Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and education use, including for instruction at the authors institution and sharing with colleagues. Other uses, including reproduction and distribution, or selling or licensing copies, or posting to personal, institutional or third party websites are prohibited. In most cases authors are permitted to post their version of the article (e.g. in Word or Tex form) to their personal website or institutional repository. Authors requiring further information regarding Elsevier’s archiving and manuscript policies are encouraged to visit: http://www.elsevier.com/copyright Author's personal copy Journal of Macroeconomics 32 (2010) 732–746 Contents lists available at ScienceDirect Journal of Macroeconomics journal homepage: www.elsevier.com/locate/jmacro One nation under the fed? The asymmetric effects of US monetary policy and its implications for the United States as an optimal currency area David Beckworth * McCoy College of Business, Texas State University, San Marco, Texas 78666, United States a r t i c l e i n f o Article history: Received 16 January 2009 Accepted 7 December 2009 Available online 23 December 2009 JEL classification: E32 E52 E58 L16 R11 a b s t r a c t Is the United States best served by a single currency? This question is explored in this paper by looking at the regional effects of US monetary policy shocks through the perspective of the optimal currency area framework. Using monthly state-level data for the period 1983:1–2008:3, this paper finds that some regions of the United States during this time may have benefited from having their own currency. Ó 2009 Elsevier Inc. All rights reserved. Keywords: Optimal currency area Regional effects of monetary policy US monetary policy Vector autoregressions ‘‘[T]he weakening labour markets in the sunbelt states are being offset by sharply falling jobless claims in rustbelt states. . . This poses a conundrum for the Fed as it determines interest rate direction: How do you interpret a split-personality labour market, and its inconsistent impact on growth and inflation?” David Parkinson in ‘‘Sunbelt vs. Rustbelt Tug-of-War”, July 7, 2007. 1. Introduction Is the United States best served by a single central bank conducting countercyclical monetary policy? According to the optimal currency area (OCA) criteria, the answer is yes if the various regions of the United States share similar business cycles or have in place flexible wages and prices, factor mobility, fiscal transfers, and diversified economies.1 In the former case, similar business cycles among the regions mean that a national monetary policy, which targets the aggregate business cycle, will be stabilizing for all regions. In the latter case, dissimilar business cycles among the regions make a national monetary policy destabilizing—it will be either too simulative or too tight—for some regions unless they have in place the above listed economic shock absorbers. Many, if not most, observers believe the United States fits these criteria and is a successful * Tel.: +1 512 245 6067. E-mail address: [email protected] 1 The seminal work in the OCA theory comes from Mundel (1961), McKinnon (1963), and Kenen (1969). See Mongelli (2002) for a survey of the OCA literature. 0164-0704/$ - see front matter Ó 2009 Elsevier Inc. All rights reserved. doi:10.1016/j.jmacro.2009.12.001 Author's personal copy D. Beckworth / Journal of Macroeconomics 32 (2010) 732–746 733 monetary union. As a result, the United States is often held up as a benchmark case of an OCA for other areas attempting to form a monetary union (Eichengreen, 1990; Bayomi and Eichengreen, 1993; Feldstein, 1997; Bordo, 2004). But is the United States truly an OCA? The above quote implies that the Rustbelt and the Sunbelt regions of the United States may have benefited from having their own currencies and monetary policy in early-to-mid 2007. This suggestion that the United States might benefit from the use of regional currencies is consistent with Rockoff (2000) who concludes that the United States’ OCA status is tenuous at best. Other studies also raise questions about the OCA status of the United States. Owyang et al. (2005) and Crone (2005, 2006) shows that business cycles continue to vary in non-trivial ways across US regions. Carlino and Defina (1998, 1999a, 1999b), Crone (2005) and Owyang and Wall (2006), meanwhile, find that monetary policy shocks generate large asymmetric effects across US regions. These findings imply that certain regions of the United States may fail to meet the OCA criteria. Kouparitsas (2001) examines this possibility directly and finds that three of the eight Bureau of Economic Analysis (BEA) regions are not in the dollar OCA. Looking at state employment business cycles, Partridge and Rickman (2005) similarly conclude the United States fails to meet the traditional criteria for an OCA. Far from being a closed case, then, the United States as an OCA remains an open question. This paper provides further evidence on the OCA status of the United States. It does so by taking another look at the asymmetric effects of US monetary policy shocks on state economies. Previous studies in this literature have tried to explain the asymmetric effects of such shocks by examining whether the different transmission channels of monetary policy were more prevalent in certain regions. This paper takes a different approach by explaining the asymmetric effects of monetary policy shocks by way of the OCA criteria. Specifically, it regresses the asymmetric effects against measures of the OCA criteria and uses them to construct a decision rule that determines whether certain states would have benefited from having their own currency and monetary policy. The paper also looks at the source of economic shocks in each state and whether they are dissimilar enough to justify monetary autonomy. This approach is different than Kouparitsas (2001) who uses the response of regional economies to and the relative importance of common and idiosyncratic shocks to determine the dollar OCA.2 This paper is also unique in several other ways. First, it makes use of a monthly data set of real economic activity at the state level. The use of this higher frequency data allows for a more complete picture of each state’s economic response to a monetary policy shock. Second, previous studies have been beset with the challenge of having to deflate nominal regional economic measures, but without having an appropriate deflator to do so. Using the above data and a unique modeling approach this paper is able to overcome this challenge. Third, the modeling approach used in this paper allows for a consistent monetary policy shock to be estimated across all states. This paper, then, is able to show whether each state economy responds differently in a statistically significant way from the national economy to the same monetary policy shock.3 Collectively, these innovations provide for fresh insights into whether the United States truly is an OCA. To be clear, though, this paper like most studies on the OCA only speaks to the potential gains for US regions having their own currency and does not address the issue of increased transaction costs. The remainder of the paper proceeds as follows. First, the paper reviews the previous research on the asymmetric effects of US monetary policy on regional economies. Second, this paper formally outlines the estimation strategy and then uses it to find the state-level effect of monetary policy shocks. Third, after finding that US monetary policy shocks generate asymmetric effects this paper explains this variation using measures of the OCA criteria. Fourth, the paper uses these findings to assess what states of the United States may have benefited from having their own currency. Finally, the paper concludes with a discussion of the implications from these findings. 