The Asymmetric Effects of Monetary Policy on Job Creation and Destructior

Pietro Garibaldi

doi:10.5089/9781451930962.024.A007

Jump to Content Jump to Main Navigation

The Asymmetric Effects of Monetary Policy on Job Creation and Destructior

Author: Mr. Pietro Garibaldi

Publication Date:: 01 Jan 1997

eISBN:: 9781451930962

Language:: English

Keywords:: SP; trade balance; liquidity constraint; income coefficient; dollar wage; credit crunch; life-cycle income; interest spending; Consumption; Income; Job creation; Effective tax rate; Real exchange rates; Central and Eastern Europe; Baltics; Asia and Pacific; Southeast Asia; Sub-Saharan Africa; net employment change; job flow; firm behavior; reservation productivity; capital gain; Job destruction; Employment; Productivity; Unemployment; Europe

Download PDF (2 MB)

The view that tight and easy monetary policy produces asymmetric effects on economic activity has long been recognized in policy debates and in the academic profession (Johnson, 1962). The behavior of the U.S. economy during the 1990–92 recession, when successive cuts in the federal fund rate failed to produce economic recovery, seemed to confirm the traditional view. Furthermore, recently collected empirical evidence for both the United States (De Long and Summers, 1988; Cover 1992; and Morgan, 1993) and Europe (Karras, 1996) strongly supports the hypothesis that negative money-supply shocks and/or increases in interest rates reduce output more than monetary expansions raise it.

Abstract

Theoretically, the asymmetric effect of monetary policy has traditionally been rationalized in models that assume price rigidity or asymmetric information. Monetary policy has asymmetric effects on real output if prices are less flexible downward than upward. Ball and Mankiw (1994), Caballero and Engel (1992), and Caplin and Leahy (1991) propose Keynesian models in which tight policy causes output to falls with little change in prices, while easy policy causes prices to rise with little change in output.

Monetary policy may also result in asymmetric output responses if asymmetric information in the banking sector produces binding credit constraints. Jackman and Sutton (1982) show that tight policy makes banks less willing to lend to riskier borrowers when market rates are high. This behavior results in credit rationing and a fall in output in a way that does not have a counterpart during periods of easy policy.

The present paper focuses on the labor market and presents theory and evidence on the asymmetric effects of monetary policy on the process of job creation and job destruction. The traditional analysis focuses only on the net effect of monetary policy, and it fails to distinguish the different effects of monetary policy on the job creation and the job destruction margins. Theoretically, we develop a minimalist version of the most recent matching literature developed by Mortensen and Pissarides (1994). ^¹ With respect to the traditional matching approach (Diamond, 1982, and Pissarides, 1990), the new matching models assume heterogeneity in the value of the labor product. The idiosyncratic and job-specific risk for existing jobs is modeled as a jump process characterized by a Poisson arrival frequency and a common distribution of productivity. In equilibrium, existing entrepreneurs endogenously select a reservation productivity at which the continuation of a job is no longer profitable. We show that when interest rates are increased, existing entrepreneurs adjust their reservation productivity and immediately destroy any job that falls short of the newly selected value. Conversely, reductions in interest rates result in higher job creation through a costly and time-consuming process. The paper shows that the asymmetry between the hiring and firing technology implies that job creation and job destruction respond asymmetrically to changes in interest rate.

We analyze empirically the effect of changes in interest rates on the U.S. manufacturing data compiled by Davis and Haltiwanger (1990 and 1992). To identify a measure of policy stance we use an econometric procedure similar to the one applied in the existing literature on asymmetry (Cover, 1992, and Morgan, 1993). We find evidence that tight policy increases job destruction and reduces net employment changes. Conversely, easy policy appears ineffective in stimulating job creation. These results are broadly consistent with the empirical implications of the model and with the view that monetary policy produces asymmetric response on job creation and destruction.

I. Existing Empirical Evidence

From the mid-1970s to the mid-1980s, most of the work in empirical monetary economics (Barro, 1977 and 1978, and Mishkin, 1982) was concerned with the distinction between expected and unexpected changes in money supply, and related papers tried to determine to what extent output fluctuations were the result of unexpected changes in the money supply. With Cover’s (1992) paper, the attention has shifted to the distinction between positive and negative shocks. Cover presents evidence that negative money-supply shocks have a larger and more important effect on output than do positive shocks. Technically, he employs the following two-step econometric procedure. The first step implies specifying different money-supply and output processes. From the residual series of shocks from the money process, Cover constructs two distinct series of positive and negative shocks. Regressing output growth on its lagged values and on the constructed variables, one can obtain an estimate of the separate effects of positive and negative shocks. Cover shows that, no matter which specification of the money supply process is used, the only significant effect on output growth is the one originated from the negative shock variable. De Long and Summers (1988) find similar results using annual data and going back to the beginning of the century. In these papers, the existence of asymmetry is robust, even though quantitative measures of the effects are difficult to calculate.

To measure the robustness of the results, scholars have considered alternative measures of the policy stance. Morgan (1993) uses the federal funds rate to identify the policy stance and, using a two-step procedure similar to the one used by Cover, shows that changes in the federal funds rate appear to have an asymmetric impact on output growth, just as changes in money growth do. In Section V, when we look at the effects of interest rate changes on job flows, we use a procedure very similar to that of Morgan (1993).

Karras (1996) examines whether asymmetry, traditionally reported in U.S. data, is also a European phenomenon. Using annual data from 1953 to 1990, he identifies money-supply shocks for a panel of 18 European countries and shows that various specifications and estimation methods strongly support asymmetry. Similarly to the Morgan paper, Karras shows that asymmetry also characterizes interest rate effects: increases in interest rates result in statistically significant decreases in output, white reductions in the discount rate have insignificant output effects.

