The Politics of Monetary Transmission in a Large Open Economy: Congressional Pressure on. the Federal Reserve, May 1999

Transcription

1 The Politics of Monetary Transmission in a Large Open Economy: Congressional Pressure on the Federal Reserve, May 1999 J. Lawrence Broz Associate Professor of Government, Harvard University Weatherhead Center for International Affairs Harvard University 1737 Cambridge St. Cambridge, MA (617) lbroz@cfia.harvard.edu Presented to the Seminar in International Political Economy, Columbia University, March 25, I'm very grateful to Irene So, Istvan Majoros, Masako Irie, and Patricia Lindenboim for valuable research assistance.

2 1 The Politics of Monetary Transmission in a Large Open Economy: Congressional Pressure on the Federal Reserve, ABSTRACT: The distributive aspects of monetary policy are thought to be too broad and diffuse to motivate members of Congress to care much about the behavior of the Federal Reserve. Yet monetary policy has clearly defined differential effects on identifiable constituency groups via the "transmission channels" through which it operates. In a large, open economy like the U.S., monetary transmission affects the traded goods sector (import-competing and export industries) through the exchange rate channel, interest rate sensitive industries (e.g. housing) via the interest rate channel, and bank dependent borrowers (e.g. small firms) by way of the credit channel. In this paper, I test to see if these channels influence congressional attentiveness to monetary policy. The results indicate that Congress is sensitive to the distributive effects of monetary policy. For example, I estimate that a one standard deviation increase in the real exchange rate will cause the number of legislative proposals to constrain the independence of the Federal Reserve to increase by 26 percent.

3 2 1. Introduction Conventional political analyses of American monetary policy typically emphasize broad electoral and related pressures and thus see little role for Congress. The basis for dismissing Congress is that monetary policy does not involve the kind of distributive or geographic effects that normally motivate members of Congress to be attentive to economic policymaking (Alt 1991, Beck 1990, Woolley 1984). The economics literature, however, has long recognized that the "channels" through which monetary policy influences real variables have distinct distributional effects (Maisel 1968, Bach and Huizenga 1961). Research on the "monetary transmission mechanism" investigates the processes or channels through which monetary policy affects economic aggregates. While there is debate about the relative importance of the various channels, there is little doubt that the channels themselves bear differentially on sectors, groups, and regions (Mishkin 1995a, Taylor 1995, Carlino and DeFina 1997). Monetary theory, in other words, suggests reasons why Fed policy actions should be relevant to members of Congress. In this paper, I test the argument that the distributive effects of monetary transmission are sufficiently concentrated on identifiable constituency groups to have an independent impact on congressional behavior. The investigation aims to bring Congress and interest groups back into the analysis of monetary politics. This focus on monetary transmission is also relevant to the study of international political economy (Frieden 1991, Goodman and Pauly 1993, Frieden and Rogowski 1996). Over the past three decades, structural changes in the economy have altered the nature of the channels through which monetary policy affects the level of economic activity in the short run (Mauskopf 1990, Bosworth 1989). Most importantly, the end of fixed exchange rates in 1973 and integration of international capital markets have introduced a new channel to the transmission mechanism in

4 3 the U.S.: the exchange rate channel. In a large open economy under floating exchange rates, a tightening of monetary policy not only drives up interest rates but also may lead to an appreciation of the dollar exchange rate (Taylor 1993). This reduces the competitiveness of domestically produced goods, causing demand for these goods to shift abroad and exports to fall. The political impact of this change is that monetary transmission operating through net exports gives well-organized industries in the traded goods sector a direct stake in monetary policy (Frieden 1996). Thus, when the currency appreciates, these actors become politically active on issues surrounding monetary policy (Destler and Henning 1989, Frankel 1994). Monetary politics today differ in a fundamentally important way from monetary politics in the Bretton Woods era, as the international sector now carries a significant portion of the burden of the economy's adjustment to changes in monetary policy. This paper's empirical analysis is a quarterly time-series covering the period from the end of the Bretton Woods System in 1973 through Three main transmission channels operated in this period: the exchange rate channel, the interest rate channel, and the credit channel. My general argument is that interest group pressures bear on the principal-agent relationship between Congress and the Fed via the differential sectoral effects of these channels. Monetary policy transmitted through the exchange rate channel affects firms in the traded goods sector (import competing and export industries). Monetary policy operating through the traditional interest rate channel bears on interest rate sensitive sectors (housing, consumer durables), and the credit channel works by way of borrowers dependent on bank loans (small and medium sized firms). While this list of channels is neither exhaustive nor mutually exclusive, the three channels are sufficiently independent to gauge their respective influence on congressional attentiveness to monetary policy and the Federal Reserve. I measure congressional attentiveness as a weighted