2. Previous research on the regional effect of US monetary policy shocks This paper follows a number of studies over the past decade that have examined the regional effects of monetary policy shocks. Carlino and DeFina (1998, 1999a,b) provide the seminal papers in this literature. They estimate a series of vector autoregressions (VARs) in these papers to determine the impact of monetary policy shocks on regional and state economies. When looking at the state-level response (1999a,b), they specifically estimate 48 VARs for the 48 contiguous states for the period 1958:Q1–1992:Q4. In each VAR they include the following variables: real personal income for the state being considered, real personal income for the remainder of the BEA region containing the state being considered, real personal income for the seven other BEA regions, energy prices, core CPI, the index of leading indicators, and the federal funds rate. Carlino and Defina use the VARs to estimate cumulative impulse response functions (IRFs) of each state’s real personal income from a shock to monetary policy (i.e. a shock to the federal funds rate) and find great variation among the states IRFs. In particular, the Great Lakes region is found to be the most adversely affected by monetary policy shocks while its impact on states in the Southwest and Rocky Mountain regions is the least affected of any area in the Unites States. Crone (2005) creates a new set of economic regions based on the similarities in state business cycles for the period 1959:Q1–1993:Q1 and uses them in a VAR identical to the one used in Carlino and DeFina (1998). Crone similarly finds the Great Lakes area to be the most adversely affected from a monetary policy shock while the area he calls the energy belt—comprised of portions of the BEA’s Southwest 2 Kouparitsas (2001) does estimate the effect of monetary policy shocks. His emphasis, however, is on common and idiosyncratic shocks in general and he does not attempt to systematically explain the asymmetric effects of monetary policy as is done in this paper. 3 Kouparitsas (2001) also estimates a consistent monetary policy shock, but as noted earlier he does not try to explain the asymmetric effects of this shock. Author's personal copy 734 D. Beckworth / Journal of Macroeconomics 32 (2010) 732–746 and Rocky Mountain regions—is the least affected. Owyang and Wall (2006) estimate the impact of monetary policy shocks on 19 sub-BEA regions. These regions consist of two to four states each and are all estimated in one VAR. Here, the VAR includes real personal income, the CPI, the federal funds rate, the 10-year Treasury rate, and commodity prices. They estimate the VAR for the periods 1960:Q1–1978:Q4, 1983:Q1–2002:Q4. They look at the regular impulse response functions and find sub-regions in the Midwest-Great Lakes area to be the most harmed by monetary shocks for all periods. They too report that the sub-regions in the Rocky Mountain, Southwest, and Far West areas tend to be the least affected. Lastly, they show that the effects of monetary policy shocks on the sub-regions lessened in the latter period. Finally, Kouparitsas (2001) estimates a VAR that includes the real personal income of the 8 BEA regions, oil prices, and the federal funds rate for the period 1969:Q1–2002:Q1. Using identifying restrictions, he is able to come up with an income shock common to all regions, regional or idiosyncratic shocks, an oil price shock, and a monetary policy shock. He uses these shocks and the responses they create in the regions to determine which areas make up the dollar OCA. His results similarly show differing responses among the regions. He specifically finds the Southwest to be the least affected while the Rocky, Plains, and Great Lakes are the most adversely affected by monetary policy shocks. While this previous research points to US monetary policy shocks generating asymmetric effects, there are some challenges for these studies in estimating the effect of monetary policy when it comes to the state level. One challenge is that the previous studies use real personal income as a measure of state-economic activity. This measure has to be constructed from the available nominal personal income using a price deflator. There are, however, no state price level measures that are at a high enough frequency or are available back far enough to be useful. As a consequence, the national CPI is used in these studies to deflate state personal income. Dowd and LeSage (1997) have shown, however, that using a national price level measure requires ignoring important variation in state price levels. Another challenge is that in order to compare the effects of a monetary policy shock across states there needs to be a consistent measure of a monetary policy shock. However, to estimate a consistent measure in a standard VAR requires all states be included in the VAR which leads to a serious degrees of freedom problem. Carlino and DeFina (1998, 1999a,b) sidestep this issue by estimating for each state a separate VAR that includes the rest of the state’s BEA region and the other remaining seven BEA regions. While this strategy allows the VAR to capture the dynamics of the whole US economy, each VAR has a different set of equations and may be estimating different monetary policy shocks. Owyang and Wall (2006) tackle this problem differently by estimating 19 sub-regions that consist of two to four states each. This leads to a consistent measure of a monetary policy shock within the sub-regions, but still results in potentially different monetary policy shocks among regions. These challenges, then, motivate the need for a better estimation strategy. The next section presents just such an estimation strategy. 3. Estimating the state-level effect of monetary policy shocks Given the degrees of freedom problem noted above, a key issue is how to properly estimate a large VAR that has both macroeconomic and state-economic variables. In order to address this issue, the estimation strategy adopted here is to impose restrictions on the model such that the state economies in the VAR cannot influence the rest of the VAR—other than the state economies bordering them—but themselves can be influenced by the macroeconomic portion.4 Doing so allows a macroeconomic shock such as a monetary policy innovation to influence a state economy upon impact and afterwards but not vice versa. This approach, therefore, not only solves the degree of freedom problem but also makes reasonable restrictions on the relationship between the state and national economy. As shown by Lastrapes (2004, 2006), this approach can be formally demonstrated by beginning with the autoregressive structural model of the form A0 zt ¼ A1 zt1 þ þ Ap ztp þ ut ; ð1Þ where A0 ; . . . ; Ap are n n structural parameters matrices, zt is a n 1 vector of endogenous variables, and ut is a n 1 vector of uncorrelated structural shocks that are assumed to be multivariate normal with mean zero and unit variance. Next, par z1t where z1t contains the macroeconomic variables or the ‘‘comtition the endogenous vector of variables so that zt ¼ z2t mon factors” while the z2t contains the state-economy variables. Here, z1t consist of a national real economic activity measure, the price level, commodity prices, and the federal funds rate while z2t consists of a state real economy activity measure for the 48 states. To avoid the degrees of freedom created by this 52-variable VAR, two sets of overidentifying restrictions can be imposed on the VAR: (1) z1t is block exogenous with respect to z2t and (2) the state-economy variables in z2t are mutually independent, except for the border economies, after conditioning on z1t . The block-exogeniety creates zero restrictions in the Ai matrices, where i = 0, 1, . . . , p, and require the macroeconomic variables to be determined independent of the state-economy variables. The second set of restrictions imposes separately for each state zeroes in the Ai matrices for the other states that do not border it and non-zero weights on states that do border it. The non-zero weights are set equal to 4 Other studies that have taken a similar approach include Barth and Ramey (2001), Davis and Haltiwanger (2001), Frantantoni and Schuh (2003), Lastrapes (2004, 2006), and Irvine and Schuh (2005). Author's personal copy D. Beckworth / Journal of Macroeconomics 32 (2010) 732–746 735 the bordering state-economy’s