II. Concepts and Notations

We consider a continuous time economy populated by a fixed quantity of risk-neutral workers normalized to one for simplicity. Workers can be either employed or unemployed and looking for a job. Since our focus is on jobs, rather than on firms, we assume that each firm is made up of only one job. Jobs are forward-looking assets that can be in two different states: filled and producing or vacant. Filled jobs produce one unit of a homogeneous product with different productivity levels ∈, and each job is characterized by an irreversible technology. The job’s productivity is subject to idiosyncratic (firm-specific) shocks and randomly switches between different values. The idiosyncratic risk is modeled as a simple Poisson process with an arrival rate equal to λ. Conditional on being hit by an idiosyncratic shock, the new productivity level is drawn from a common productivity distribution F(∈), with finite support over the interval ∈_c. - ∈_u, and with no point mass, other than at the upper support ∈_u. Existing firms observe only the current productivity level ∈ and must decide whether to keep the job open or whether to immediately destroy it at no cost. Firm behavior is described by the choice of a reservation productivity, ∈_d, below which the continuation of production is no longer profitable, and job destruction takes place when a filled job draws a productivity below the reservation level.

Vacant jobs incur a search cost equal to γ and are slowly matched to unemployed job seekers; similarly to the definition used in the empirical literature, job creation takes place when a vacant job meets an unemployed job seeker. We assume that newly created jobs have the option to select the best productivity in the market and create a job at the upper support of the distribution ∈_u. However, after a job has been created, the technology becomes irreversible and firms cannot adjust their productivity levels. The meeting of unemployed job seekers and vacant firms is described by a homogeneous of degree one matching function m(u, ν), where u and ν are, respectively, the unemployment and the vacancy rate. In what follows, we shall indicate with

\begin{matrix} q (θ) = \frac{m (u, v)}{v}, & θ \equiv \frac{v}{u}, \end{matrix}

the probability of filling in an existing vacancy, where θ is a measure of market tightness from the firms’ standpoint. Similarly, the job finding rate for unemployed job seekers reads $θ_{q} (θ) = \frac{m (u, v)}{u}$ . Monetary policy’s stance is exogenous and is indicated by the interest rate r. In Section III, monetary policy is fixed and existing firms and unemployed job seekers take as given the interest rate r. Conversely, in Section IV, monetary policy can be tight or easy and there is a probability μ of switching between the two regimes.

The existence of the matching frictions and the search costs generate a pure economic rent to be shared between firms and workers already matched, and to close the model it is necessary to specify a wage rule. In most of the matching literature (Pissarides, 1990) the rent-sharing problem is solved with a Nash-bargaining game at the level of the job. Since the focus of this paper is on job creation and job destruction, rather than on wages, we assume that firms extract all the surplus from the match and fix the wage at the workers’ reservation utility level b. As Diamond (1971) has shown, this outcome is an equilibrium in a wage-setting game in which workers search sequentially at some positive cost and have only the power to accept or reject offers. Given this outcome, workers have no incentive to search on the job and their behavior is fully described by the reservation utility level b. Finally, with respect to the existing Mortensen and Pissarides (1994) model, we do not model common aggregate shocks and the technology is fully described by the idiosyncratic productivity ∈.

III. A Minimalist Model: Steady State

In steady state, the policy stance is fixed and each firm takes as given the exogenous interest rate r. In line with the traditional matching literature (Diamond, 1982, and Pissarides, 1990), we model jobs as forward-looking assets that can be in two different states: filled and producing or vacant. We present and solve the steady-state model in three steps. First, we solve for the job destruction decision of existing firms. Second, we focus on job creation and solve for the vacancy posting decision. Finally, we introduce aggregate unemployment and close the model.

Since the productivity of each job is stochastic and randomly jumps across the productivity distribution F, the value of an existing job depends on its current productivity ∈, and in what follows we shall indicate with 7(e) the value-function of a filled job with idiosyncratic productivity ∈. At each point in time the asset function J(∈) yields a (net) flow of production equal to the difference between the idiosyncratic productivity ∈ and the exogenous wage rate b. Furthermore, with finite probability λ, an idiosyncratic shock arrives, the productivity ∈ changes, and the firm draws a productivity level from the distribution F(.). Thus, conditional on a change in productivity, the firm incurs a capital gain (loss), equal to the positive (negative) difference between the expected value of a job and the value of the current job. Accordingly, the asset valuation of a job of productivity ∈ reads

\begin{matrix} r J (ϵ) = ϵ - b + λ {E [J | J (ϵ) > 0] - J (ϵ)}, & (1) \end{matrix}

where E[J|J(∈) > 0] is the expected value of a job, conditional on a positive value J(∈). Equation (1) shows that the return from a job is equal to the sum of a dividend, ∈ - b, and an expected capital gain (loss). Since a firm always has the option to shut down a job that commands a negative present value, the expectation term in equation (1) is conditional on those values of ∈ that yield a positive value function.J(.). Making use of the probability distribution F(.), the conditional expectation of equation (1) can be conveniently expressed to reflect the firm choice between the value of each job and the firm’s option to freely destroy a job. The firm’s decision problem in equation (1) reads

\begin{matrix} E [J | J (ϵ) > 0] = \int {\max [J (y), 0]} d F (y) & (2) \end{matrix}

Differentiating equation (1) with respect to ∈ yields

\begin{matrix} J^{'} (ϵ) = \frac{1}{r + λ}, & (3) \end{matrix}

and shows that J(∈) is a monotonic increasing function of ∈. Firm behavior can be fully described by the choice of a reservation productivity, say ∈,_d, so that for each productivity level lower than ∈_d, a job is immediately destroyed. Formally, the reservation productivity or cut-off value ∈_d is defined as

J (ϵ_{d}) \equiv 0.

