables are defined as follows: “PAC-Favor-5k”—contributions from Political Action Committees that favored the passage of the “Currency Reform for Fair Trade” bill in $5000 units (i.e. 1=$5000); “PAC-Against-5k”—contributions from Political Action Committees that were against the passage of the “Currency Reform for Fair Trade” bill, in $5000 units; “DW Nom 1” is the DW Nominate ideology score from Poole and Rosenthal, first dimension; “DW Nom 2” is the second dimension of the DW Nominate ideology score; “Democrat?” is an indicator variable equal to 1 if the legislator is a member of the Democratic party, 0 otherwise; “Manufacturing Sector” is the log of manufacturing employment relative to total employment in the Congressional District; “Trade Sector” is the log of wholesale trade employment relative to total employment in the Congressional District; “Yes Vote” is an indicator variable equal to 1 if the congressman voted in favor of the “Currency Reform for Fair Trade” bill, 0 otherwise. Standard errors are included in parenthesis under the coefficients. P-values are reported in italics under the standard errors.

233C.D. Ramirez / China Economic Review 27 (2013) 227–237

Fig. 2 displays the CCMBI series. The displayed pattern is consistent with expectations about the timing of the acceleration of U.S. criticism of China for its currency management policy. Around 2003, the incidence of currency-manipulation bashing began to rise. Although it has oscillated in recent years, it still continues to increase. In addition, it is worth noticing that the index peaks precisely during periods of unusually high incidence of commentaries from legislators regarding the RMB. For example, in May 2006, newspaper articles reported that many congressmen and policymakers were disappointed and dissatisfied with the decision of John Snow, then Treasury Secretary, not to officially classify China as a “currency manipulator,” even though his report

Fig. 2. China currency manipulation bashing index (CCMBI), January 2000 to December 2010. Notes: The colored line tracks the fraction (in percent) of all news about China that touches on one or more of the following keywords: “currency manipulation,” “currency manipulator,” “exchange rate manipulation,” and “exchange rate manipulator.” The black solid line is the 12-month moving average.

234 C.D. Ramirez / China Economic Review 27 (2013) 227–237

heavily criticized China's exchange rate policy. Many congressmen explicitly voiced their frustration regarding his decision. For example, Rep. Phil English (R-Pa), regarding Snow's decision, commented: “Today the administration chose to plant its head in the sand.”12

The March 2010 peak corresponds to the time the Senate chose to introduce legislation targeting China's currency policy. Many newspapers printed articles directly or indirectly related to this event. For example, on March 16 the Wall Street Journal reports: “A bipartisan group of U.S. senators on Tuesday introduced legislation aimed at forcing the Obama administration to take action against China over its currency policy, reflecting growing anger on Capitol Hill over the issue.”13

The peaks in April and September 2010 also correspond to periods of high Congressional activity regarding China's currency stance. The majority of newspaper articles in April 2010 report discussion on legislation regarding the RMB issue. And, of course, the September 2010 peak corresponds to the House vote on H.R. 2378.

5.2. VAR model of “currency manipulation” bashing

While congressmen debated and criticized China's policy regarding the RMB, its value relative to the U.S. dollar increased in nominal terms. Between 2004 and 2010, the RMB appreciated approximately 20% in nominal terms against the U.S. dollar as Chinese authorities relaxed their currency peg (see Fig. 3). Undoubtedly, there are many reasons for the appreciation. But the obvious question that this sequence of events raises is: Did the currency-manipulation bashing that China endured during this period had anything to do with the nominal appreciation of the RMB? In other words, did bashing work?