share of the total border economy and allow for interdependencies among border state economies.5 In short, the restrictions imposed on z2t mean that state economies cannot influence each other unless they are physically contiguous. As noted by Lastrapes (2004, 2006), these overidentifying restrictions make it possible to estimate the large VAR in a twostep procedure: first, the macroeconomic portion of the VAR is estimated and second, the state economies conditional on the macroeconomic variables, the bordering economies, and themselves can be individually estimated. The estimation strategy adopted in this paper is to effectively do this by estimating the following vector of endogenous variables by SUR for each state: zt ¼ ð yusa Þ; cpit ; ppit ; ffrt ; ystate ; yborder t ; t t ð2Þ where yusa is the national real economic activity measure, cpit is the price level, ppit is the commodity price index, and ffr t is t is the state real economic activity measure, and yborder , is the border economy measure. This last the federal funds rate, ystate t t variable is a composite measure created by taking the weighted average of all state real economies bordering the state being estimated and reflects the non-zero weighting discussed above.6 SUR estimation is used here since the block exogeneity restrictions are imposed by not allowing the two state-economy measures to enter the right-hand side of the macroeconomic . This ordering in conjunction with a variable equations.7 The z2t restrictions are effectively imposed by only including yborder t Choleski decomposition of the covariance matrix for the entire vector of endogenous variables allows for the recursive identification strategy outlined by Christiano et al. (1999) whereby the federal funds rate—the instrument of monetary policy—is able to respond instantaneously to shocks in the other macroeconomic variables but can only affect them with a lag. Any change in the federal funds rate not forecasted by the VAR is interpreted as a shock to monetary policy.8 Given the block exogeneity restrictions and the recursive identification strategy—which imposes contemporaneous restrictions—the state variables in no way affect the macroeconomic variables, both upon impact and afterwards. As a result, only one macroeconomic system is being estimated and hence, only one set of monetary policy shocks is being estimated. The same monetary policy shocks, therefore, get applied across all state economies as they are individually estimated.9 Once the model is estimated and the structural monetary policy shocks are identified, innovation accounting can be done to see the effect of these shocks on real economic activity. In particular, cumulative impulse response functions (IRFs) are used to show the cumulative dynamic response of both the US and state economies to a one standard deviation monetary policy shock. To the extent that there are statistically significant differences among these IRFs, monetary policy is generating asymmetric effects. The measure of real economic activity for the United States and the 48 states is the monthly coincident indicator produced the Philadelphia Federal Reserve bank. The coincident indicator is based upon Stock and Watson (1989) and is designed to be a summary measure of real economic conditions in each state. It is derived using nonfarm payroll employment, average hours worked in manufacturing, the unemployment rate, wage and salary disbursements deflated by the consumer price index (US city average), and trend GDP growth rate for each state and the nation. Owyang et al. (2005) use this measure in their study of state-level business cycles. Since this measure is already in real terms, the problem of having to find a proper deflator is avoided. The consumer price index and the producer price index of commodities are used as well. These latter variables plus the federal funds rate come from the St. Louis Federal Reserve bank’s Fred database. All variables are in a monthly frequency and have observations running from 1983:1 to 2008:3. This time period is used to account for the well-documented structural break in monetary policy that occurred in the early 1980s (Boivin and Giannoni, 2006).10 All variables are transformed into log form and first differenced—except for the federal funds rate which was just first differenced—since standard unit root test indicate nonstationarity in the levels of the variables and because there was no convincing evidence found for cointegrating relationships among the variables.11 The VARs are estimated using 13 lags since the Ljung-Box Q test indicates that this many lags are sufficient to whiten the residuals of any serial correlation. 5 The weights are specifically constructed by dividing a bordering state’s personal income in current dollars by the all bordering state economies’ personal income in current dollars for each quarter. The quarterly data is then linearly interpolated into a monthly frequency. The state personal income comes from the BEA and is available only on a quarterly basis. 6 Specifically, the weights explained in footnote 5 are multiplied by the state real economy measure for each of the bordering states and these are then summed up into the border economy composite measure. 7 Other than this restriction, the rest of the system is set up like a standard VAR with lags of all the endogenous variables entering the right-hand side of the equations. This reduced-form VAR is estimated and turned into a structural VAR using the recursive identification procedure outlined above. 8 Given its dynamic nature and the inclusion of four standard macroeconomic variables, the VAR forecast of the federal funds rate can be viewed as the expected path of monetary policy. 9 It is worth nothing that since the equations for the macroeconomic variables have identical right-hand side variables and cannot be influenced by the state variables, the SUR estimation of the macroeconomic portion of the system amounts to the same thing as estimation by least squares equation-by-equation. However, this is not true for the state-economy variables since they are allowed to be influenced by the macroeconomic variables and are estimated in a system of equations that does not have the same number of right-hand side variables. 10 An interesting exercise would be to examine this OCA question before the 1980s. However, the coincident indicator measure only goes back to 1979 and, thus, the analysis is limited to the post 1979 period. 11 The augmented Dicker–Fuller and the Phillips–Perron unit root test were used. First differencing the data was sufficient to remove the unit root. The twostep Engle–Granger cointegration test was used to test for cointegrating relationships. Author's personal copy 736 D. Beckworth / Journal of Macroeconomics 32 (2010) 732–746 Alabama 0.2 Arizona -0.0 -0.2 -0.2 -0.4 -0.4 -0.6 -0.6 -0.8 -1.0 -0.8 10 20 30 40 50 Colorado 1.0 0.8 0.6 0.4 0.2 0.0 -0.2 -0.4 10 20 30 40 50 Conneticut -0.0 0.25 0.00 -0.6 -0.25 -0.8 -0.50 30 40 50 Georgia 0.2 -0.2 -0.4 -0.6 -0.8 -1.0 10 20 30 40 50 Iowa -0.2 -0.4 -0.6 -0.8 10 20 30 40 50 20 30 40 50 Idaho -1.0 30 40 50 20 40 30 40 50 Illinois -0.2 -0.2 -0.4 -0.4 -0.6 -0.6 -0.8 -0.8 20 40 50 30 40 50 40 50 Indiana 0.2 -0.0 -1.0 10 20 30 40 50 10 20 Louisiana 0.75 0.50 0.25 -0.4 0.00 -0.6 -0.25 -0.8 50 10 -0.0 -0.2 40 30 -0.8 20 Kentucky 30 50 -0.4 -0.0 20 40 -0.6 Kansas 10 30 -0.2 10 50 20 Florida -0.0 0.2 30 10 Delaware -1.0 10 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.0 -0.8 20 -1.00 10 1.2 1.0 0.8 0.6 0.4 0.2 0.0 -0.2 -0.4 -0.0 -0.4 -0.6 -0.75 -1.2 20 -0.2 0.50 -0.4 California 0.2 -0.0 10 -0.2 -1.0 10 Arkansas 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 -0.0 -0.50 10 20 30 40 50 10 20 30 Fig. 1. IRF to a standard deviation monetary policy shock. 