After a simple integration by parts,^² and making use of the definition of ∈_d, the expected value in equation (2) reads

\begin{matrix} \int {\max [J (y), 0]} d F (y) = \int_{ϵ_{d}}^{ϵ_{u}} J (y) d F (y) = \frac{1}{r + λ} \int_{ϵ_{d}}^{ϵ_{u}} [1 - F (y)] d y . & (4) \end{matrix}

Making use of equation (4) in equation (1), and evaluating equation (1) at J(∈_d) = 0, the reservation productivity (or cut-off value) ∈_d solves

\begin{matrix} ϵ_{d} - b = - \frac{λ}{λ + r} \int_{ϵ_{d}}^{ϵ_{u}} [1 - F (y)] d y . & (5) \end{matrix}

Equation (5) is the first key equation of the model and uniquely solves for the reservation productivity ∈_d, as a function of the parameters of the model and of the distribution F(∈). The left-hand side of equation (5) is the profit from the marginal job and is negative. This implies that a zero present value for the marginal job is consistent with a negative marginal profit. Since firms have the option to shut down the job at no cost, they hoard labor in the event that conditions improve. Several comparative static results are derived in Appendix I. First, when $\frac{\partial ϵ_{d}}{\partial r} \geq 0$ , higher interest rates increase the marginal productivity. Intuitively, higher discount rates reduce the present value of future returns and make the firm less willing to hold on to existing jobs. Second, when $\frac{\partial ϵ_{d}}{\partial λ} \leq 0$ , the higher arrival rate of the idiosyncratic shock reduces the reservation productivity. Since the persistence of the idiosyncratic shock is $\frac{1}{λ}$ , a higher λ reduces the persistence of each shock and makes the firm more willing to hold on to existing jobs.

Job creation is a costly and time-consuming process, and comes from the matching of vacant jobs and unemployed job seekers. Vacancies are forward-looking assets owned by the firm and their yield is the sum of a dividend and a capital gain. Since vacancy posting involves a flow cost of γ per period, the dividend is equal to -γ per period. Furthermore, with instant probability q(θ), a vacant job meets an unemployed job seeker and creates a job at the upper support of the distribution. If we let V be the present value of a vacant job, the asset valuation of a vacancy reads

\begin{matrix} r V = - γ + q (θ) [J (ϵ_{u}) - V], & (6) \end{matrix}

where J(∈_u) - V is the capital gain associated with the meeting between a vacant job and an unemployed job seeker. If there is free entry and full exhaustion of rents, as in Pissarides (1990), vacancies will be created as long as their present value is positive. Thus the free-entry condition implies that, in equilibrium, V = 0. Substituting in equation (6) for V= 0 yields

\begin{matrix} J (ϵ_{u}) = \frac{γ}{q (θ)} . & (7) \end{matrix}

Since 1/_q(θ) is the average duration of a vacancy, γ_q(θ) is the expected search cost, and equation (7) shows that the value of a job at the upper support of the distribution is equal to the expected searching cost. Making use of equation (5) and of equation (1) evaluated at the upper support ∈_u, J(∈_u) can be written as

J (ϵ_{u}) = \frac{ϵ_{u} - ϵ_{d}}{r + λ} .

and the job creation condition becomes

\begin{matrix} \frac{ϵ_{u} - ϵ_{d}}{r + λ} = \frac{γ}{q (θ)} . & (8) \end{matrix}

Equation (8) is the second key equation of the model and, given the reservation productivity ∈_d, uniquely solves for θ. Appendix I derives the comparative static results. In particular, higher interest rates reduce market tightness. Higher interest rates reduce the present value of a new job and make vacancy posting less attractive.

To close the model, we need to introduce unemployment. In steady state, unemployment is constant and job creation is equal to job destruction. Aggregate job creation is the fraction of unemployed job seekers that meet a vacant job and can be written as

J C \equiv θ q (θ) u .

Aggregate job destruction is the fraction of existing jobs hit by a shock below the reservation productivity ∈_d and reads

J D \equiv λ F (ϵ_{d}) (1 - u) .

If job creation must be equal to job destruction, equilibrium unemployment solves

\begin{matrix} u = \frac{λ F (ϵ_{d})}{λ F (ϵ_{d}) + θ q (θ)} . & (9) \end{matrix}

Equation (9) is the Beveridge curve and completes the description of the model. The model is recursive and is fully described by three equations. Equation (5) determines the reservation productivity ∈_d. Equation (8), given ∈_d, uniquely solves for θ. Finally, equation (9), given θ and ∈_d, determines equilibrium unemployment. From equation (9), an increase in interest rate r unambiguously increases unemployment. In steady state, changes in the interest rate reduce firms’ job creation and increase firms’ job destruction and permanently affect equilibrium unemployment. To look at the asymmetric effect of interest rate changes on job flows, we need to expand the model and explicitly allow for changes in the interest rate.

IV. The Model with Tight and Easy Policy

This section extends the model of the previous section and considers an economy in which the policy stance can be tight or easy and, from the agents’ standpoint, randomly switches between the two states. In what follows we assume that the interest rate can take a high value r* and a low value r. We assume that the stochastic process that drives the policy change is Poisson with arrival probability μ. We solve the extension of the model and highlight the empirical implications through a simple simulation.

Model

From the steady-state model solved in the previous section, it is clear that in each policy regime, firms’ behavior is characterized by a reservation productivity. In what follows, we shall indicate with ∈*_d and ∈_d the reservation productivity during tight and easy policy, respectively. Furthermore, since the interest rate during periods of tight policy is higher than the interest rate during periods of easy policy (r* > r), the comparative static results of Appendix I suggest that ∈*_d > ∈_d. To solve the model, however, we need to specify a policy state contingent value function. Thus, we need to explicitly model the behavior of a filled job in each policy stance, and in what follows we shall indicate with J*(∈) and J(∈) the forward-looking value of a job with productivity ∈ under tight and easy policy, respectively.

For values of the idiosyncratic productivity ∈>∈*_d, the value of a job during a period of tight policy, J*(∈), reads

\begin{array}{l} r * J * (ϵ) = ϵ - b + λ \int_{ϵ_{d}^{*}}^{ϵ_{u}} {\max [J * (y), 0] - J * (ϵ)} d F (y) \\ \begin{matrix} + μ [J (ϵ) - J * (ϵ)], & ϵ \geq ϵ_{d}^{*}, & (10) \end{matrix} \end{array}

where at rate λ an idiosyncratic shock arrives and at rate μ a policy switch takes place. With respect to the steady-state specification of equation (1), equation (10) features an extra capital gain term: at rate μ policy switches from tight to easy, the firm loses its current value J*(∈) and gets the value of ajob with the same productivity level during an easy regime, J(∈).

