Economic Dynamics Newsletter

Larry Christiano and Martin Eichenbaum write about their current research program on the monetary transmission mechanism

Much of our research focuses on the nature of the monetary transmission mechanism. In part this reflects our desire to understand the interaction between real and monetary phenomena. And in part, this reflects our interest in policy questions, such as: what causes periodic outbreaks of high inflation and what are the causes of currency and banking crises? Making progress on questions like these requires empirically plausible models of the monetary transmission mechanism. To construct such models, we have looked at two types of evidence: US time series data, as well as evidence from countries that have experienced currency crises. In what follows we discuss what we have learned from these two types of data and our ongoing efforts to model the monetary transmission mechanism.

Evidence From US Time Series

One branch of our work develops and implements econometric strategies for selecting between competing models of the monetary transmission mechanism. In working with US data, the strategy that we have emphasized is a limited information one. Specifically, it involves comparing the response of a model economy and the actual economy to a particular shock. To be useful for diagnostic purposes, the shock must satisfy two properties: different model economies must react differently to it, and we must know how the actual economy responds to it. In our view, monetary policy shocks satisfy both properties. This strategy has been implemented using other kinds of shocks. For example, Gali (1999) studies the effects of technology shocks, while Burnside, Eichenbaum & Fisher (1999a,b), Edelberg, Eichenbaum & Fischer (1999), Ramey & Shapiro (1999) and Rotemberg & Woodford (1992) consider the effects of shocks to government purchases. It is well known that different model economies react differently to monetary policy shocks, see Christiano, Eichenbaum & Evans [thereafter CEE] (1997). Moreover, much progress has been made in understanding the economic impact of a monetary policy shock. As a result, we know a great deal about what it will take to have an empirically plausible model of the monetary transmission mechanism.A second branch of work focuses more narrowly on particular episodes such as the high inflation in the 1970’s. Our work on this begins with Chari, Christiano & Eichenbaum (1998). Christiano has pursued this line further with co-authors.

The Facts

In a series of papers, we have argued that the key consequences of a contractionary monetary policy shock are as follows: (i) interest rates, unemployment and inventories rise; (ii) real wages fall, though by a small amount; (iii) the price level falls by a small amount, after a substantial delay; (iv) there is a persistent decline in profits and the growth rate of various monetary aggregates; (v) there is a hump-shaped decline in consumption and output; and (vi) the US exchange rate appreciates and there is an increase in the differential between US and foreign interest rates. See CEE (1998b) for a discussion of the literature and the role of identifying assumptions that lie at the core of these claims.

Implications for Models

What kind of models are consistent with the facts? Some models obviously are not.

Monetized RBC, Misperception and Pure Sticky Wage Models

By a monetized Real Buiness Cycle (RBC) model, we mean an RBC model with no frictions other than a cash in advance constraint or a transactions technology for purchases that involves real balances. Simple monetized RBC models cannot account for the sluggish response of prices or the rise in the nominal interest rate that follow in the wake of a contractionary policy shock. The `monetary misperception’ model (Lucas, 1972) is also inconsistent with the facts. Its problem is that movements in the price level are central to the mechanism by which a monetary shock is transmitted to the real economy. In reality, the price level isn’t the first thing to move. It’s the last thing. Models in which sticky nominal wages are the sole friction also run afoul of this last observation. In those models, output falls after a contractionary monetary policy shock because a fall in the price level raises the real wage. There are two problems with this story: prices don’t fall and the real wage doesn’t rise.

Pure Sticky Price Models

What about models in which the only friction is stickiness in goods prices? These models too, cannot account for the facts. In our work, we have emphasized the failings of these models for profits, prices, output and interest rates. In CEE (1997) we argued that these models predict that profits rise after contractionary monetary policy shocks. In effect, these models imply that business people should be crying out for a surprise monetary contraction! This peculiar implication for profits reflects the sharp fall in wages and capital costs that occurs as output falls.A closely related problem is the inability of sticky goods price models to reproduce the observed degree of price inertia and persistence in output after a monetary policy shock. In CEE (1997) we argued that both shortcomings reflect the absence of strategic complementarity in price setting. To be specific, imagine that an expansionary monetary policy shock occurs in a particular period. If all prices are sticky, then the shock will have no effect on the price level and a large effect on output. Now consider the next period, and imagine that a fraction x of the firms can reset their prices. If x were unity, prices would fully adjust and output would revert to its pre-shock level. It turns out that the outcome is not very different when x is small. The reason is simple.