5 4 count of legislative proposals to reduce the independence of the Federal Reserve. Since the observations are non-negative and discrete, I estimate the effects with a Generalized Event Count (GEC) regression model. The results indicate that Congress is sensitive to the differential effects of monetary transmission, controlling for other political and economic influences. For example, I estimate that a one standard deviation increase in the real exchange rate--which bears on firms in the traded goods sector--will cause the number of "threats to the Fed" to increase by 26 percent. In light of the findings, the dismissal of congressional oversight of monetary policy may be premature. The plan of the paper is as follows: Section 1 briefly surveys the literature on politics of monetary policy, highlighting the reasons why analysts have downplayed the relevance of Congress. Section 2 reviews the channels of monetary transmission and draws out their political implications. Section 3 describes the empirical model, the variables used in the event count analysis, and the data sources. Results are presented in Section 4 and Section 5 concludes with a discussion of implications and future research. 1. Congress and the Fed Scholarly interest in the relationship between politics and monetary policy in the United States has grown rapidly in recent years, motivated in large part by the literature on the new political business cycle (electoral and partisan). The role of Congress is not central in this literature (Grier 1991 is a notable exception) even though the Fed is a creation of Congress and Congress gives much attention to overseeing monetary policy and the Federal Reserve. 1 The electoral cycle approach begins with the assumption that presidential administrations seek a path 1 Akhtar and Howe (1991) identify over 200 bills containing 307 distinct proposals to amend the Federal Reserve Act in the 1979 to 1990 period.

6 5 for monetary policy timed to produce favorable macroeconomic conditions just prior to the next election (Nordhaus 1975, Tufte 1978). Introducing rational expectations yields the modified implication that electoral cycles should be short-lived on monetary instruments, but absent in respect to real outcomes (Persson and Tabellini 1990, Alesina and Roubini 1992). The partisan cycle approach assumes that political parties differ in their preferences for macroeconomic outcomes, which implies a relationship between the party of the president and economic outcomes (Hibbs 1987). Democratic presidents tend to tolerate more inflation to lower unemployment while Republican presidents are more likely to tolerate unemployment. Alesina and Sachs (1988) develop a rational expectations variant of the partisan business cycle model, predicting transient shocks in real variables from changes in the partisan identity of the president but no long-term link between party and output. All political business cycle models focus on the relationship between the macroeconomic preferences of politicians and macroeconomic outcomes. Whereas presidents and parties are understood to have well-defined preferences for macroeconomic outcomes, members of Congress are thought to have very weak preferences. This is because efforts to influence economic aggregates offer seemingly few electoral benefits to legislators beholden to local and sectoral interests rather than to national constituencies. Indeed, following Mayhew (1974), Woolley (1984) argues that Congress has no specifically electoral incentive to supervise the making of monetary policy. Beck (1990: 133) concurs: "A spot on the evening TV news attacking the devil himself, or his agent Paul Volcker, is never to be turned down, but there is no electoral reason for any MC [Member of Congress] actually to care about whether monetary

7 6 policy is too tight or too easy." Congressional oversight, when observed, is thus dismissed as a symbolic activity of the "blame shifting" or "position taking" variety (Alt 1991, Fiorina 1982). Kevin Grier (1991, 1996) is rare advocate of the position that Congress has an electoral stake in monetary policy. In the tradition of other studies, Grier builds his theory of congressional committee oversight of the Fed on the basis of a presumed relationship between macroeconomic preferences and macroeconomic outcomes. His idea is that inflation has sufficiently differential effects on identifiable interest groups to make it possible to treat legislators on powerful banking committees as representing either pro-inflation or anti-inflation constituencies. Yet the sectoral and geographic basis of the distributional consequences of inflation are insufficiently clear in Grier's work to justify the sort of specialization by committee required by his theory (see Alt 1991: 56). Indeed, it is difficult to accept the notion that an individual congressperson gains specific electoral benefits from pressuring the Fed on the basis of any macroeconomic objective - inflation, unemployment, or output. For legislators to focus their resources on the Federal Reserve, monetary policy ought to have the characteristics of a "distributive" activity (Fiorina 1982). My claim is that the differential effects of monetary policy transmission make monetary policy just such an activity for Congress. While there is a large literature on the politics of American monetary policy, no one to my knowledge has given systematic attention to the sectoral effects of monetary transmission in analyzing congressional activism on monetary policy. Yet the idea that policy influences the economy primarily through its impact on particular sectors has long been a standard feature of monetary theory, and leading money and banking textbooks give attention to the exchange rate, the interest rate, and the credit channels (Mishkin 1995b). The study of monetary transmission is a growth industry in economics, and research focuses on the way