Similarly to equation (10), the value of a job during a period of easy policy, J(∈), for ∈ > ∈*_d reads

\begin{array}{l} r J (ϵ) = ϵ - b + λ \int_{ϵ_{d}}^{ϵ_{u}} {\max [J (y), 0] - J (ϵ)} d F (y) \\ \begin{matrix} + μ [J * (ϵ) - J (ϵ)], & ϵ \geq ϵ_{d}^{*} . & (11) \end{matrix} \end{array}

For idiosyncratic values below ∈*_d, a job is operative only during periods of easy policy and J(∈) reads

\begin{array}{l} r J (ϵ) = ϵ - b + λ \int_{ϵ_{d}}^{ϵ_{u}} {\max [J (y), 0] - J (ϵ)} d F (y) - μ J (ϵ), \\ \begin{matrix} ϵ_{d} \leq ϵ \leq ϵ_{d}^{*} . & (12) \end{matrix} \end{array}

The last term in equation (12) reflects the capital loss deriving from a change in the policy regime: a job with productivity level ∈ < ∈*_d is immediately destroyed if policy turns tight, and the firm loses its current value J(∈). Differentiating equations (10) and (11) with respect to ∈ yields

\begin{matrix} \frac{\partial J * (ϵ)}{\partial ϵ} = \frac{r + λ + 2 μ}{(r + λ + μ) (r * + λ + μ) - μ^{2}} & \forall ϵ \geq ϵ_{d}^{*} & (13) \end{matrix}

and

\begin{matrix} \frac{\partial J (ϵ)}{\partial ϵ} = \frac{r * + λ + 2 μ}{(r + λ + μ) (r * + λ + μ) - μ^{2}} & \forall ϵ \geq ϵ_{d}^{*} . & (14) \end{matrix}

Similarly, differentiating equation (12) with respect to ∈ yields

\begin{matrix} \frac{\partial J (ϵ)}{\partial ϵ} = \frac{1}{r + λ + μ} & \forall ϵ : ϵ_{d} \geq ϵ \geq ϵ_{d}^{*} . & (15) \end{matrix}

Making use of equations (13). (14), and (15), after integrating by parts, the reservation productivity during tight policy, ∈*_d, solves as follows:

\begin{array}{l} ϵ_{d}^{*} - b = - \frac{λ (1 + λ + 2 μ)}{(r + λ + μ) (r * + λ + 2 μ) - μ^{2}} \int_{ϵ_{d}^{*}}^{ϵ_{u}} [1 - F (y)] d y \\ \begin{matrix} - \frac{μ}{r + λ + μ} (ϵ_{d}^{*} - ϵ_{d}) . & (16) \end{matrix} \end{array}

Similarly, the reservation productivity during easy policy solves

\begin{matrix} ϵ_{d} - b = - \frac{λ}{1 + λ + μ} \int_{ϵ_{d}}^{ϵ_{d}^{*}} [1 - F (y)] d y \\ \begin{matrix} - \frac{λ (r * + λ + 2 μ)}{(r + λ + μ) (r * + λ + μ) - μ^{2}} \int_{ϵ_{d}^{*}}^{ϵ_{u}} [1 - F (y)] d y . & (17) \end{matrix} \end{matrix}

To obtain market tightness during tight and easy policy it is first necessary to solve for the value of a job at the upper support of the distribution. Subtracting from both equation (10) and equation (11) the reservation productivity (16), it is possible to write

\begin{matrix} (r * + λ + μ) J * (ϵ_{u}) = ϵ_{u} - ϵ_{d}^{*} + μ [J (ϵ_{u}) - J (ϵ_{d}^{*})], & (18) \end{matrix}

and

\begin{matrix} (r + λ + μ) [J (ϵ_{u}) - J (ϵ_{d}^{*})] = ϵ_{u} - ϵ_{d}^{*} + μ J * (ϵ_{u}) . & (19) \end{matrix}

Equations (18) and (19) form a system of two equations that can be solved for J*(∈_u) and J(∈_u) - J(∈*_d). Similarly, subtracting equation (17) from equation (12), it is possible to write

\begin{matrix} (r + λ + μ) J (ϵ_{u}) = σ (ϵ_{u} - ϵ_{d}) + μ J * (ϵ_{u}) . & (20) \end{matrix}

Equation (20), given J*(∈_u) from (18) and (19), uniquely solves for J(∈_u). Given J*(∈_u) and J(∈_u) from equations (18), (19), and (20), two expressions analogous to equation (6) yield θ and θ*.

To complete the model, we need two differential equations that describe the dynamics of unemployment during the different regimes. During periods of tight policy, the unemployment dynamics are given by the difference between job creation and job destruction and read

\begin{matrix} \dot{u} = λ F (ϵ_{d}^{*}) (1 - u) - θ* q (θ*) u, & (21) \end{matrix}

while during periods of easy policy the differential equation becomes

\begin{matrix} \dot{u} = λ F (ϵ_{d}) (1 - u) - θ q (θ) u, & (22) \end{matrix}

where the difference between equations (21) and (22) is the fact that during periods of tight policy job creation and destruction are driven by ∈*_d and θ*, while during periods of easy policy they are driven by ∈_d and θ.

Discussion and Simulation

In the model with tight and easy policy, job creation and job destruction change over time, and it is possible to analyze the response of job flows to changes in policy. Since ∈*_d > ∈_d, equations (21) and (22) show that, other things equal, there is more job destruction during periods of tight policy than during periods of easy policy. Similarly, since θ* < θ, there is more job creation during periods of easy policy.