4. Empirical findings Figs. 1–3 report the estimated cumulative IRFs for each state’s coincident indicator from a positive one standard deviation shock to the federal funds rate. Since the IRFs are cumulative they can be interpreted as the percent change in the level of the real economy. For each state, the black line shows the IRF, the gray line shows the IRF for the United States, and the dashed line shows the simulated standard error bands to account for the precision of the estimates and to show which states are significantly different than the point estimates of the United States.12 Using the United States as a benchmark, these figures show that most of the economic decline from a positive one unit federal funds rate shock occurs by 24 months. For the United States, the decline is 0.25% by that time. Figs. 1–3 also show wide variation among the states’ IRFs. Looking at those states whose standard error bands fall outside the US IRF for at least 25 months, these figures show 12 states—Colorado, Idaho, Louisiana, Montana, Nebraska, New Mexico, North Dakota, Oklahoma, South Dakota, Texas, Utah, and Wyoming—did significantly better than the United States and eight states—Connecticut, Maryland, Michigan, New Jersey, Ohio, Pennsylvania, South Carolina, and West Virginia—that did worse. Consistent with Crone (2005), all the states that did better are in the Energy belt part of the country, running from Texas up to Montana, while most of those that did worse mostly fall into the Rustbelt. These results alone suggest there are vast regions of the United States not in the dollar OCA. To help summarize the information in these figures, Table 1 reports the 12-month, 24-month, and 36-month IRF for each state relative to the United States since this is the region targeted by US monetary policy. The states are ranked in descending order according to the size of their decline. This table makes plain the stark contrast between some of the states. Michigan, for example, declines relative to the United States 0.97% at the 36-month horizon while Wyoming increases 1.60%. Other states like Arkansas, Maine, New York, and Iowa, on the other hand, have declines very similar to the United States. Two key insights emerge from Figs. 1–3 and this table. First, US monetary policy shocks do generate asymmetric effects among the states. Second, there appears to be a regional pattern to those state economies whose response to such shocks is significantly better or worse than the United States. The next section attempts to systematically explain these findings. 5. Explaining the asymmetric effects of monetary policy Prior studies (Carlino and DeFina, 1998, 1999a,b; Owyang and Wall, 2006; Owyang et al., 2005) have tried to explain monetary policy’s asymmetric effects by looking at three monetary policy transmission channels: the interest rate channel, 12 The standard error bands are technically fractiles that come from using Monte Carlo integration techniques to estimate the posterior density of the response coefficients. Sims and Zha (1999) recommend with this approach, which characterizes the likelihood shape, the use of a 68% posterior probability. Author's personal copy 737 D. Beckworth / Journal of Macroeconomics 32 (2010) 732–746 Maine 0.2 -0.0 -0.2 -0.4 -0.6 -0.8 -1.0 -1.2 10 20 Maryland -0.0 30 40 -0.2 -0.0 -0.4 -0.2 -0.6 -0.4 -0.8 -0.6 -1.0 -0.8 -1.2 -1.0 50 10 Minnesota 20 30 40 50 Mississippi -0.0 -0.0 -0.2 -0.2 -0.4 -0.4 -0.6 -0.6 -0.8 -0.8 Massachusetts 0.2 Michigan 0.00 -0.50 -1.00 -1.50 -2.00 10 20 30 40 50 10 Missouri 0.2 20 30 40 50 Montana 1.2 -0.0 -0.2 0.8 -0.4 0.4 -0.6 10 20 30 40 50 Nebraska 0.2 -1.0 10 20 30 40 50 Nevada 0.50 0.1 0.00 -0.1 -0.25 -0.2 -0.3 -0.50 -0.4 -0.75 10 20 30 40 50 New Mexico 0.3 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 10 20 30 10 50 30 40 50 20 30 40 50 New Hamsphere 40 50 40 50 40 50 30 40 50 30 40 50 30 40 50 -1.0 10 20 30 40 20 30 40 50 20 30 North Dakota 0.4 0.3 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 -0.8 10 10 50 -0.4 50 50 -0.8 North Carolina 40 40 -0.6 -0.6 30 30 -0.2 -0.2 20 20 New Jersey -0.0 -0.0 10 10 -0.4 New York 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 40 20 -0.4 10 0.25 0.00 -0.25 -0.50 -0.75 -1.00 -1.25 -1.50 -1.75 0.25 -0.0 0.0 -0.8 10 20 30 Fig. 2. IRF to a standard deviation monetary policy shock. Ohio 0.2 -0.0 -0.2 -0.4 -0.6 -0.8 -1.0 10 20 30 40 50 Rhode Island -0.0 -0.4 -0.6 -0.8 -1.0 -1.2 20 30 40 50 Texas 0.2 -0.0 -0.2 -0.4 20 30 40 Washington 0.6 0.2 0.0 -0.2 -0.4 -0.6 10 20 30 40 50 40 10 50 20 30 40 50 South Dakota 0.2 -0.0 -0.2 -0.3 -0.4 20 30 40 50 Utah 10 20 30 40 50 20 30 50 30 40 10 50 30 20 Virigina -0.0 -0.2 -0.4 -0.6 -0.8 10 20 30 40 10 50 West Virgina 0.2 -0.0 -0.2 -0.4 -0.6 -0.8 -1.0 -1.2 40 20 20 Tennessee Vermont 0.2 -0.0 -0.2 -0.4 -0.6 -0.8 -1.0 -1.2 Wisconsin 10 10 10 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 -0.1 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 0.4 30 Pennsylvania 0.2 -0.0 -0.2 -0.4 -0.6 -0.8 -1.0 -1.2 0.1 10 50 20 Oregon 0.2 -0.0 -0.2 -0.4 -0.6 -0.8 -1.0 -1.2 South Carolina 0.3 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 0.4 10 10 0.2 -0.0 -0.2 -0.4 -0.6 -0.8 -1.0 -1.2 -0.2 10 Oklahoma 1.0 0.8 0.6 0.4 0.2 0.0 -0.2 -0.4 20 Wyoming 2.5 2.0 1.5 1.0 0.5 0.0 -0.5 10 20 30 40 50 10 20 Fig. 3. IRF to a standard deviation monetary policy shock. the broad credit channel, and the narrow credit channel. These studies regress the asymmetric effects against measures of these channels for each state. Only the first channel, measured by the share of interest rate-sensitive industries in each state, Author's personal copy 738 D. Beckworth / Journal of Macroeconomics 32 (2010) 732–746 Table 1 Summary of the IRFs. Relative response of state economy to a standard deviation monetary policy shock. State Levels %D 12 Months out Levels %D 24 Months out Levels %D 36 Months out Significantly different than US > 24 Months Michigan West Virginia South Carolina Maryland New Hampshire Connecticut Rhode Island Pennsylvania New Jersey Ohio Oregon Missouri Vermont Alabama Massachusetts Illinois Virginia Indiana Mississippi Georgia Minnesota Florida Arizona North Carolina Kentucky California Delaware Arkansas Maine New York Iowa Wisconsin Tennessee Kansas Washington Utah Nebraska South Dakota Nevada New Mexico North Dakota Texas Oklahoma Colorado Louisiana Idaho Montana Wyoming 0.56 0.30 0.20 0.20 0.12 0.18 0.16 0.12 0.22 0.15 0.10 0.09 0.09 0.13 0.07 0.03 0.11 0.02 0.10 0.06 0.06 0.03 0.05 0.06 0.05 0.02 0.04 0.01 0.02 0.01 0.02 0.02 0.06 0.07 0.12 0.08 0.08 0.12 0.09 0.12 0.06 0.05 0.15 0.24 0.12 0.32 0.37 0.37 0.83 0.39 0.37 0.36 0.34 0.34 0.33 0.31 0.30 0.28 0.25 0.21 0.19 0.18 0.15 0.13 0.13 0.12 0.11 0.11 0.10 0.10 0.09 0.09 0.07 0.06 0.03 0.01 0.00 0.00 0.01 0.05 0.06 0.15 0.15 0.16 0.17 0.21 0.23 0.24 0.24 0.33 0.33 0.40 0.44 0.80 0.82 1.02 0.97 0.37 0.40 0.41 0.50 0.36 0.33 0.35 0.30 0.30 0.28 0.25 0.24 0.19 0.09 0.19 0.14 0.14 0.09 0.14 0.10 0.09 0.10 0.07 0.11 0.12 0.03 0.01 0.06 0.00 0.00 0.04 0.10 0.14 0.28 0.23 0.19 0.24 0.28 0.28 0.35 0.39 0.49 0.58 0.50 0.94 1.13 1.60 Y Y Y Y N Y N Y Y Y N N N N N N N N N N N N N N N N N N N N N N N N N N Y Y N Y Y Y Y Y Y Y Y Y Relative response is the state IRF minus the US IRF. has been consistently successful in explaining some of the asymmetric effects.13 What these studies find is that those states with a relatively high share of their state economy in interest sensitive industries, particularly manufacturing, are more susceptible to negative monetary policy shocks. Although these studies are not clear why, they also find that those states with relatively high shares of their economy in extractive industries are less affected by such shocks. These findings are consistent with the geographical patterns of the asymmetric effects found in the previous section. Although interesting, these prior studies have only found differences in industry mix to explain the asymmetric effects of monetary policy shocks. None of them have viewed this issue through the perspective of the Optimal Currency Area (OCA) framework. The OCA is useful here because it provides criteria to determine whether multiple regions are best served by a single monetary union. Although there is still some debate, a common set of criteria that comes out of the OCA literature is that if one monetary authority will be conducting countercyclical monetary policy for multiple regions then the regions must either share similar business cycles or have in place the economic shock absorbers of factor mobility, flexible prices, federal fiscal transfers, and a diversified economy.14 If the regions share the 13 See Crone (2007) for a survey of these studies. The seminal work in the OCA literature comes from Mundel (1961), McKinnon (1963), and Kenen (1969). See Mongelli (2002) for a survey of the OCA literature. There are other OCA criteria in the literature, specifically economic openness, political commitment, financial integration, and similarity of assets used as money as noted by Swofford (2000). They, however, are ignored here since they are not an issue for the United States. 