This proposition can be explicitly tested by estimating a Vector Autoregression (VAR) model that includes the rate at which the RMB appreciated against the dollar, and the frequency of currency-manipulation bashing (i.e. the CCMBI series). When estimating the effect of bashing on the RMB appreciation rate, it is important to also control for the effect of changes in the equilibrium exchange rate as well as the effect of economic fundamentals. Thus, the VAR model includes the rate of change of China's real effective exchange rate (to control for changes in the equilibrium exchange rate), as well as the rate of change of industrial production (to control for changes in economic fundamentals).14

Formally, the adopted VAR model is:

12 Wa 13 Wa 14 Bot d=330 15 In s criterio present

yt ¼ c þ Xn

i¼1 Aiyy−i þ εt

yt≡ Δet; ccmbit; ΔREERt; ΔIPtð Þ′ ð2Þ

Δet is the first difference of the log of the monthly RMB per U.S.$ at time t, ΔREERt is the percentage change in the real effective

where exchange rate, and ΔIPt is the rate of change in industrial production in China. In Eq. (2), n stands for the number of lags.

15

One advantage of using the VAR model is that it assumes that all variables are endogenous, and that they can potentially cause each other. In the context of this paper, the endogenous variables of interest are the rate of change of the nominal exchange rate, Δet and the currency manipulation bashing index, ccmbi. Causality between these two variables is tested by performing a Granger causality test. This test can help to clarify the direction of causality—whether the currency bashing series explains subsequent changes in the nominal exchange rate, or, instead, whether changes in the nominal exchange rate affect subsequent bashing, controlling for other factors. The first direction—bashing affecting the exchange rate—is, in essence, the test of whether bashing works. The reverse direction—exchange rate changes affecting bashing—evaluates the proposition that the intensity of the bashing may be a function of past changes in the exchange rate.

The results from the Granger causality test are presented at the bottom of Table 3. The Chi-squared statistics indicate that the direction of causality runs from bashing to the exchange rate (Chi-squared statistic of 7.281, with a P-value of 0.026 when there are 2 lags in the VAR model; Chi-squared statistic of 14.871, with a P-value of 0.002 when there are 3 lags in the VAR model), but not the other way around (Chi-squared statistic of 0.264, with a P-value of 0.876 when there are 2 lags in the model, and Chi-squared statistic of 1.788, with a P-value of 0.618 when there are 3 lags in the model). Hence, the results indicate that bashing Granger causes changes in the nominal exchange rate. In other words, the exchange rate appears to respond to bashing. Changes in the exchange rate, by contrast, do not appear to influence bashing in any meaningful way.

The Granger causality tests, however, do not say much about the magnitude of the effect of bashing on the exchange rate. Worse, they do not say anything about the sign of the effect—whether positive or negative. To find out more about these effects, it is more helpful to look at the actual regressions.

The top of Table 3 presents the key VAR regressions of interest: in regressions (1) and (3) the dependent variable is the change in the log of the nominal exchange rate, while in regressions (2) and (4) the constructed bashing index is the dependent variable. Regressions (1) and (2) report the results when the lag-order is 2, while regressions (3) and (4) report the results when the lag-order is 3. Since the Granger causality tests indicate that the only statistically significant direction of causality is from bashing

shington Post, May 11, 2006, Financial Section page D06, “Treasury Resists Calling China a Manipulator.” (Accessed: May 10, 2011.) ll Street Journal Online, March 16, 2010, “Senators Introduce China Currency Manipulation Bill.” (Accessed: May 11, 2011.) h series were obtained from the International Monetary Fund International Financial Statistics database: elibrary-data.imf.org/FindDataReports.aspx? 61&e=169393. electing the number of lags, I rely on standard lag-order selection statistics. Both the Akaike information criterion and the Hannan and Quinn information n recommended the use of 3 lags, while the Schwarz's Bayesian information criterion recommended 2 lags. In order to be as comprehensive as possible, I the regression results for both lag-order selections.

Fig. 3. RMB per U.S.$, monthly average, January 2000 to December 2010.

235C.D. Ramirez / China Economic Review 27 (2013) 227–237

to the exchange rates, I focus the discussion on regressions (1) and (3) (where dependent variable is Δet) for the remainder of this section.