Let us now consider what happens when there is a switch from tight to easy policy. The intuition goes as follows. On impact, θ jumps upward and ∈ falls downward. Those jobs whose productivity level is now higher than ∈_d and lower than ∈*_d will no longer be destroyed and job destruction will fall. The jump in θ causes a jump in the number of vacancies and an increase in job creation through the slow matching described by θ_q(θ). This process will continue as long as job creation is equal to job destruction or there will be another policy shock. Let us now consider what happens when the policy shifts from easy to tight. As θ* < θ, job creation falls in a symmetric way to the increase in job creation during a shift from tight to easy policy: the number of vacancies falls and so does job creation. The behavior of job destruction during a switch from easy to tight is the driving force behind the asymmetric effect of interest rate changes. First, since ∈*_d > ∈_d, job destruction is higher, and those jobs whose productivity is lower than ∈*_d will now be destroyed. But on impact, as the reservation productivity jumps to ∈*_d, all jobs whose idiosyncratic productivity lies between the two reservation values are immediately destroyed. This jump in job destruction does not find a counterpart in the behavior of job destruction during a switch from a tight to an easy regime, nor in the behavior of job creation during switches from tight to easy policy. The intuition would suggest that job flows respond to interest rate changes in an asymmetric way.

To simulate the model’s response to policy switches in a formal way we need to discretize time and to characterize the distribution of employment at each reservation productivity ∈. Appendix II shows how to obtain the distribution of employment and how to simulate the model when time is discrete and uncertainty is resolved at the beginning of each period. Table 1 gives the parameter values used for such simulations. The choice of the parameter is standard and is in line with the values used by Mortensen and Pissarides (1994) for simulating the U.S. job flows, but in this paper the simulations are meant to be suggestive rather than fully realistic.

^{Table 1.}

Baseline Parameter Values

Source: Author’scalculations.

^{Table 1.}

Baseline Parameter Values

Variables	Notation	Value
Matching elasticity	α	0.500
Friction parameter	k	1.000
Reservation utility	b	0.000
Tight interest rate	r*	0.050
Easy interest rate	r	0.020
Idiosyncratic shock rate	λ	0.100
Price distribution	F(.)	uniform
Upper support	∈_u	1.000
Lower support	∈_to	-1.000
Switching probability	μ	0.100

Source: Author’scalculations.

^{Table 1.}

Baseline Parameter Values

Variables	Notation	Value
Matching elasticity	α	0.500
Friction parameter	k	1.000
Reservation utility	b	0.000
Tight interest rate	r*	0.050
Easy interest rate	r	0.020
Idiosyncratic shock rate	λ	0.100
Price distribution	F(.)	uniform
Upper support	∈_u	1.000
Lower support	∈_to	-1.000
Switching probability	μ	0.100

Source: Author’scalculations.

Figure 1 plots the time profile of net employment changes (the difference between JC and JD in equations (22) and (21)) along a full policy cycle. The economy starts off in an easy regime, switches to tight, and finally goes back to the easy regime. In correspondence to the switch to tight policy (in period 3 in Figure 1), there is a clear spike in net employment change that does not find a counterpart in the behavior of net employment changes when the policy switches back to the easy regime (in period 13 in Figure 1). At the time of the policy switch, the model predicts an asymmetric effect of monetary policy on the behavior of net employment changes. Eventually, net employment changes go back to zero and unemployment is constant. Thus, in the very long run the asymmetry fades away.

Figure 2 reports, for the same time profile as Figure 1, gross job creation and gross job destruction. From Figure 2, it is clear that the asymmetric response of net employment changes to interest rates is driven by the behavior of job destruction. Furthermore, not only does the jump in job destruction not find similarities in the behavior of job creation during the switch from tight to easy policy, but the response of job creation is much smoother, due to the existence of the matching function, which acts as a stochastic filter on the increase in vacancies. The other empirical implication of the model is thus that the marginal effect of tight policy on job destruction should be larger than the marginal effect of easy policy on job creation.

To conclude, our theoretical model implies that, on impact, changes in monetary policy have asymmetric effects on job creation and job destruction. In particular, we pointed out two empirical implications. First, net employment changes should be more responsive to increases in interest rates than to proportional cuts. Second, the increase in job destruction in response to increases in interest rates should be higher than the increase in job creation in response to cuts in interest rates. In the next section we turn to the empirical investigation and we test the implications of the model.

V. Empirical Evidence

Data

We use two sets of data. Job flows data are the statistics compiled by Davis and Haltiwanger (1990 and 1992) for the U.S. manufacturing service between 1972:2 and 1988:4. Gross job creation (destruction) is measured as the sum of all employment gains (losses) at the level of the establishment in a given period in the U.S. manufacturing industry. If we divide job creation (destruction) by the number of jobs in the sample, we get the job creation (destruction) rate. The difference between the job creation and destruction rates is the traditional measure of net employment changes.

Figure 3 is from Davis, Haltiwanger, and Shuh (1996) and plots the quarterly time series of job creation, job destruction, and net employment change. The fact that job destruction experiences wider fluctuations than job creation is evident and has stimulated much interest and research (Mortensen and Pissarides, 1994; Caballero and Hammour, 1995; and Garibaldi, 1997). Policy data are constructed using the federal fund rate for the United States between 1971:2 and 1988:4, as reported in the IMF’s International Financial Statistics (line 60b). To carry out the empirical analysis, we need also the growth rate of GDP and the inflation rate, as measured by the consumer price index. In the subsection where we check the robustness of our findings, we also use the oil price in U.S. dollars, the growth rate of the money supply (line 34), and the government net borrowing (line 84).

The only problem with the comparability of the data concerns the different definition of quarter in the two data sets. In the flows data, the first quarter is defined as the change in employment between end-November of the previous year and end-February. Conversely, in International Financial Statistics the first quarter is defined as the traditional end-December to end-March period. In what follows we do not adjust the series but we keep in mind that the contemporaneous value of the two time series has a lag of approximately one month.

Evidence of Asymmetry

We employ the econometric procedure developed in the empirical literature on the asymmetric effects of monetary policy (Morgan, 1993, and Cover, 1992) and briefly discussed in Section I. First, we need to specify the interest rate process.