14 Author's personal copy D. Beckworth / Journal of Macroeconomics 32 (2010) 732–746 739 same business cycle then monetary policy, which typically targets the aggregate business cycle or an anchor region’s business cycle, should be stabilizing for all regions. If, on the other hand, there are regional economic shocks generating dissimilar business cycles among the regions, then one monetary policy will be destabilizing for some of the regions unless the above mentioned economic shock absorbers are in place. Consider, for example, a region in a currency union whose economy is not welldiversified and is slowing down because of a series of negative shocks to its primary industries. If the monetary authorities in this currency union decide to tighten because the other regional economies are expanding too fast then the region slowing down needs price flexibility, labor mobility, and federal fiscal transfers in order to offset the effects of the contractionary monetary policy. If these economic shock absorbers are absent, then this region would find this tightening of monetary policy to be further destabilizing to its economy.15 In general, the greater the dissimilarity of a region’s business cycle with the rest of the currency union the more important these economic shock absorbers become for the region to be a successful part of an OCA. This understanding implies that how a region responds to countercyclical monetary policy relative to the other regions provides a summary measure of whether that region shares a similar business cycle with the rest of the monetary union or has in place the appropriate economic shock absorbers, the very criteria defining an OCA. One implication, then, is that the asymmetric effects of monetary policy found in the previous section may be the result of some states not being a part of the dollar OCA. Fig. 4 provides some evidence on this interpretation. This figure plots for the three time horizons used in Table 1 the correlation between the state and the national economies against the absolute value of the relative IRF from a standard deviation monetary policy shock. The correlation among the state and the US economies is measured using the coincident indicator measure from the previous section and is based on the entire sample period.16 Here again, the relative IRF is used since it shows the asymmetry of the state responses relative to that of the region targeted by monetary policy, the United States. Like Carlino and DeFina (1998) the absolute value of this IRF (AIRF) is used since the objective is to see whether the asymmetric responses, regardless of sign, can be explained by the OCA criteria. Two important findings emerge from this figure. First, many state economies were not highly correlated with the US economy over the sample period. Specifically, 24 states had economies that were correlated with the national one at a rate less than a 70%. This finding is consistent with those of Owyang et al. (2005) and Crone (2005, 2006) who similarly show wide variation in US regional business cycles. Second, the greater the asymmetric impact of a US monetary policy shock on a particular state the less correlated that state’s economy was with the US economy. This relationship becomes more pronounced the longer the horizon, with as much as 44.43% of the variation explained at the 36-month horizon.17 One interpretation of this finding is that US monetary policy significantly exacerbated business cycles in state economies that were not in sync with the national one and lacked the economic shock absorbers needed to offset the monetary policy shocks. Some states, then, may not have been a part of the dollar OCA over the 1983:1–2008:3 period. Further evidence on this issue can be found in Table 2. This table reports the estimates from a series of regressions where measures of the economic shock absorbers plus the same correlation measure are regressed against the AIRF used above. For robustness and to be consistent with the previous analysis, the regressions are run using the 12-month, 24-month, and 36month AIRF as the dependent variable. The various OCA criteria are measured as follows First, wage flexibility is measured by each state’s relative manufacturing wage while price flexibility is measured by each state’s relative inflation rate. The relative manufacturing wage is constructed by taking the average percent deviation of a state’s hourly manufacturing wage from the US hourly manufacturing wage for the period 1983:1–2008:3. The data comes from the Bureau of Labor Statistics’ Current Employment Statistics survey. The state’s inflation rate is constructed by taking the Bureau of Economic Analysis’ (BEA) nominal state GDP growth rate minus the real state GDP growth rate for the years 1983–2006.18 It is turned into a relative inflation rate by subtracting from it the US inflation rate and taking the average difference. Since the wage and inflation measures used here are rates of change relative to the national average, they should have a mean of zero over the long-run if they are perfectly flexible and the law of one price holds. However, if a state has wages or prices that are persistently higher than the national average then its wages and prices are downwardly sticky and less flexible compared to the nation. If so, the state’s economy will be more susceptible to a negative monetary policy shock. Second, following Eichengreen (1990), labor mobility is measured by looking at the relative persistence in a state to an unemployment shock.19 The measure is constructed by taking the difference between the IRF 5 years out from a 1 unit shock to a state’s monthly unemployment rate and a similar shock to the US monthly unemployment rate for period 1983:1–2008:3. A simple 13-lag autoregressive model is estimated to get the IRFs.20 Here, the larger the IRF 5 years out the greater the unemployment persistence and the greater the labor immobility. 15 This region would have benefited at this time from having monetary autonomy through which it could have eased monetary policy and engineered a competitive devaluation of its currency. 16 Correlations are calculated using the monthly growth rate of the coincident indicator. 17 All relationships were significant at the 1% level. 18 In both the manufacturing wage and state price level measures there is a break in the series due to the US government switching from the Standard Industrial Classification (SIC) system to the North American Industrial Classification System (NAICS). However, both of these measures are constructed relative to the US value making this break a non issue. 19 Only labor is considered for factor mobility since capital mobility is not an issue among the states. Kalemli-Ozcan et al. (2005) has shown that capital flows among states are effectively frictionless. 