Regressions (1) and (3) indicate that estimated coefficient of the first lag of the bashing index variable is negative and statistically significant at standard levels, even after controlling for changes in the real effective exchange rate and changes in industrial production. However, the coefficient in the second lag is positive (and significant), and its magnitude is at least as large as that of the first lag (in regression (1)) and even slightly larger in regression (3). Thus, when the lag order is 2, there does not appear to be any sustained effect of ccmbi on Δet. However, when the lag-order is 3, the regression indicates that the third lag of the bashing variable is negative and significant at standard levels. For the lag order 3, then, the results appear to indicate that over the entire 3-month period, a surge in bashing decreases the rate at which the exchange rate changes. Additional lags were not statistically significant.

There are no hard and fast rules about which model is the “preferred” one. As stated above, different information criterion advocated different lag orders (either 2 or 3), which is why both are presented here. However, given that the 3rd lag of the ccmbi is still highly significant, it evidently carries helpful information for modeling changes in the nominal exchange rate. From this perspective, it can be argued that this ought to be the preferred regression specification.

To confirm what regression (3) is suggesting, it is useful to examine the impulse response function implied by the VAR equation. The impulse response function describes the reaction of the dependent variable (in this case, the change in the exchange rate), to a one-standard deviation shock to the bashing index series, after orthogonalizing the residuals in the system in order to ensure that they are not contemporaneously correlated.16

A graph of the impulse response function, along with the 90% confidence intervals, is presented in Fig. 4. As can be readily discerned, a surge in the bashing index results in an almost 4% decline on the rate at which the RMB appreciates against the dollar after only one month. After a reversal in month 2, the third lag effect of the shock further reduces the appreciation rate of the RMB by nearly 5%. The error bands in the impulse response function suggest that the effect of bashing is most precisely estimated in the first and the third months. By the fourth month, the effect of the shock has largely dissipated. Hence, although bashing tends to slow down the rate at which the RMB appreciate against the dollar, the effect is clearly temporary.

But the fact that the effect is negative, even if temporary, suggests that bashing is associated with a subsequent slowdown in the rate at which the RMB appreciates against the U.S. dollar. From the perspective of the United States, this result indicates that bashing is actually counterproductive—it actually works to delay the rate of currency appreciation.

A skeptical reader may claim that this result is purely coincidental. That in reality, bashing should not have any effect on the rate at which the RMB appreciates. But the chances of committing a Type I error—falsely rejecting the null hypothesis of no effect—are relatively small (about 1 in 200 when the VAR includes 3 lags). In addition, it should be noted that there is a plausible explanation for this effect. As already mentioned, China is not the only country that may have an undervalued currency. Yet, time and again, it is singled out in Congress for its currency management policy. Given that bashing is largely the product of pressure

16 Identification is achieved using the Cholesky decomposition. Since the percent change in REER can arguably be identified as the most “exogenous” in the model, it is included first. The percent change in industrial production is included second, and the ccmbi variable is included last. It should be noted, however, that deleting the percent change in REER and IP from the model does not affect the effect of ccmbi on the nominal exchange rate. A comprehensive discussion of impulse response functions can be found in Lutkepohl (2005), Chapter 2. In addition, it is worth pointing out that, as a robustness check, the order in which the variables were included in the system was reversed (for both versions of the model). The results, however, were very similar.

Table 3 Evaluating the Effectiveness of Currency Manipulation Bashing.