8 7 specific sectors (traded-goods, housing, consumer durables, small business) respond to monetary policy shocks. By extension, sectorally-based interest group pressures could influence the behavior of the Congress with respect to monetary policy oversight. It should be noted that several studies of Congress and monetary policy have found evidence that is consistent with this view. The relevant finding is of a positive relationship between congressional activism and nominal interest rates, which is attributed to lobbying by interest rate-sensitive sectors, such as housing (Kettl 1986, Woolley 1984, Woolley and LeLoup 1989, Alt 1991, Havrilesky 1995). In a simple model using the short-term interest rate as the only explanatory variable, Kettl (1986, ) accounts for 56 percent of the variance in proposed legislation aimed at the Federal Reserve or monetary policy between 1951 and Alt (1991) extends Kettl s analysis to 1986 and obtains equivalent results. The findings conform to anecdotal evidence on Congress during the high rate periods of the late 1970s and early 1980s (Beck 1990, Pierce 1978, Weintraub 1978). Congress appears to exercise its statutory power to monitor the Federal Reserve when interest rates are high and interest rate-sensitive sectors are lobbying for action. That lobbying by a particular sector is seen as a source of congressional activism shows implicit recognition that monetary transmission has distributive effects and that Congress responds to differentially affected pressure groups. Flaws in these studies, however, raise questions about the implied role of interest groups. First, the empirical work does not control for aggregate business-cycle conditions, or any other likely influence (Corder 1998). The relationship between congressional action and interest rates might dissolve if inflation and 2 The constant in Kettl s model is 5.045, β = (p < 0.001), and R 2 = See also Woolley (1984).

9 8 unemployment are included as controls. This seems all the more likely since interest rates are typically measured in nominal terms, and thus contain inflation expectations. It is the real (inflation adjusted) interest rate that affects spending and investment decisions, by way of the interest rate channel. Second, monetary transmission is a complex process, and the operative channels have changed over time, most notably with the move to floating exchange rates and international financial integration (Mishkin 1995a). Past studies treat the U.S. as a financially closed economy and thereby do not consider this an important change in monetary transmission. 3 Monetary transmission has also been shaped by regulatory policy changes (e.g. the elimination of Regulation Q) and financial innovation, such as the rise of adjustable rate mortgages and secondary markets for packaged home and auto loans (Thornton 1994, Pozdena 1990, Kahn 1989). Thus, evidence that interest rates alone drive congressional oversight of the Federal Reserve may be anachronistic in light of these structural changes in monetary transmission. In sum, political analyses of monetary policy presume that the actions of the Federal Reserve have few electoral implications for members of Congress. Outside of the housing industry, analysts see little role for interest group pressures as determinants of congressional interest in the Federal Reserve and monetary policy. The key empirical question I address is: when does Congress become interested in the Federal Reserve and monetary policy? My argument is that members of Congress respond to those mobilized sectoral interests most directly affected by monetary policy. Identifying the affected groups is the subject of the next section. 3 The articles in Mayer (1990) assume a closed economy context. For the ramifications of international capital mobility, see Cukierman (1992), Frieden (1991), and Goodman and Pauly (1993). Woolley (1994) reviews some of the literature.

10 9 2. The Distributional Effects of Monetary Transmission in a Large Open Economy Any change in monetary policy is "distributive" in the sense that it alters the allocation of resources from what would have prevailed in the absence of the policy change. 4 Research on monetary policy transmission systematically describes and measures the types of reallocation and redistribution that follow a change in policy. In this section, I demonstrate that this research can inform a systematic political analysis of monetary policy. I discuss three transmission channels that have operated in the United States since 1973: the exchange rate channel, the interest rate channel, and the credit channel. 5 I show that each channel bears on a particular sector of the economy, which is to say that some sectors carry more of the burden of adjusting the economy to monetary policy than others. 2.1 The Exchange Rate Channel The exchange rate is a relatively recent addition to the monetary transmission mechanism in the United States and the channel with the most intriguing political implications. With the advent of flexible exchange rates in March 1973 and the international integration of capital markets, economists have given more attention to the effects of monetary policy transmission operating through exchange rates on net exports (Taylor 1995, Mauskopf 1990). From Mundell (1962), we know that flexible exchange rates in the context of international capital mobility alter the operation and increase the effectiveness of domestic monetary policy. The operational consequence is that the exchange rate becomes a key channel of monetary transmission, via net 4 The discussion assumes prices are sticky in the short run; hence monetary policy can influence output and employment at least temporarily. In the long run, however, all changes in the money supply will be reflected as changes in the price level. Long-run output and employment cannot be changed by monetary policy.

11 10 exports. In the case of the U.S., the effect is not trivial; the exchange rate accounted for roughly one-third of monetary transmission to output after 1982 (dekock and Deleire 1994). In turn, adding this new channel to the traditional mix of channels renders monetary policy more effective as a tool of macroeconomic management (Taylor 1995, Bosworth 1989, Akhtar and Harris 1987). As a large, open economy in the context of high capital mobility, the exchange rate channel complements the effects of other channels. 6 For example, a reduction in the money supply raises domestic real interest rates and restricts investment via the interest rate channel (see Section 2.2 below). Yet higher interest rates also induce an inflow of capital from abroad, causing the currency to appreciate. The appreciation makes the country s tradable goods less competitive, resulting in a deterioration of the trade balance. Monetary contraction thus decreases national output through its effect on foreign spending and the trade balance, as well as through its effects on domestic investment spending. In sum, with capital mobility and flexible exchange rates, the effect of monetary policy on the current account reinforces the interest rate effect on investment and other components of domestic demand. Monetary policy is more potent because it also involves the exchange rate. Spending on domestic output is reduced via the substitution of foreign for domestic goods, thereby reducing the demand for domestic production. 5 I do not address the effects of monetary policy on other assets prices, such as stocks and bonds, and thereby ignore the so-called "wealth effect." 6 For a small, open economy that floats its exchange rate, monetary policy operates entirely through the international sector. With perfect capital mobility, a small economy cannot maintain a risk-adjusted interest rate that differs from the world interest rate, as any differential is promptly contravened by offsetting capital flows.