20 Arguably, some of the persistence in the unemployment rate could come from sticky wages rather than labor immobility alone. However, in the regression that follows the possibility of sticky wages is controlled for by relative manufacturing wage measure. Author's personal copy 740 D. Beckworth / Journal of Macroeconomics 32 (2010) 732–746 Fig. 4. Monetary policy shocks and regional business cycles. Note: AIRF = absolute value of state IRF minus US IRF. The correlation between the state economy and US economy is based on the coincident indicator growth rates for states and the United States. Third, for the diversification of each state’s economy, two measures are used in the analysis. The first one is a measure of how similar the state and US economy are diversified among economic sectors. The more similar the state economy’s diversification is to that of the US economy, the more similar should be the state and US economies response to a monetary policy shocks. This form of diversification is measured by taking the share of a state economy in a particular economic sector minus 0.103 (0.90) Constant 54.75 46.83 (3.60)*** 0.061 (0.12) 12.219 (2.34)** (3.10)*** 0.095 (0.30) 6.188 (2.09)** 44.60 34.90 3.543 59.25 52.12 (3.19)*** 0.334 (0.50) 17.921 (2.54)** 5.636 0.142 (1.90)* 0.065 (0.40) 0.850 (2.00)* 11.112 (1.11) 0.126 (0.50) (3) 36-Month response 65.01 57.83 (3.38)*** (1.60) 47.30 36.49 0.477 (1.38) 0.178 (0.40) 13.865 (2.71)*** 1.491 0.112 (2.60)** 0.003 (0.02) 0.613 (2.60)** 7.606 (1.04) 0.298 (1.40) (5) 24-Month response 0.120 (1.05) 0.155 (0.48) 6.603 (2.18)** 0.572 0.053 (2.67)** 0.083 (1.22) 0.469 (2.18)** 4.352 (1.31) 0.012 (0.10) (4) 12-Month response Note: The dependent variable is the AIRF. *, **, and *** indicate 90%, 95%, and 99% significance levels, respectively. R2 (%) Adjusted R2 (%) Similar business cycle Coincident indicator correlation Industry portfolio volatility Relative share: manufacturing Economic diversification Relative share: extractive industries 1.089 0.136 (2.34)** 0.059 (0.46) Factor mobility and fiscal transfers Unemployment persistence 0.059 (2.56)** Fiscal transfers 0.098 (1.39) Relative inflation rate 0.688 (2.00)* 7.452 (0.99) 0.061 (0.30) (2) 24-Month response 0.488 (2.18)** 4.313 (1.46) Wage/price flexibility Relative manufacturing wage (1) 12-Month response Variables Table 2 Using the OCA criteria to explain the asymmetric effects of monetary policy shocks. 71.22 65.31 (4.03)*** 0.706 (3.21)* 0.020 (0.04) 20.354 (3.21)*** 2.603 0.106 (1.99)* 0.026 (0.19) 0.739 (2.53)** 11.340 (1.25) 0.406 (1.54) (6) 36-Month response 42.43 37.07 (3.07)** 0.160 4.330 (2.17)** 0.061 (2.69)*** 0.416 (1.93)* 0.149 (2.98)*** (7) 12-Month response 62.51 59.02 (5.50)*** 0.552 13.235 (3.11)*** 0.115 (2.84)*** 0.651 (3.22)*** 0.365 (3.57)*** (8) 24-Month response 67.89 64.90 (4.55)*** 0.871 20.668 (3.26)*** 0.113 (2.06)** 0.791 (2.59)*** 0.495 (4.17)*** (9) 36-Month response Author's personal copy D. Beckworth / Journal of Macroeconomics 32 (2010) 732–746 741 Author's personal copy 742 D. Beckworth / Journal of Macroeconomics 32 (2010) 732–746 the share of the US economy in that economic sector. The closer to zero this measure is for a state economy, the closer it is to being diversified along the lines of the US economy. The second type of diversification measure is based on the idea that the state economy should be allocated among economic sectors in a way that minimizes the volatility of its economic portfolio. Following Conroy (1975), the volatility of the state economic portfolio can be measured as follows, r2P ¼ N X N X i¼1 wi wj rij ; ð3Þ j¼1 where wi and wj are the proportion of the state economy in economic sectors i and j, and rij is the covariance between returns in the two sectors. The square root of Eq. (3) is taken to get volatility in terms of the portfolio’s standard deviation. Here, the more diversified the state economy is, the lower is its volatility. Lower volatility, then, should be associated with fewer asymmetric effects arising from a monetary policy shock. With the first form of diversification, two specific measures are constructed for each state: the relative share of manufacturing industries and the relative share of extractive industries. The relative shares of manufacturing and extractive industries come from dividing their share of state personal income over total state personal income and subtracting from the same ratio for the United States for the years 1983–2007.21 These latter two measures are used given the importance previous studies have given them in explaining the asymmetric effects of monetary policy. For the latter form of diversification, one measure is constructed for each state: the volatility of a state economic portfolio divided among the 10 major SIC industries making up personal income for the period 1983–2001.22 Here, sectoral returns are measured by the personal income generated in the 10 SIC industries. Finally, for fiscal transfers this paper turns to the data provided by The Tax Foundation. This organization collects data on net fiscal transfers between states and the federal government. Here, their ‘‘deficit neutral federal expenditures per dollar of federal taxes” is used as the measure of fiscal transfers for the years 1983–2005.23 This measure captures the net federal inflow of funds into a state controlling for any implicit changes in each state’s tax liability arising from changes in the federal budget deficit. The first three columns of Table 2 report the regression results where only the economic shock absorbers are used to explain the 12-month, 24-month, and 36-month horizon of the AIRF.24 Only wage flexibility, labor mobility, extractive industry diversification, and industry portfolio diversification are statistically significant. The signs for these estimates are as expected: the higher the relative manufacturing wage (i.e. the lower the wage flexibility), the greater the unemployment persistence (i.e. the less is labor mobility), the larger the share of extractive industries relative to the US economy, and the greater the industry portfolio volatility the greater are the absolute responses of the state economies to a negative monetary policy shock. Interestingly, the estimated coefficients on these variables and the overall explained variation get larger the farther out the horizon. Columns (4) through (6) repeat these regressions but now include the correlation measure. The same variables plus the correlation measure are statistically significant now except that the extractive industry share only becomes significant at the 36month horizon. These regressions indicate that variation in the amount of the economic shock absorbers and the correlation among state and US economies can explain much of US monetary policy’s asymmetric effects. Columns (7) through (9) report the best fitting models. These models explain between 42.43% at the 12-month horizon and 67.89% at the 36-month horizon. This amount of explained variation in the AIRF is notably more than Owyang and Wall (2006), the only other study to look at the post-1983 period, who are able to explain only about a third of the variation using the monetary policy transmission channels approach. Table 3 reports on the economic significance of the OCA criteria in the best fitting models. Here, the standard deviation or typical change of each OCA variable across the states is multiplied by its regression estimate. This product can then be compared against the average AIRF to see if the OCA variables individually are economically meaningful. The table also reports the semi-partial R2 for each OCA variable. This table indicates that there are economically meaningful contributions made by all the OCA variables. Their importance, however, changes over the horizons. Wage flexibility and labor mobility are more important during the first two horizons while the industry portfolio volatility and the correlation measure become more important during the latter two horizons. Overall, then, the OCA criteria appears to provide a good framework for understanding the asymmetric effects of US monetary policy shocks. These findings suggest that some states of the United States may have gained from having their own currency and monetary policy. This possibility is further explored in the next section of the paper. 6. States that may have benefited from having their own currency The previous section found that a significant portion of the asymmetric effects generated by monetary policy shocks can be explained by the OCA criteria. This suggests that the 20 states that had IRFs significantly different than the United States 21 Like before, the switch from the SIC to NAICS is not an issue since the relative manufacturing and relative extractive industry shares are being used. Instead of splicing the series together at a particular year, though, the average of the relative shares calculated from both the SIC and NAICS are used for those years where there is overlap (i.e. 1990–2001). 