The 1-x firms with preset prices must expand output. In the act of doing so, they drive up factor prices and, therefore, marginal costs. The x firms resetting prices, respond to the higher marginal costs by raising prices substantially and reducing output. This effect is greater, the smaller is x. Put differently, the lower is x, the less is the incentive of flexible price firms to act like the firms with preset prices. This absence of strategic complementarity among price setters means that the response of aggregate prices and output in the period after a shock is roughly the same whether x is unity or smaller.

An additional problem with sticky goods price models is that they have difficulty accounting both for the liquidity effect and the hump-shaped fall in consumption that are associated with a contractionary monetary policy shock. This is because, absent asset market frictions, the standard intertemporal Euler equation holds in these models. For standard specifications of utility, this implies that a hump-shaped fall in consumption is associated with an immediate fall in the real interest rate. With persistently sticky prices, this translates into an immediate fall in the nominal interest rate. But, in the data monetary contractions are associated with a rise in the nominal interest rate.

Limited Participation Models

A different class of models that we have considered stress frictions in households’ portfolio decisions (see Lucas (1990) and Fuerst (1992)). The effect of this friction is that a monetary contraction creates a shortage of liquidity in the financial sector, reducing the supply of loanable funds. The resulting rise in interest rates induces firms who need working capital to cut back on their scale of operations and aggregate output declines. In many ways, these models can be thought of as stylized representations of the type of credit market frictions emphasized in the lending channel model of the monetary transmission mechanism, see for example Kashyap, Stein & Wilcox (1993) and Bernanke, Gertler & Gilchrist (1999). In effect, these credit market frictions mean that a contractionary monetary policy shock acts on the economy like a negative supply shock. Of course, the model also imbeds the standard channel of monetary policy, which operates via demand.Versions of these models that we have worked with can reproduce a number of features of the stylized facts discussed above. Because they break the standard intertemporal Euler equation they are in principle compatible with the liquidity effect and hump shaped decline in consumption after a contractionary monetary policy shock. These models can generate some sluggishness in prices and persistent movements in output. To see this, think of a contractionary monetary policy shock as a negative shift to the nominal demand for goods. Other things equal, this reduces the price level. But, it also reduces the supply of goods. Other things equal, this raises the price level. So, if the demand and supply effects are equally strong, there is a substantial decline in output and little movement in the price level, see CEE (1997).

This being said, there are at least two important problems with such models: for the supply side effect to be strong enough to cancel the demand side effect one must appeal to a counterfactually large labor supply elasticity and markup. And even then, these models cannot account for the observed degree of sluggishness in the price level.

Implications for Future Research

The discussion above suggests that single friction models will not provide a convincing account of what happens after a monetary policy shock. The obvious question is: what combination of frictions will? To provide a quantitative answer to this question, we are constructing and estimating a model which incorporates limited participation constraints and sticky prices. To place these on a symmetric footing, we do two things. First, we model sticky prices in the way suggested by Calvo (1983). Second, we capture the portfolio restrictions implicit in the limited participation assumption in the same way. Specifically, we imagine that households can change their money holdings at stochastic intervals of time.We also think it is important to allow for labor market frictions into this extended framework. This is because the problems of both the limited participation and sticky goods price models revolve in part around their labor market implications. The first model relied on implausibly high labor supply elasticities, while the second model foundered on the rock of highly cyclical marginal costs.

One way to deal with both problems is to assume that wages are sticky too. In practice, we will model these using Calvo-style wage contracts. Perhaps a more interesting route is to allow for efficiency wages, modeled along the line of the general equilibrium shirking model in Alexopoulous (1998). An attractive feature of that framework is that real wages do not move in response to a monetary policy shock, and changes in employment involve only the extensive margin of unemployed workers.