Regression 1 Regression 2 Regression 3 Regression 4

Dependent variable: Δet

Dependent variable: ccmbit

Dependent variable: Δet

Dependent variable: ccmbit

Δe−1 0.382⁎⁎ 0.000 0.398⁎⁎ 0.001 (0.083) (0.001) (0.087) (0.001)

Δe−2 0.283⁎⁎ 0.000 0.210⁎⁎ 0.001 (0.081) (0.001) (0.088) (0.002)

Δe−3 0.094 −0.001 (0.088) (0.001)

ccmbi−1 −11.894⁎⁎ 0.386⁎⁎ −8.944⁎ 0.373⁎⁎ (5.195) (0.086) (5.073) (0.087)

ccmbi−2 11.820⁎⁎ 0.182⁎⁎ 15.982⁎⁎ 0.187⁎⁎

(5.300) (0.089) (5.442) (0.093) ccmbi−3 −14.914⁎⁎ 0.013

(5.365) (0.092) Industrial Production, % change−1 −0.008 −0.000 −0.006 −0.000

(0.007) (0.000) (0.008) (0.000) Industrial Production, % change−2 −0.006 0.000 −0.002 0.000

(0.007) (0.000) (0.007) (0.000) Industrial Production, % change−3 −0.001 0.000

(0.008) (0.000) Real Effective Exchange Rate, % change−1 −6.241 −0.050 −5.773 −0.051

(4.358) (0.074) (4.380) (0.075) Real Effective Exchange Rate, % change−2 −1.438 0.014 −0.316 −0.027

(4.339) (0.074) (4.624) (0.079) Real Effective Exchange Rate, % change−3 3.009 0.166

(4.334) (0.074) Constant 0.000 0.002 0.171 −0.000

(0.000) (0.002) (0.132) (0.002) Number of observations 129 129 128 128 R squared 0.489 0.302 0.524 0.335 Granger Causality test: ccmbi→Δe, Chi-sq. 7.281 14.871

0.026 0.002 Granger Causality test: Δe→ccmbi, Chi-sq. 0.264 1.788

0.876 0.618 Does ccmbi Granger cause Δe? Yes Yes Does Δe Granger cause ccmbi? No No

Notes: This table presents regressions of the change in the log of the nominal exchange rate (RMB/$), Δe, and of the China currency manipulation bashing index, ccmbi, on their own lags, controlling for lags of the growth rate of industrial production in China, “Industrial Production, % change,” and the percent change in the real effective exchange rate, “Real Effective Exchange Rate, % change”. Subscripts indicate the lag. Regressions 1 and 2 present the results when only 2 lags are included, while regressions 3 and 4 present the results when 3 lags are included. Standard errors are included in parenthesis under each coefficient. The first Granger causality test (ccmbi→Δe) examines whether the ccmbi variables affect Δe. The second Granger causality test evaluates the reverse causality: whether the Δe variables affect ccmbi. P-values for each Granger causality test are included under the estimated Chi-squared statistics. ⁎⁎ Significant at the 5% level or better. ⁎ Significant at the 10% level.

236 C.D. Ramirez / China Economic Review 27 (2013) 227–237

groups' influence, continual criticism on the currency issue is likely met with more distrust in China, thereby adversely affecting bilateral trade and economic relations.

6. Concluding remarks

Why are congressmen so concerned with China's exchange rate regime? The most benevolent, if not naïve, explanation is the contention that congressmen are genuinely concerned about the well-being of American citizens, and thus, anything that threatens their welfare, including the potential loss of jobs from competition from abroad, will motivate Congressional action in their defense. However, a more realistic view, advanced by the public choice and political economy literature, contends that politicians are no different than other individuals in society who seek to maximize their own well-being (Buchanan & Tollison, 1984). Although they may have inherent preferences and ideological proclivities, the calculus involved in their decision-making process is also influenced by how it would affect their chances of re-election.

This more realistic interpretation of congressional behavior is consistent with congressional “currency manipulation” bashing. Using data for the 111th Congress (2009 to 2010), I find that an additional $5000 in PAC contributions from interest groups that supported the passage of H.R. 2378 (the “Currency Reform for Fair Trade Act”) increases the odds that a legislator would engage in China bashing by about 1.35 times. In addition, I find that constituent interests from the manufacturing sector in the congressional district also influence the incidence to which legislators criticize China's exchange rate policy. This result is robust to the inclusion of other variables, including how the legislators ultimately voted on this legislation.