12 11 The obvious distributional consequence is that the international sector is directly affected by monetary policy. Tight monetary policy, for example, causes the currency to appreciate, which switches expenditures from domestically-produced goods to foreign-produced goods. Export and import-competing industries suffer most from this switch in expenditure patterns. Conversely, depreciation stimulates spending on domestic goods by lowering their relative price. Export and import-competing sectors tend to gain from depreciation relative to consumers of imported goods and firms that are heavy users of imported inputs. More generally, when the exchange rate channel is operative, the traded-goods sector is directly exposed to the effects of monetary policy, implying that industries in this sector will have a stake in monetary policy decisions. Frieden (1996) suggests that monetary politics will come to resemble trade politics in this setting, as both policies implicate the same constituencies. Yet the analog to trade politics is complicated by two factors. First, coalitional patterns on trade issues can and do cut across the traded goods sector: export industries typically lobby for freer trade while import-competing industries tend to lobby for protection (Magee 1989). The exchange rate, by contrast, produces no such division in the international sector as export- and import-competing industries both benefit from currency depreciation. Second, trade policy can be narrowly targeted to create rents for specific industries in a sector (Gowa 1988). Unlike a tariff or an export subsidy, the exchange rate has an indiscriminate affect across industries in the sector. From a collective action standpoint (Olson 1965), both differences imply that interest group activity will be more limited with respect to the exchange rate since it is a less "excludable" good than trade policy. Although the barriers to interest group lobbying on exchange rates are higher than in trade policy, they are not insurmountable. A growing literature covers periods in which the

13 12 monetary policy debate was organized around open economy pressure group politics (Frieden 1997, Frankel 1994, Destler and Henning 1989, Hefeker 1995, Eichengreen 1993). A plausible explanation can be derived from the "privileged group" model of collective action (Olson 1965). If the gains from lobbying on exchange rates are not distributed symmetrically across industries in the sector, then pressure group activity cannot be ruled out. With asymmetric gains, some industries will care more than others about a policy action that leads to appreciation (industries heavily dependent on exports). When the gains of lobbying outweigh the costs, these industries will lobby the government, even if other beneficiaries in the sector free ride. Exchange rate politics in the 1980s and elsewhere appear consistent with this logic as certain extremely tradedependent industries were very active in "dollar politics" (Destler and Henning 1989, Frenkel 1994). Another factor aiding collective action is that most industries in the sector have preestablished facilities in Washington devoted to shaping trade policy, which lowers the start-up costs for lobbying Congress on monetary policy. Despite these complications, Frieden's basic point is incontrovertible: with floating exchange rates and high capital mobility, monetary policy bears on the international sector. This gives rise to the following prediction: Given the political prominence and collective action potential of large import-competing and export industries on Capitol Hill, congressional attention to monetary policy and the Federal Reserve should be positively related to the level of the real exchange rate. Note that the prediction assumes that lobbying is asymmetric; that lobbying from the "winners" of real appreciation does not counteract pressure from the "losers". The assumption is based on the following considerations: The effect of a real appreciation is to make tradable goods cheaper in the domestic currency. This harms the profitability of domestic producers of import-

14 13 competing and exported goods. Yet even producers of non-tradables are adversely affected, since the change in relative prices induces a switch from consumption of non-tradables to tradables. Who then are the beneficiaries of currency appreciation? First, foreign producers gain but are prohibited by law from lobbying Congress. Second, domestic consumers benefit but face high barriers to collective action and thus do not lobby. While traded goods industries that use imports as intermediate goods benefit from real appreciation, they are harmed by the effect of appreciation on their profits (Campa and Goldberg 1997). The asymmetry in lobbying incentives and capabilities is thus closely aligned with the division into gainers and losers, which is why real exchange rate misalignments tend to be resolved between countries rather than via domestic lobbying (Funabashi 1988). 2.2 The interest rate and credit channels The exchange rate channel is not the only operative channel in the U.S. The other channels also have distributional effects that imply interest group politics. Consider the conventional interest rate channel, wherein changes in monetary policy (open market operations) influence interest rates and thereby the cost of capital. Policy-induced changes in real interest rates discourage or encourage investment spending, and thereby affect aggregate demand and output. For example, if the Fed tightens monetary policy, short-term nominal interest rates rise more or less immediately, reflecting the reduced supply of bank reserves in the market. An increase in nominal rates results in an increase in real rates over the time period in which prices and expectations are adjusting. Sustained increases in short-term rates lead to higher real longterm interest rates on such instruments as home mortgages, consumer loans, and corporate bonds. 7 Higher long rates then depress demand for housing, cars, and other consumer durables,