The Causes of High Inflation

An important policy question is, why has inflation alternated between periods of being high and periods of being low. Christiano has studied this question in recent work with co-authors (Stefania Albanesi and V.V. Chari in one case, Christopher Gust in another).Christiano & Gust (1999) are studying the Great Inflation of the 1970s. They consider the hypothesis that this episode was due to an increase in expected inflation that became self-fulfilling because of the nature of monetary policy. They explore this hypothesis in the context of two models, a sticky price model constructed by Clarida, Gali & Gertler and the limited participation model of CEE (1998a). Christiano & Gust (1999) find that the limited participation model is a more reasonable explanation of this episode because it is also consistent with the stagflation of the time: inflation was high and output and employment were low.

Albanesi Chari & Christiano (1999) study the potential for the Inflation Bias Hypothesis to account for the high and variable inflation that has been observed. This hypothesis, originally due to Kydland & Prescott and Barro & Gordon, is that high inflation reflects a lack of commitment in monetary institutions. Albanesi, Chari & Christiano begin by exploring the magnitude of this bias in standard sticky price and limited participation models. They find that, for a surprisingly wide range of parameter values, the models predict that the inflation bias is zero: models with and without commitment predict the same inflation. So, these models imply that the inflation bias hypothesis is rejected: it cannot account for the high observed inflation rates. They argue that this result reflects in part the poor implications of these models for money demand. When money demand is modified in a plausible direction, the model has the potential to resolve several classic, outstanding puzzles in the money demand literature. At the same time, the inflation bias hypothesis is restored with these modifications. In ongoing work, the authors are exploring in greater detail the empirical implications for money demand of their model. They are also exploring the empirical performance of the version of their model that is potentially consistent with the volatility of inflation.

Evidence From Currency Crisis Countries

Another strand of our research is motivated by evidence on the nature of currency crises in the post-Bretton Woods era. In our view, the key characteristics of these crises are: (i) they have increasingly coincided with banking crises, (ii) there are implicit government guarantees to domestic and foreign bank creditors prior to these crises, (iii) banks generally do not hedge exchange rate risk and many go bankrupt following a currency devaluation, (iv) a lending boom precedes the crises, (v) after twin banking and currency crises, domestic interest rates and the current account rise sharply, while output declines dramatically. Accounting for these observations is obviously an important task. At the same time, any theory which can account for these facts may also shed light on the monetary transmission in countries like the US. In ongoing work, Eichenbaum and co-authors (Craig Burnside and Sergio Rebelo) have been exploring the role of government guarantees to foreign and domestic bank creditors in precipitating twin banking and currency crises. They proceed in three stages. First they have explored the effects of banking crises on currency crises. Second they study the effects of currency crises on banking crises. Finally in ongoing work they are endogenizing both types of crises as reflecting government policy vis a vis the banking system.Burnside Eichenbaum & Rebelo [thereafter BER] (1998) argues that banking crises caused the recent currency crises in Southeast Asia. It begins from the assumption that the banks were in trouble in Southeast Asia prior to the currency crises. According the BER model, private agents realized this and assumed that governments would eventually fund bank bailouts at least in part via seigniorage. It implies that the fixed exchange rate regime should have collapsed after agents understood that the banks were failing, but before governments actually started to monetize their deficits. In this way, their explanation of the Asian crisis lays the blame at the feet of bad government policy. Yet, the model is also consistent with the fact that standard indicators of bad government policy – high deficits, excessive money growth, etc. – were not observed prior to the crises.

BER (1999a) takes as given the probability of an exchange rate collapse and investigates how government guarantees to domestic and foreign bank creditors can convert a currency crisis into a banking crisis. According to BER’s model, when there are government guarantees, it is optimal for banks to not hedge foreign exchange risk. Indeed, it may be optimal for banks to magnify their exchange rate exposure. According to their analysis, prior to an exchange rate collapse, government guarantees act like a subsidy to domestic loans and reduce the interest rate that firms must pay to fund their ongoing operations. This leads to higher levels of output and economic activity than would be the case absent government guarantees. But, if there is a currency devaluation, it is optimal for banks to declare bankruptcy and renege on their foreign debt. Moreover, domestic interest rates rise and there is a permanent decline in wages, employment and output. In short, a currency crisis leads to a banking crisis and a permanent decline in aggregate economic activity.