In what follows, we follow Morgan and regress the level of the federal fund rate on its own lagged values, on current and lagged values of output growth, and on current and lagged values of inflation. The federal fund process reads

\begin{matrix} F_{t} = a + \sum_{i = 1}^{i = M} α_{i} F_{t - i} + \sum_{i = 0}^{i = M} β_{i} G_{t - i} + \sum_{i = 0}^{i = P} γ_{i} π_{t - 1} + e_{t}, & (23) \end{matrix}

where F_t is the federal fund rate, G_t is the real growth of output, π_t is the inflation rate, and e_t is a white noise. Following Cover (1992), we define a negative shock in monetary policy as

\begin{matrix} t i g h t = \max (e_{t}, 0), & (24) \end{matrix}

and a positive shock as

\begin{matrix} e a s y = \min (e_{t}, 0) . & (25) \end{matrix}

To analyze the effect of positive and negative policy shocks on job flows, we specify the following process for job flows data. Net employment changes are regressed on a constant term, on their own lagged values, and on lagged values of the series defined in equations (24) and (25). Formally, the job flow process reads

\begin{matrix} N E T_{t} = β + \sum_{i = 1}^{i = Q} δ_{i} N E T_{t - 1} + \sum_{i = 1}^{i = R} ρ_{i} t i g h t_{t - 1} + \sum_{i = 1}^{i = S} ω_{i} e a s y_{t - 1} + z_{t} . & (26) \end{matrix}

As the quarterly definition of NET_t is different from the quarterly definition in F_t, equation (26) does not include the contemporaneous value of shocks tight_t, and easy_t. A regression similar to that in equation (26) can be estimated for job creation (JC_t) and for job destruction (JD_t).

We estimate the system described by equation (23) and equation (26) using two different econometric methods. The first method is the two-step procedure used by Barro (1977 and 1978) in his study of the effects of unanticipated and anticipated money shocks. In the first step, equation (23) is estimated by ordinary least squares (OLS), and the residuals are used as regressors in equation (26). Several authors, and Mishkin (1982) in particular, have pointed out the advantages of a simultaneous estimation of the system in equations (23) and (26). Thus, the second method requires the use of multivariate maximum likelihood for a simultaneous estimate of the system specified in equations (23) and (26). Nevertheless, from the empirical analysis that follows, it is obvious that the overall results do not depend on the particular method applied.

The first issue we address concerns the specification of the number of lags in equation (23). As a starting point, we used the Akaike information criterion and, from equation (23), we select M = N = 1 and P = 2. The results are reported in Table 2.

Table 3 reports the results of the second stage of the procedure, that is, the regression specified in equation (26), using the residual obtained from the regression of Table 2. Table 3 uses job flows data that include employment changes originated in existing establishments as well as in newly created (destroyed) establishments.

The second column of Table 3, labeled NET, reports the regression of net employment change (NET) on its own lag value, and on the policy shocks variables and their lags. As predicted, the coefficients on tight have a negative sign and the coefficients on easy have, overall, a positive sign. Both shocks have a significant effect on net employment changes but the tight shock has a bigger (negative) impact than the easy shock. Quantitatively, the sum of the coefficients on tight are several times bigger than the sum of the coefficients on easy and the hypothesis that the sum of the two coefficients is equal is confidently rejected. In particular, the impact of tight policy on net employment changes takes place predominantly in the first lag, while the coefficients on easy policy are significant at both the first and the second lag, and the second lag has a negative sign.

^{Table 2.}

Federal Fund Rate Process 1971:1 to 1988:4 Ordinary Least Squares

Source: Author’s calculations.Notes: Standard errors are in parentheses. One, two, and three asterisks indicate significance at the 10, 5, and i percent level, respectively.

^{^a}GDP growth rate.

^{^b}Inflation rate.

^{Table 2.}

Federal Fund Rate Process 1971:1 to 1988:4 Ordinary Least Squares

F = Federal fund rate
Variables	Estimate
Constant	-0.49
	(0.54)
F_t-1	0.92***
	(0.05)
$G_{t}^{a}$	0.58***
	(0.17)
$π_{t}^{b}$	0.49***
	(0.17)
π_t-1	-0.37**
	(0.17)
Number of observations: 71
F(4,66) = 110.42
R² = 0.87
Ar(1) coefficients of the residuals = -0.16 (0.12)
Akaike criterion = 0.30

Source: Author’s calculations.Notes: Standard errors are in parentheses. One, two, and three asterisks indicate significance at the 10, 5, and i percent level, respectively.

^{^a}GDP growth rate.

^{^b}Inflation rate.

^{Table 2.}

Federal Fund Rate Process 1971:1 to 1988:4 Ordinary Least Squares

F = Federal fund rate
Variables	Estimate
Constant	-0.49
	(0.54)
F_t-1	0.92***
	(0.05)
$G_{t}^{a}$	0.58***
	(0.17)
$π_{t}^{b}$	0.49***
	(0.17)
π_t-1	-0.37**
	(0.17)
Number of observations: 71
F(4,66) = 110.42
R² = 0.87
Ar(1) coefficients of the residuals = -0.16 (0.12)
Akaike criterion = 0.30

Source: Author’s calculations.Notes: Standard errors are in parentheses. One, two, and three asterisks indicate significance at the 10, 5, and i percent level, respectively.

^{^a}GDP growth rate.

^{^b}Inflation rate.

^{Table 3.}

Job Flows Estimate 1972:2–1988:4: All Firms

(Ordinary least squares, second stage regression)

Source; Author’s calculations.Notes: Standard errors are in parentheses. NET, JD, and JC are net employment, job destruction, and job creation. One, two, and three asterisks indicate significance at the 10, 5, and 1 percent level, respectively.

^{^a}Sum of the tight coefficients.

^{^b}Sum of the easy coefficients.

^{^c}F-statistic of the hypothesis that the coefficients on tight are jointly zero.

^{^d}F-statistic of the hypothesis that the coefficients on easy are jointly zero.

^{^e}F-statistic of the hypothesis that the sum of the coefficients on tight equal the sum of the coefficients on easy.