15 14 and additional plant and equipment. In short, this transmission channel affects real economic activity by way of interest-sensitive expenditures. Industries respond differently to changes in real interest rates because interest rate elasticities differ across industries (Maisel 1968). Empirically, the earliest and largest decline in demand occurs in residential housing, with spending on consumer durables and fixed business investment declining less and with longer lags (Bernanke and Gertler 1995; deleeuw and Gramlich 1969). The sectoral variation in interest-sensitivity depends on the proportion of purchases made with credit, the amount of credit required per unit of expenditure, the institutional character of the market, and the degree to which lenders are influenced by policy changes. Housing and construction rank highest in sensitivity on all these accounts, followed by consumer durable industries (e.g. autos). Indeed, restrictive policy has been found to fall earliest and heaviest on these sectors (Carlino and DeFina 1997, Mauskopf 1990). Note that interest-sensitive sectors are also politically organized, with powerful trade associations operating in Washington. 8 Moreover, residential construction is an important component of expenditures in all states, implying that Congress will be especially sensitive to the plight of this sector. Given the national political importance of these industries, domestic political activity on monetary policy should be positively associated with real interest rates. 7 The relationship between short rates and long rates is based on the expectations model of the yield curve, wherein the long rate is determined by the expected future path of short rates. 8 The National Association of Home Builders, for example, currently has 195,000 members, onethird of whom are homebuilders. The other members are in closely related fields, such as mortgage finance and building products and services.

16 15 Lobbying by the interest-sensitive sector during periods of high real interest rates is, in turn, expected to increase congressional concern for monetary policy. 9 A third transmission conduit is known as the credit channel. As distinct from the costs of borrowing, tighter monetary policy can also reduce the supply of bank credit through increases in the perceived riskiness of private loans or through higher funding costs. The credit view asserts that in addition to affecting real interest rates, monetary policy affects aggregate demand by altering the availability or terms of new bank loans (Bernanke 1993). Analysts further distinguish between the bank lending channel and the broad lending channel, yet both require asymmetry in the access of small and large firms to bank credit (Mishkin 1995a, Hubbard 1995). The bank lending view stresses that some borrowers (small firms) are dependent upon banks for external finance, and that monetary policy can thereby work directly on the supply of bank loans (Bernanke and Blinder 1992). 10 When banks are subject to reserve requirements, monetary tightening drains reserves, thus decreasing the ability of banks to lend. As a result, credit to bank-dependent borrowers falls, causing these borrowers to curtail spending. Large firms, by contrast, can directly access credit through the commercial paper, stock, and bond markets without going through banks. The burden of a decline in bank loans following a monetary contraction is thus borne by smaller firms (Bernanke and Gertler 1995). 9 See Kettl (1986, 187) for anecdotal evidence. On one occasion, homebuilders mailed to the Fed 2 X 4 boards by the thousands as a protest for lower interest rates. 10 The bank lending channel requires that there are firms for which bank loans are the only source of external funds. Credit market imperfections (asymmetric information) leave some firms incapable of issuing securities, and banks fill this void by specializing in gathering information and monitoring such borrowers.

17 16 The broad lending (or financial accelerator) view emphasizes the impact of policy changes on the balance sheets and creditworthiness of borrowers (Bernanke, Gertler, and Gilchrist 1996). For example, tightening monetary policy can reduce the creditworthiness of borrowers by increasing interest expenses, reducing net cash flow, and lowering asset values. With fewer eligible borrowers in the market, lending contracts, leading to a reduction in aggregate demand. Like the bank lending channel, the broad lending channel affects smaller, financially weaker firms disproportionately. Unlike large firms that can access other forms of credit to smooth a policy-induced drop in their cash flows, small firms account for a heavier share of the decline in output that follows a monetary shock (Gertler and Gilchrist 1994). Both credit channels have much the same distributional impact: small businesses that depend on bank loans to maintain production and investment and to finance inventories bear the brunt of a policy-induced squeeze in bank credit (Kashyap, Stein and Wilcox 1993). Small firms are, of course, a potent lobby in Washington. 11 My prediction is that pressure from this sector during periods of tight money (when the supply of bank credit contracts relative to non-bank credit) will show up in increased small business pressure on Congress. As monetary contraction shifts the composition of business finance away from bank loans toward internal (nonintermediated) finance, small bank-dependent firms will pressure Congress for relief. In summary, monetary policy transmission has differential effects as it works through the economy, implying pressure group analysis of monetary politics. In an open economy with floating exchange rates, monetary transmission occurs through the real exchange rate, the real 11 The U.S. Chamber of Commerce is the world's largest business federation representing more than three million businesses and organizations. More than 96 percent of the Chamber's members are small businesses with 100 or fewer employees and 71 percent have 10 or fewer employees.