8.00%

6.00%

4.00%

2.00%

0.00% 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

P e

rc e

n t

Month

-2.00%

-4.00%

-6.00%

-8.00%

-10.00%

Fig. 4. The effect of “currency manipulation” bashing on the exchange rate: impulse response function. Notes: This figure presents the results of a one-standard deviation shock to the China “currency manipulation” bashing index (CCMBI) to the rate at which the RMB appreciates against the U.S. dollar. The dash lines represent 90% confidence intervals.

237C.D. Ramirez / China Economic Review 27 (2013) 227–237

The second objective of this paper is to investigate the extent to which “currency manipulation” bashing works. More specifically, I examine whether bashing accelerates the rate at which the RMB appreciates against the dollar. The results indicate that it does not. Instead, if anything, the results suggest that bashing may actually slow down the rate of appreciation of the RMB against the dollar. If the ultimate objective of U.S. politicians is to try to persuade China to let its currency float against the dollar, or at least let it appreciate some 20 to 40%, the results here indicate that bashing China may not seem to be the most productive way of going about it.

References

Ansolabehere, Stephen, de Figueiredo, John M., & Snyder, James M. (2003). Why is there so little money in U.S. politics? Journal of Economic Perspectives, 17(no. 1), 105–130.

Buchanan, James M., & Tollison, Robert D. (1984). The theory of public choice—II. Ann Arbor, MI: University of Michigan Press. Cheung, Yin-Wong, Chinn, Menzie D., & Fujii, Eiji (2010). China's current account and exchange rate. (Chapter 7). In Robert C. Feenstra, & Shang-Jin Wei (Eds.),

China's growing role in world trade, Chicago (pp. 231–279). IL: University of Chicago Press. Chinn, Menzie D., & Shang-Jin Wei (forthcoming). “A faith-based initiative meets the evidence: Does a flexible exchange rate regime really facilitate current

account adjustment?” Review of Economics and Statistics. Grossman, Gene M., & Helpman, Elhanan (2001). Special interest politics. Cambridge, MA: MIT Press. Grossman, Gene M., & Helpman, Elhanan (2002). Interest groups and trade policy. Princeton, NJ: Princeton University Press. Hilbe, Joseph M. (2011). Negative binomial regression. New York, NY: Cambridge University Press. Kalt, Joseph P., & Zupan, Mark A. (1984). Capture and ideology in the economic theory of politics. American Economic Review, 74(3), 279–300. Kalt, Joseph P., & Zupan, Mark A. (1990). The apparent ideological behavior of legislators: Testing for principal–agent slack in political institutions. Journal of Law

and Economics, 33(1), 103–131. Kau, James B., & Rubin, Paul H. (1979). Self-interest, ideology, and logrolling in congressional voting. Journal of Law and Economics, Vol. 22(No. 2), 365–384. Krehbiel, Keith (1993). Constituent characteristics and legislative preferences. Public Choice, Vol. 76(1–2), 21–37. Levitt, Steven D. (1996). How do senators vote? Disentangling the role of voter preferences, party affiliation, and senator ideology. American Economic Review,

86(3), 425–441. Lutkepohl, Helmut (2005). New introduction to multiple time series analysis. Berlin, Germany: Springer. Poole, Keith T., & Rosenthal, Howard (1997). Congress: A political-economic history of roll call voting. Oxford: Oxford University Press. Poole, Keith, & Rosenthal, Howard (2007). Ideology and congress. Piscayute, NJ: Transaction Publishers. Staiger, Robert W., & Sykes, Alan O. (2010). Currency manipulation' and world trade. World Trade Review, 9(4), 583–627. Stratmann, Thomas (2005). Some talk: Money in politics, a (partial) review of the literature. Public Choice, 124(nos. 1–2), 135–156. Xie, Tao (2010). U.S.–China relations: China policy on Capitol Hill. New York, NY: Routledge.