Finally, BER (1999b) endogenizes both types of crises as arising from government guarantees to banks’ creditors. The key argument is that in the presence of government guarantees to banks’ creditors, a currency crisis transforms potential government liabilities into actual government liabilities. This opens up the possibility of self-fulfilling currency attacks. In BER’s model self-fulfilling currency attacks are not possible in the absence of government guarantees. But, in the presence of such guarantees, self-fulfilling currency attacks occur almost surely. So, both fundamentals and expectations play important roles in currency crises. The former determine whether twin currency/banking crises can occur. But, the latter determine the precise timing of the crises.

In work with co-authors (Christopher Gust and Jorge Roldos), Christiano studies the nature of the monetary transmission mechanism in the aftermath of financial crises. This work does not investigate the causes of currency crises, and simply assumes that it is a time when international lenders suddenly impose binding collateral constraints. The collateral constraints specify that the stock of foreign debt cannot exceed the value of land and capital. They carry out their analysis in the context of a small 2-sector traded goods, non-traded goods model modified to incorporate features of the limited participation model. In particular, firms require two forms of working capital to conduct operations: domestic funds to finance the wage bill and foreign funds to pay for imports of an intermediate good. Just before the crisis the model economy is in a steady state with a zero current account balance, no collateral constraints and a large stock of foreign debt relative to the value of domestic assets. This steady state is hit by an unexpected imposition of collateral constraints. If monetary policy does not react, the current account responds by turning positive as the foreign debt is paid off, now with a high shadow cost because of the binding collateral constraint. At equilibrium the foreign debt is reduced to exactly the point where the collateral constraint becomes non-binding. During the transition, output and employment are low because the binding collateral constraint inhibits the import of the intermediate good.

Building on this baseline analysis, Christiano, Gust & Roldos [thereafter CGR] ask what is the economic effect of a monetary policy action which expands domestic liquidity and reduces the domestic nominal interest rate. They show that, under plausible circumstances, the policy causes an even greater fall in output and employment because the expansion of domestic liquidity operates on the wrong margin in an economy in the aftermath of a crisis. At a time like this, it is the shortage of foreign liquidity that holds back economic activity. Then, actions designed to expand domestic liquidity may paradoxically just serve to further tighten the constraint on foreign liquidity. The reasoning is simple. The expansion of domestic liquidity reduces the cost of labor. This has a relatively greater impact on the marginal cost of producing nontraded goods, because that sector is assumed to be relatively labor intensive. As a result, there is a fall in the price of nontraded versus traded goods. This in turn produces a real depreciation, and a fall in the international value of the economy’s collateral. The latter causes the collateral constraint to bind even more strongly, so that imports of the intermediate good must be curtailed. The result is an amplification in the drop in employment and output.

This analysis is a response to observers who argue that the appropriate way to cushion the fall in output in the aftermath of a currency crisis is for the central bank to cut the domestic rate of interest. Those arguments appeal to the conventional wisdom, based on experience in countries like the US, that interest rate cuts stimulate economic activity. This holds in the CGR model, but only in the version with no collateral constraints. The imposition of binding collateral constraints fundamentally changes the nature of the monetary transmission mechanism in the model, reversing the conventional wisdom. This suggests that experience with the effects of policy actions in economies in `normal times’, when collateral constraints are not binding, may have little relevance for predicting the effects of these actions in economies in `crisis times’, when collateral constraints are binding. In continuing work, CGR are investigating the exact nature of international lending contracts and the role of collateral constraints in international lending during normal and crisis times. They are also studying the empirical plausibility of their model’s assumptions required to guarantee the results on the effects of monetary policy with a binding collateral constraint. The key assumptions are limited substitutability between foreign intermediate goods and domestic factors in production and the relative importance of labor in the nontraded good sector.

References