18 17 interest rate, and the supply of bank loans. The exchange rate channel gives industries in the international sector a direct stake in monetary policy. The involvement of the traded-goods sector shifts the politics of monetary policy (somewhat) in the direction of trade politics, as it implicates a price that this sector cares about deeply. The interest rate channel bears heaviest on industries that are most sensitive to interest rates, such as housing. The credit channel tends to work through the small business sector. The general point is that the effects of monetary policy are not distributed evenly throughout the economy: the international sector, the interest-sensitive sector, and the small business sector carry disproportionate shares of the burden. It is difficult, however, to directly test the political implications of these arguments. On the one hand, accurate measurement of sectoral political activity is probably impossible (Havrilesky 1995, ). On the other, clear distinctions regarding sector sensitivities to the transmission channels tend to break down once it is realized that some interest-sensitive industries are also sensitive to exchange rates e.g. autos, aircraft, and capital goods generally (Frankel 1994). However, the sector most sensitive to interest rates is residential housing, which means this overlap does not cripple the argument. Given these problems I focus on aggregate congressional behavior. The following discussion lays out a procedure for evaluating the predictions by way of congressional activity on monetary policy. 3. Data Description and Empirical Specification Congress is constitutionally the final arbiter in the conduct of monetary policy. The Federal Reserve System was created by an Act of Congress in 1913, and Congress continues to

19 18 guard its prerogative to oversee the System and to supervise monetary policy. 12 Since data on sectoral monetary policy preferences and political activity are not available, I analyze the activities of Congress with respect to monetary policy and the Federal Reserve. Congressional attentiveness to monetary policy can serve as a proxy for the sectoral implications of the argument because members of Congress are tied to constituents via campaign contributions and elections. I expect congressional action with respect to monetary policy whenever legislators reasonably anticipate an electoral payoff. McCubbins and Schwartz (1984) provide one of the best discussions on this topic. Congress sets up fire alarm mechanisms that allow constituents to make it clear when an agency is doing something that might cause members of Congress harm. This is more efficient for Congress than engaging in costly direct oversight. In regard to monetary policy, the fire alarms are likely to take the form of sustained lobbying by interest groups harmed by current monetary conditions. The data are a quarterly time-series covering the floating exchange rate period (1973II III). The dependent variable is congressional attentiveness to monetary policy, ATTNFED, measured as a weighted count of formal legislative proposals to alter the status of the Federal Reserve or to mandate specific monetary policy objectives. Kettl (1986), Alt (1991), and Havrilesky (1995) measure congressional attentiveness to monetary policy in a similar fashion. I add weights to legislative proposals for which congressional hearings are held as a means to gauge the "seriousness" of proposals. To elaborate, formal legislative proposals are defined as bills, simple resolutions, joint resolutions, and concurrent resolutions. The series includes all proposals in these forms that 12 For principal-agent models of the relationship between Congress and the Federal Reserve, see Mayer (1990), Alt (1991), and Grier (1991, 1996).

20 19 address: (1) the membership or tenure of the Federal Reserve s monetary policy decision-making bodies; (2) the provision of information on monetary policy decisions and decision-making procedures; and (3) the Federal Reserve s policy instruments, objectives, and final goals. Proposals that deal exclusively with the regulatory role of the Federal Reserve are not included (see Appendix A for more detail on the coding criteria). I weigh legislative proposals for which congressional hearings are held twice as heavily as proposals without hearings. While the weighting procedure is purely arbitrary, the operation picks up some of the difference in the seriousness and costs of legislative proposals. Hearings require a larger commitment of individual and collective resources than proposals, and so give a sense of the seriousness of proposals in terms of opportunity costs. Hearings may also indicate that a proposal has support beyond the legislator(s) who introduced it, and thus help to identify bills that have a higher likelihood of becoming law. Woolley and LeLoup (1989) employ a similar weighting strategy. This measure does not capture all aspects of congressional oversight of monetary policy, however. Some forms of congressional oversight of monetary policy do not appear as legislative proposals. Discussions occur as congressional hearings unconnected to proposed legislation and congressional studies and reports occasionally address monetary policy and the organizational structure of the Federal Reserve System. Additionally, there are formal (e.g., Humphrey- Hawkins testimony) and informal interactions between legislators and monetary authorities. My measure does not include these non-legislative activities. Despite these omissions, ATTNFED should pick up variation in congressional attentiveness to monetary policy reasonably well. Given the Fed's budgetary independence, threats to alter the Federal Reserve Act may be the most effective way for Congress to signal its monetary policy preferences to the Fed. My