^{Table 3.}

Job Flows Estimate 1972:2–1988:4: All Firms

(Ordinary least squares, second stage regression)

	NET	JD	JC
Constant	0.37	2,06***	1.66***
	(0.32)	(0.77)	(0.67)
tight_t-1	-1.01***	0.73***	-0.20**
	(0.25)	(0.21)	(0.11)
tight_t-2	-0.12	0.17	-0.06
	(0.27)	(0.21)	(0.11)
easy_t-1	0.99***	-1.03***	0.08
	(0.32)	(0.32)	(0.14)
easy_t-2	-0.84***	0.39*	-0.38***
	(0.35)	(0.34)	(0.14)
NET_t-1	0.72***	—	—
	(0.12)
NET_t-2	-0.46***	—	—
	(0.13)
NET_t-3	0.34***	—	—
	(0.11)
JD_t-1	—	0.64***	—
		(0.12)
JD_t-2	—	-0.42***	—
		(0.13)
JD_t-3	—	0.29***	—
		(0.11)
JC_t-1	—	—	0.55***
			(0.12)
JC_t-2	—	—	-0.08
			(0.13)
JC_t-3	—	—	0.18*
			(0.11)
SUM(tight)^{^a}	-1.13	0.91	-0.27
SUM(easy)^{^b}	0.15	-0.63	-0.37
tight = 0^{^c}	8.48***	9.14***	1.96
easy = 0^{^d}	6.96***	8.21***	3.62*
easy = tight^{^e}	4.06**	8.40***	0.15

^{^a}Sum of the tight coefficients.

^{^b}Sum of the easy coefficients.

^{^c}F-statistic of the hypothesis that the coefficients on tight are jointly zero.

^{^d}F-statistic of the hypothesis that the coefficients on easy are jointly zero.

^{^e}F-statistic of the hypothesis that the sum of the coefficients on tight equal the sum of the coefficients on easy.

^{Table 3.}

Job Flows Estimate 1972:2–1988:4: All Firms

(Ordinary least squares, second stage regression)

	NET	JD	JC
Constant	0.37	2,06***	1.66***
	(0.32)	(0.77)	(0.67)
tight_t-1	-1.01***	0.73***	-0.20**
	(0.25)	(0.21)	(0.11)
tight_t-2	-0.12	0.17	-0.06
	(0.27)	(0.21)	(0.11)
easy_t-1	0.99***	-1.03***	0.08
	(0.32)	(0.32)	(0.14)
easy_t-2	-0.84***	0.39*	-0.38***
	(0.35)	(0.34)	(0.14)
NET_t-1	0.72***	—	—
	(0.12)
NET_t-2	-0.46***	—	—
	(0.13)
NET_t-3	0.34***	—	—
	(0.11)
JD_t-1	—	0.64***	—
		(0.12)
JD_t-2	—	-0.42***	—
		(0.13)
JD_t-3	—	0.29***	—
		(0.11)
JC_t-1	—	—	0.55***
			(0.12)
JC_t-2	—	—	-0.08
			(0.13)
JC_t-3	—	—	0.18*
			(0.11)
SUM(tight)^{^a}	-1.13	0.91	-0.27
SUM(easy)^{^b}	0.15	-0.63	-0.37
tight = 0^{^c}	8.48***	9.14***	1.96
easy = 0^{^d}	6.96***	8.21***	3.62*
easy = tight^{^e}	4.06**	8.40***	0.15

^{^a}Sum of the tight coefficients.

^{^b}Sum of the easy coefficients.

^{^c}F-statistic of the hypothesis that the coefficients on tight are jointly zero.

^{^d}F-statistic of the hypothesis that the coefficients on easy are jointly zero.

^{^e}F-statistic of the hypothesis that the sum of the coefficients on tight equal the sum of the coefficients on easy.

Looking at the regression on job destruction, in the third column of Table 3, there is again some evidence of asymmetry, even though the magnitude of the effect is somewhat smaller. Tight policy increases job destruction more than easy policy reduces it, similarly to the theory’s prediction. Finally, looking at the effect of the tight and easy variables on job creation, Table 3 finds that easy monetary policy reduces job creation. This finding is certainly counterintuitive, but in Table 3 it is impossible to reject the hypothesis that the overall effects of tight policy and easy policy on job creation are different. Basically, from the last column of Table 3, it is impossible to conclude that there is clear evidence of the effect of monetary policy on job creation. Table 4 reports results for the same regressions as Table 3, but uses job flows data that exclude the employment changes originated in newly created and destroyed establishments. Overall, the results in Table 4 confirm the results of Table 3, and find clear evidence of asymmetry in the regressions that use net employment change and job destruction.

Table 5 reports simultaneous estimates of the system specified in equations (23) and (26) using a multivariate maximum likelihood. The overall evidence on asymmetry is still significant, even though the magnitude of the effect is smaller. The standard errors of the policy coefficients on net employment change are very similar to those of Table 3, but in Table 5 the effect of tight policy is only twice as great as the effect of easy policy. Similar considerations hold for the coefficients on job destruction. In the case of job creation, maximum likelihood estimates show that the negative effect of easy policy on job creation is not statistically significant. From the results of Table 5 it is clear that the evidence of asymmetry does not depend on the particular econometric method applied.

Robustness Checks

This subsection investigates the robustness of the evidence of asymmetry to a number of specifications and generalizations. First, we check whether the results of the preceding subsection depend on the particular number of lags of the policy shocks. Second, we change the specification of the federal fund process and include a higher number of lags. Third, we include in equations (23) and (26) the effects of other variables, such as the dollar price of oil, government net borrowing, and growth of the money supply.

The model developed in the first part of the paper suggests that the transmission of easy policy on job creation takes more time to produce its effect than does the transmission of tight policy on job destruction. Thus, the model predicts that in the very long run the asymmetric effect of tight and easy on job creation would disappear. In this respect, it seems important to check whether the existence of asymmetry depends on the number of lags included in equation (26). Table 6 performs simultaneous maximum likelihood estimates of the system specified in equations (23) and (26). and includes three lags of the policy shocks. With respect to net employment changes, the coefficients on the third lags are not significant and a test on the absence of asymmetry is confidently rejected. In the estimates of job destruction, the overall negative effect of easy policy on job destruction is greater than the positive effect of tight policy, but the coefficient on the third lag of the policy shock is not significant. In the case of job creation, the presence of further lags in equation (26) does not change the poor performance of the overall regression. Regressions with a higher number of lags (not reported here) do not change the results of Table 6. These results go beyond the predictions of the model and suggest that the asymmetric effect of monetary policy on job flows is a permanent phenomenon.