22 The 10 major SIC industries are as follows: agriculture; mining; construction; manufacturing; transportation and utilities; wholesale trade; retail trade; financial insurance and real estate; service; and the public sector. Here, SIC and NAICS observations cannot be combined since relative values are not being used. 23 2005 is the latest year of data availability. 24 All regressions were checked for heteroskedasticity using the Breusch–Pagan Test. Those regressions found to be plagued by heteroskedasticity were reestimated with robust standard errors. Author's personal copy 743 D. Beckworth / Journal of Macroeconomics 32 (2010) 732–746 Table 3 Economic significance of OCA criteria. 12-Month response Relative manufacturing wage Unemployment persistence Industry portfolio variation Coincident indicator correlation Average AIRF 24-Month response 36-Month response Standard deviation OCA estimates Semi-partial R2 (%) Standard deviation OCA estimates Semi-partial R2 (%) Standard deviation OCA estimates Semi-partial R2 (%) 0.045 0.035 0.022 0.036 0.122 15.32 8.18 3.72 9.14 – 0.070 0.065 0.067 0.124 0.245 9.06 7.02 8.41 26.39 – 0.085 0.064 0.105 0.195 0.302 7.14 3.64 10.94 35.03 – Fig. 5. Variance decomposition of the state-specific shock. for 25 months or more may have benefited from having monetary autonomy and their own currencies over the sample period. To see if this implication is robust, two more lines of analysis are considered: a decision rule based on the regression results and an examination of the state-specific variance decomposition (VDC). Author's personal copy 744 D. Beckworth / Journal of Macroeconomics 32 (2010) 732–746 Table 4 States that could gain from having their own currency. State Significantly different Cum. IRF Sufficient variation explained by OCA criteria Idiosyncratic shocks > rest of US svg. Other considerations Gains from own currency? CO CT ID LA MD MI MT NE NJ NM ND OH OK PA SC SD TX UT WV WY Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y N Y Y Y Y Y Y Y Y Y Y Y Y Y N Y Y Y Y Y Y N Y Y N N Y Y N Y Y Y Y Y N Y Y Y Y Y Large sectoral – – Large sectoral – Rustbelt Large sectoral – Rustbelt Large sectoral Large sectoral Rustbelt Large sectoral Rustbelt – – Large sectoral Large sectoral Rustbelt Large sectoral Possibly Possibly Y Y Possibly Y Y Y Possibly Y Y Y Y Y N Y Y Y Y Y share of extractive industries share of extractive industries share of extractive industries share of extractive industries share of extractive industries share of extractive industries share of extractive industries share of extractive industries share of extractive industries First, information from columns (7) through (9) from the above regressions allows for a simple decision rule as to which of the 20 significantly different responses can be attributed to the OCA criteria. In particular, those states for which the regression models predict an AIRF that as a percent of their actual AIRF are equal to or exceed the R2 from the regression model should be counted as a state that may have gained from having its own currency. For example, since 67.89% of the variation in the 36-month AIRF can be explained by the OCA criteria, then a particular state’s predicted AIRF should be at least be that much as a percent of its actual AIRF at that horizon if it is to be meaningfully explained by the OCA criteria. Since there are three 2 horizons, the decision rule is based on the average over the three horizons as follows: if P R PN ðPAIRFi =AIRFi Þ P Ni¼1 Ni , then the state’s AIRF is explained sufficiently by the OCA criteria. Here, PAIRF is the predicted AIRF i¼1 N and i = 1, 2, and 3 are the regressions for the three horizons. Using this decision rule all the states except for Colorado and South Carolina had AIRFs that could be sufficiently explained by the OCA criteria. Second, the variance decomposition (VDC) of each state’s coincident indicator growth rate attributed to economic shocks originating from within the state can be examined to see how similar it is with other states. This measure shows the relative contribution (out of a 100%) of the state-specific shocks to the mean squared forecast error (MSFE) generated by the VAR. The larger the VDC number, the greater the percentage of MSFE that is explained by the state-specific shock. In turn, the more a state’s MSFE is dissimilar with the rest of the states the more likely it would have benefited from having its own currency. Fig. 5 presents the VDCs for the 20 states and for ease of reading splits them into Eastern and Western regions. This figure shows that state-specific shocks explain most of the MSFE early on, but then declines until about 2 years out. The thick, solid black shows the weighted average state-specific shock VDC for the other 28 states.25 This line shows that for the rest of the United States the state-specific shock explains on average about 31% of the MSFE two years out. Except for West Virginia, this response is roughly the same for the rustbelt states in the Eastern region. It is however, noticeably greater than the other Eastern states of Connecticut, Maryland, New Jersey, and South Carolina. This suggests these other Eastern states may actually be in the dollar OCA. On the other hand, all of the Western states have far more of their MSFE explained by the state-specific shock indicating there may have gained from having monetary autonomy. Interestingly, the states in the Western region were the ones that significantly better than US AIRF while the ones in the Eastern region did worse. Together, these findings lend support to the view that parts of the United States may have benefited from having their own currency. Table 4 summarizes these findings along with the results. The table also makes note of other pertinent considerations. For many of the states, it is noted that they are either part of the Rustbelt or have a large share of their economy is in the extractive industries compared to the US average.26 The last column in the table forms a tentative conclusion as to whether the state would have gained from having monetary autonomy and its own currency. A state was determined to be in this group if it either (1) met all three criteria listed in the table or (2) met two of the three criteria but had a sufficiently strong justification for being included anyways. These criteria resulted in 15 states falling outside the dollar OCA: Idaho, Louisiana, Michigan, Montana, Nebraska, New Mexico, North Dakota, Ohio, Oklahoma, Pennsylvania, South Dakota, Texas, Utah, West Virginia, and Wyoming. Of these states, only Michigan was counted under the latter criteria given it had the largest negative IRF in the nation. These 14 states can be divided into those that fall in the Western and Eastern region or what Crone (2005) calls the 25 26 The states are weighted by their size of their economies. Industry shares are calculated using sectoral shares of personal income. Author's personal copy D. Beckworth / Journal of Macroeconomics 32 (2010) 732–746 745 Energy Belt and the Rustbelt. While the findings of this paper can only speak to the potential gains from a state having its own currency, it does suggest that if these gains were greater than the increased transaction costs that came with such a move to regional monetary autonomy then the Energy Belt, which did better than the US economy, and the Rustbelt, which did worse than the US economy, would most likely form two separate currency areas distinct from the dollar union. 