21 20 intention, however, is not to show that Congress controls the Fed by way of proposed legislation or any other oversight procedure. Rather, the immediate goal is to explain the conditions under which Congress becomes interested in the Federal Reserve and potentially influential. 3.1 Variables of Interest The central prediction is that congressional activity is sensitive to the distributional (sectoral) effects of monetary policy via three channels of policy transmission. I use continuous variables to measure these channels, each lagged one period to allow for the time it takes for lobbies to organize and for the delay in the congressional response. 13 To capture the impact of exchange rates on traded-goods industries the "open economy" channel I use the real effective exchange rate of the dollar, REER. The real exchange rate is what matters to individuals when choosing whether to buy a U.S. or a foreign product since it captures the relative price of goods at home and in a foreign country. The real effective exchange rate (or trade-weighted real exchange rate) is the weighted average of the real exchange value of the dollar against the currencies of major U.S. trading partners. Since it includes a basket of currencies, REER provides a good sense of the international competitiveness of the American traded-goods sector. I use the new "Major Currency" index from the Federal Reserve Board, which is calculated as the inflation-adjusted value of the dollar relative to the real values of the currencies of the nation's top 16 trading partners. 14 I then divide the index number by 10 to aid in the interpretation of the results Longer lags of two and three periods do not alter the signs or have much impact on the size of the coefficient estimates, but significance levels are reduced for some terms. 14 For a description of the Fed's indexing and weighting procedures see, Leahy (1998). Other real dollar indexes (International Financial Statistics, Federal Reserve Bank of Dallas) have little

22 21 To examine the effect of interest rates, I use the real one-year Treasury bill interest rate, REALTBIL. Short-term rates are most directly affected by the Federal Reserve and real interest rates are what influence household and business spending decisions. 16 The variable is calculated as the one-year nominal T-bill rate minus the mean expected inflation rate one-year forward. Inflation expectations are from the University of Michigan's Surveys of Consumers. Lastly, I evaluate the impact of the credit channel by way of the variation in the "mix" of finance, FINMIX, which is measured as the ratio of commercial bank loans to the sum of commercial bank loans and commercial paper (following Kashyap, Stein, and Wilcox 1993). Kashyap et. al. (1993) found that a monetary contraction was associated with a decline in this ratio, which is consistent with the predictions of the credit channel. Their intuition is that if a monetary contraction leads to a decline in this ratio, the supply of bank loans must be shrinking since the demand for both types of finance should fall in rough proportion to one another. Since small firms cannot issue commercial paper as readily as large firms, a decline in the ratio is also a rough proxy for the differential financial pressure facing the small business sector. I expect the coefficient for FINMIX to be negative; an increase in the ratio of bank loans to commercial paper means financial relief to bank-dependent small businesses, and thus Congress should respond with fewer proposals targeting the Fed and monetary policy. impact on the results since they are closely correlated with the Fed Board index (IFS: r =.99, Dallas: r =.81). 15 The Fed's currency index base is set at 100. Dividing by 10 scales the variable to the level of the other explanatory variables. 16 See Darin and Hetzel (1995) for the problems of measuring ex-ante real interest rates.

23 Control variables There are two procedural features of Congress that influence the volume of proposed legislation. First, Congress runs on two-year sessions, and legislation introduced in one session does not carry over to the next session. Proposals not voted into law expire at the end of a congressional session. This procedure leads to far more proposals early in a new session. In fact, thousands of measures are introduced in the first weeks of a new session as members perceive the pressures of time and other proposals and realize that it may take two full years for their proposals to be enacted into law. This pattern is strikingly evident in the annual data displayed in Figure 1. More bills and resolutions are introduced in the first (odd-numbered) year of a new Congress, more than the difference in volume of business in the two years. Though quarterly data are not available on total legislative activity, the rush of proposals typically occurs in the first three months of a new session. Therefore, I include a dummy variable, NEWCONG, in the regression to control for this aspect of the political process. The variable takes the value of 1 if the observation is in the first quarter of a new Congress, zero otherwise. --Figure 1 about here-- Second, procedures governing the sponsorship of legislative proposals also influence the volume of legislative activity. In Figure 1, note the sharp drop in total legislative activity after Before that year, congressional rules so restricted the number of co-sponsors allowed on an individual bill that many members introduced identical versions of popular proposals. In 1978, unlimited co-sponsorship was allowed and the new rule sharply reduced the total number of bills (Wilson and Young 1997). As Mayhew (1974, 63) notes, sponsorship of a bill is a

24 23 relatively inexpensive signal to constituencies, suggesting that unlimited cosponsorship would reduce the volume of bills introduced. Indeed, before 1978, the average annual number of legislative proposals was 12,136. After 1978, the mean volume dropped to 5,450. To control for this change in sponsorship procedures, I include a dummy variable, COSPON, which takes the value of 1 for all observations before 1979, zero otherwise. I also include two macroeconomic variables to control for business cycle effects on congressional activism. Some evidence suggests that aggregate economic conditions affect the electoral success of congressional incumbents (Tufte 1975, Kieweit 1983, Erikson 1990). Therefore, members of Congress may respond to "bad" macroeconomic conditions by giving more attention to monetary policy and the Federal Reserve. Variation in the dependent variable might then be explained without any direct reference to monetary transmission or lobbying by sectors affected by the channels of transmission. Evidence to this effect would support the conventional "blame shifting" and "position taking" view of congressional attentiveness (Beck 1990, Alt 1991). To measure macroeconomic conditions, I use the inflation rate, INFLA, and the unemployment rate, UNEMP. 17 After running the basic model, I control for congressional elections and partisan control of Congress to see whether factors underlying conventional political business cycle analysis influence congressional attentiveness to the Fed. To examine the role of elections, I created a dummy variable, ELECTION, coded 1 for the two quarters prior to an election, zero otherwise. For partisan effects, I use GOPCONG, a dummy variable for Republican control of both houses of Congress, and DIVCONG, a dummy for divided control of Congress. The omitted reference 17 Since I have subtracted inflation expectations from the interest rate variable (by using the real rate), there is less chance of a serious collinearity problem between INFLA and REALTBIL.

25 24 category is thus Democratic control of Congress. To the extent that the parties have different preferences for monetary policy, they should also have different preferences for the independence of the Federal Reserve. Thus, when Republicans control Congress, there should be fewer legislative threats to the Fed than when Democrats are in power. Table 1 contains a description of the variables and data sources. Summary statistics are reported in Table 2. --Tables 1 and 2 about here Model Specification As the dependent variable is a count that takes on positive integer values for the number of occurrences in a fixed time interval (quarter), ordinary least squares is an inappropriate estimation technique. With event count processes, the error terms cannot be normally distributed because the observations are non-negative and discrete. Nor can the expected value of the observed variable be a linear function of the co-variates because linear functions permit negative values. Consequently, least squares analyses are inefficient estimation techniques (King 1989). Therefore, I estimate the effects of the monetary transmission channels with a Generalized Event Count (GEC) regression model. 18 The GEC model is attractive because it is consistent in the presence of unknown forms and levels of "dispersion". 19 While event count models have been 18 To estimate the GEC model, I use the "Count" program developed by Gary King. This program is available on-line at 19 Dispersion concerns the independence and variance of events occurring within an arbitrary period of observation. "Overdispersion" occurs when one event makes other within-observation events more likely. "Underdispersion" occurs when one event makes other events less likely. Rather than assume the type of dispersion in my data I have no clear priors on whether one

26 25 developed to deal with specific forms of dispersion (e.g., Poisson and the Negative Binomial models), they impose a priori assumptions about the unobserved processes generating the event counts. In the absence of prior knowledge on the type of dispersion, the GEC model is the most appropriate estimator since rather than assuming the type of dispersion, it estimates a separate parameter that allows from any form and produces estimates and standard errors that are efficient and consistent. 4. Results Regression results are reported in Table 3. Model 1 is for the entire Post-Bretton Woods period; Models 2 and 3 split the sample differently to account for changes in international capital mobility. The dependent variable is congressional attentiveness to the Federal Reserve and monetary policy. The variables of interest are REER, REALTBIL, and FINMIX. These terms capture the differential effects of monetary policy operating through the exchange rate, the interest rate, and the credit channels, respectively. I expect REER and REALTBIL to have positive coefficients; FINMIX should be negative. NEWCONG and COSPON are variables that control for the influence of congressional procedures on bill proposals. INFLA and UNEMP are controls for business cycle (macroeconomic) conditions; the relationship is expected to be positive in keeping with the more general finding that aggregate economic conditions affect electoral prospects. Since collinearity between the aggregate and the price variables seems an obvious concern, I report the co-variance matrix in Table 4. Surprisingly, little multicollinearity is evident, perhaps because the price variables affect macroeconomic conditions with various lags. Lastly, including a lagged dependent variable as a regressor has almost no impact on the legislative proposal targeting the Fed makes additional proposals more or less likely in any given quarter I use the GEC model to measure it.

27 26 results (all parameter estimates have the same sign and significance levels), indicating that serial correlation in the ATTNFED series is not a problem. --Tables 3 and 4 about here-- The sign of the relationship for the REER parameter estimate indicates that as the dollar appreciates, Congress becomes more attentive to monetary policy. The estimate is highly significant (p =.008). The result bears out a key intuition of the paper: floating exchange rates and financial integration have led Congress to be sensitive to the exchange rate's impact on the international sector in its dealings with the Fed. When the real value of the dollar appreciates, Congress feels pressure from the international sector, and congressional activity on monetary policy increases. Furthermore, the positive sign supports the claim that there are strong asymmetries in lobbying on exchange rates; neither tradables nor non-tradables producers gain from real appreciation, and the beneficiaries (consumers, foreign producers) do not lobby. The results also provide evidence of a political effect for the interest rate channel. The sign of the REALTBIL estimate indicates that Congress responds to increases in the real interest rate with more "threats" to the Federal Reserve. The REALTBIL coefficient is positive and significant at the.05 level, suggesting that high real interest rates induce Congress to be attentive to monetary policy by way of pressure from the economy s interest-sensitive sectors. This result is in keeping with earlier work on Congress and monetary policy (Kettl 1986, Woolley 1984). However, I think these results are more compelling because (1) the interest rate effect is