7. Conclusion This paper has shown that the regional asymmetric effects generated by US monetary policy over the period 1983:1 through 2008:3 can be attributed, in part, to regions of the United States being constrained by the dollar currency union and its monetary policy. These findings imply that some states may have benefited from having their own currency at this time. Two regions in particular were found to be likely beneficiaries of such monetary autonomy: the Energy Belt and the Rustbelt. The Rustbelt may have benefited over the past decade and a half from a depreciated currency coming from a looser monetary policy that would have made its manufacturing exports more competitive. Instead, it was stuck with the US dollar that until 2002 was appreciating. The Energy Belt, on the other hand, may have benefited from having a slightly tighter monetary policy. Although this paper speaks only to the potential gains of breaking up the US monetary union and not to the added transaction costs, it does imply there may be some benefit to regional monetary autonomy. A move to regional monetary policies may seem radical, but it is not something new to the United States. Between 1838 and 1860 states were able to, and in some cases did, undertake state-level monetary policy (Shambaugh, 2006).27 During the US Civil War and up through 1879 there were two currencies and thus two regional monetary policies. Yellowbacks—gold-backed dollars—circulated in the far West and Greenbacks—fiat dollars—were everywhere else (Rockoff, 2000). In 1913, the original framers of Federal Reserve designed the system so that discount lending and open market operations could be independently conducted at the district banks. The Federal Reserve System, therefore, originally allowed for monetary policy to be determined at the district level (Meltzer, 2003).28 While regional currencies in the United States presently seem unlikely, limited regional monetary autonomy could still emerge if Federal Reserve district banks were allowed more discretion in running their discount windows. Doing so would allow for regional changes in liquidity to offset national monetary policy deemed inappropriate for the regional economy.29 A final implication of this paper is that since the United States is most likely not an OCA, policies should be promoted that keep wages and output prices flexible, labor mobile, and fiscal transfers meaningful and targeted appropriately. Together, these two policies should serve to bring the United States closer to becoming a true OCA. Acknowledgement I would like to thank William Lastrapes, Danny Hughes, Lenny Gashugi, Jim Swofford, and several anonymous referees for helpful comments and suggestions. Of course, all remaining errors are mine. References Barth III, M., Ramey, V., 2001. The cost channel of monetary transmission. NBER Macroeconomics Annual 16, 199–240. Bayomi, T., Eichengreen, B., 1993. Shocking aspects of European monetary integration. In: Torres, F., Giavazzi, F. (Eds.), Adjustment and Growth in the European Monetary Union. Cambridge University Press, New York, pp. 193–229. Boivin, J., Giannoni, M., 2006. Has monetary policy become more effective? The Review of Economics and Statistics 88, 445–462. Bordo, M., 2004. The United States as a monetary union and the euro: a historical perspective. Cato Journal 24, 163–170. Carlino, G., DeFina, R., 1998. The differential regional effects of monetary policy. Review of Economics and Statistics 80, 572–587. Carlino, G., DeFina, R., 1999a. The differential regional effects of monetary policy: Evidence from the US states. Journal of Regional Science 39, 339–358. Carlino, G., DeFina, R., 1999b. Do states respond differently to changes in monetary policy? Federal Reserve Bank of Philadelphia Business Review (July– August), 17–27. Christiano, L., Eichenbaum, M., Evans, L., 1999. Monetary policy shocks: what have we learned and to what end? In: Taylor, J., Woodford, M. (Eds.), Handbook of Macroeconomics. The University of Chicago Press, Chicago, pp. 65–148. Conroy, M., 1975. The concept and measurement of regional industrial diversification. Southern Economic Journal 41, 492–505. Crone, T., 2005. An alternative definition of economic regions in the United States based on similarities in state business cycles. Review of Economics and Statistics 87, 617–626. Crone, T., 2006. What a new set of indexes tell us about state and national business cycles. Federal Reserve Bank of Philadelphia Business Review Q1, 11–24. Crone, T., 2007. Pattern of regional differences in the effects of monetary policy. Federal Reserve Bank of Philadelphia Business Review Q3, 9–19. Davis, S., Haltiwanger, J., 2001. Sectoral job creation and destruction responses to oil price changes. Journal of Monetary Economics 48, 465–512. Dowd, M., LeSage, J., 1997. Analysis of spatial contiguity influences on state price level formation. International Journal of Forecasting 13, 245–253. Eichengreen, B., 1990. One money for Europe? Lessons from the U.S. Currency Union. Economic Policy 5, 156–187. Feldstein, M., 1997. EMU and international conflict. Foreign Affairs 76, 60–73. Frantantoni, M., Schuh, S., 2003. Monetary policy, housing, and heterogeneous regional markets. Journal of Money, Credit, and Banking 35, 557–589. Irvine, F., Schuh, S., 2005. The roles of comovement and inventory investment in the reductions of output volatility. Federal Reserve Bank of Boston Working Paper No. 05-9. Kalemli-Ozcan, S., Reshef, A., Sorensen, B., Yosha, O., 2005. Net capital flows and productivity: Evidence from US states. NBER Working Papers 11301. 27 Money during this time was issued by banks which were chartered and regulated by state governments. Some state governments used this control to conduct state-level monetary policy during this time. 28 Of these tools, the discount lending was the one most often used for monetary policy at this time. District banks lost their ability to conduct regional monetary policy with the Banking Act of 1935. This act consolidated power within the Federal Reserve System at the Board of Governors. 29 Regional economic adjustments, therefore, could occur through regional quantity changes in liquidity rather than through regional price changes in exchange rates associated with having regional currencies. Author's personal copy 746 D. Beckworth / Journal of Macroeconomics 32 (2010) 732–746 Kenen, P., 1969. The theory of optimum currency areas: An electric view. In: Mundell, R., Swoboda, A. (Eds.), Monetary Problems in the International Economy. University of Chicago Press, Chicago, pp. 41–60. Kouparitsas, M., 2001. Is the United States an optimum currency area? An Empirical Analysis of Regional Business Cycles. Federal Reserve Bank of Chicago Working Paper, WP-01-22. Lastrapes, W., 2004. Estimating and identifying vector autoregressions under diagonality and block exogeneity restrictions. Economic Letters 87, 75–81. Lastrapes, W., 2006. Inflation and distribution of relative prices: The role of productivity and money supply shocks. Journal of Money, Credit and Banking 38, 2159–2198. McKinnon, R., 1963. Optimum currency areas. American Economic Review 53, 717–724. Meltzer, A., 2003. A History of the Federal Reserve, vol. 1. University of Chicago Press, Chicago. pp. 1913–1951. Mongelli, F., 2002. New views on the optimum currency area: What is EMU telling us. ECB Working Paper 138. Mundel, R., 1961. A theory of optimum currency areas. American Economic Review 51, 509–517. Owyang, M., Piger, J., Wall, H., 2005. Business cycle phases in US states. Review of Economics and Statistics 87, 604–616. Owyang, M., Wall, H., 2006. Regional VARs and the channels of monetary policy. Federal Reserve Bank of St. Louis Working Paper 2006-002A. Parkinson, D., 2007. Sunbelt versus Rustbelt in the US tug-of-war. The Globe and Mail. <http://www.theglobeandmail.com/servlet/story/ LAC.20070707.MKECOLAB07/TPStory/Business>. Partridge, M., Rickman, D., 2005. Regional cyclical asymmetries in an optimal currency area: An analysis using US state data. Oxford Economic Papers 57, 657–665. Rockoff, H., 2000. How long did it take the United States to become an optimal currency area? NBER Working Paper H124. Shambaugh, J., 2006. The American Monetary System from 1838-60: an experiment with multiple currencies. Explorations in Economic History 43, 609– 645. Sims, C., Zha, T., 1999. Error bands for impulse responses. Econometrica 67, 1113–1155. Stock, J., Watson, M., 1989. New indexes of coincident and leading economic indicators. NBER Macroeconomics Annual 351, 394. Swofford, J., 2000. Microeconomic foundations of an optimal currency area. Review of Financial Economics 9, 121–128.