^{Table 4.}

Job Flows Estimate 1972:2–1988:4— Continuing Firms

(Ordinary least squares, second-stage regression)

^{^a}Sum of the tight coefficients.

^{^b}Sum of the easy coefficients.

^{^c}F-statistic of the hypothesis that the coefficients on tight are jointly zero.

^{^d}F-statistic of the hypothesis that the coefficients on easy are jointly zero.

^{^e}F-statistic of the hypothesis that the sum of the coefficients on tight equals the sum of the coefficients on easy.

^{Table 4.}

Job Flows Estimate 1972:2–1988:4— Continuing Firms

(Ordinary least squares, second-stage regression)

	NET	JD	JC
Constant	0.42*	2.27***	2.12***
	(0.30)	(0.78)	(0.67)
tight_t-1	-0.99***	0.73***	-0.16
	(0.22)	(0.20)	(0.18)
tight_t-2	0.02	0.08	-0.09
	(0.23)	(0.20)	(0.10)
easy_t-1	0.92***	-1.07***	-0.04
	(0.28)	(0.23)	(0.13)
easy_t-2	-0.91**	0.41	-0.37**
	(0.31)	(0.29)	(0.13)
NET_t-1	0.72***	—	—
	(0.12)
NET_t-2	-0.51***	—	—
	(0.12)
NET_t-3	0.39***	—	—
	(0.10)
JD_t-1	—	0.54***	—
		(0.13)
JD_t-2	—	-0.40***	—
		(0.12)
JD_t-3	—	0.26**	—
		(0.11)
JC_t-1	—	—	0.45***
			(0.12)
JC_t-2	—	—	0.10
			(0.11)
JC_t-3	—	—	0.1
			(0.11)
SUM(tight)^{^a}	-0.96	0.82	-0.25
SUM(easy)^{^b}	0.01	-0.65	-0.41
tight = 0^{^c}	9.06***	11.37***	1.82
easy = 0^{^d}	10.15***	8.54***	4.2***
easy = tight^{^e}	8.23***	7.74***	0.44

^{^a}Sum of the tight coefficients.

^{^b}Sum of the easy coefficients.

^{^c}F-statistic of the hypothesis that the coefficients on tight are jointly zero.

^{^d}F-statistic of the hypothesis that the coefficients on easy are jointly zero.

^{^e}F-statistic of the hypothesis that the sum of the coefficients on tight equals the sum of the coefficients on easy.

^{Table 4.}

Job Flows Estimate 1972:2–1988:4— Continuing Firms

(Ordinary least squares, second-stage regression)

	NET	JD	JC
Constant	0.42*	2.27***	2.12***
	(0.30)	(0.78)	(0.67)
tight_t-1	-0.99***	0.73***	-0.16
	(0.22)	(0.20)	(0.18)
tight_t-2	0.02	0.08	-0.09
	(0.23)	(0.20)	(0.10)
easy_t-1	0.92***	-1.07***	-0.04
	(0.28)	(0.23)	(0.13)
easy_t-2	-0.91**	0.41	-0.37**
	(0.31)	(0.29)	(0.13)
NET_t-1	0.72***	—	—
	(0.12)
NET_t-2	-0.51***	—	—
	(0.12)
NET_t-3	0.39***	—	—
	(0.10)
JD_t-1	—	0.54***	—
		(0.13)
JD_t-2	—	-0.40***	—
		(0.12)
JD_t-3	—	0.26**	—
		(0.11)
JC_t-1	—	—	0.45***
			(0.12)
JC_t-2	—	—	0.10
			(0.11)
JC_t-3	—	—	0.1
			(0.11)
SUM(tight)^{^a}	-0.96	0.82	-0.25
SUM(easy)^{^b}	0.01	-0.65	-0.41
tight = 0^{^c}	9.06***	11.37***	1.82
easy = 0^{^d}	10.15***	8.54***	4.2***
easy = tight^{^e}	8.23***	7.74***	0.44

^{^a}Sum of the tight coefficients.

^{^b}Sum of the easy coefficients.

^{^c}F-statistic of the hypothesis that the coefficients on tight are jointly zero.

^{^d}F-statistic of the hypothesis that the coefficients on easy are jointly zero.

^{^e}F-statistic of the hypothesis that the sum of the coefficients on tight equals the sum of the coefficients on easy.

^{Table 5.}

Job Flows Estimate 972:2–1988:4—Joint Estimates

(All firms, maximum likelihood)^{^a}

Source: Author’s calculations.Notes: NET, JD, and JC are net employment, job destruction, and job creation respectively. Standard errors are in parentheses. One, two, and three asterisks indicate significance at the 10, 5, and 1 percent level, respectively.

^{^a}Regressions include a constant and three lags of the dependent variable.

^{^b}Sum of the tight coefficients.

^{^c}Sum of the easy coefficients.

^{^d}F-statistic of the hypothesis that the coefficients on tight are jointly zero.

^{^e}F-statistic of the hypothesis that the coefficients on easy are jointly zero.

^{^F}F-statistic of the hypothesis that the sum of the coefficients on tight equals the sum of the coefficients on easy.

^{Table 5.}

Job Flows Estimate 972:2–1988:4—Joint Estimates

(All firms, maximum likelihood)^{^a}

	NET	JD	JC
tight_t-1	-1.02***	0.75***	-0.20
	(0.22)	(0.21)	(0.16)
tight_t-2	0.04	0.10	-0.06
	(0.23)	(0.21)	(0.17)
easy_t-1	1.48***	-1.21***	0.08
	(036)	(0.32)	(0.2)
easy_t-2	—0.97***	0.41*	-0.39**
	(0.38)	(0.34)	(0.42)
F_t-1	0.99***	0.98***	0.91***
	(0.04)	(0.04)	(0.04)
G_t	0.49***	0.55***	0.56***
	(0.11)	(0.12)	(0.15)
π_t	0.61***	0.52***	0.47***
	(0.11)	(0.11)	(0.13)
π_t-1	-0.55***	-0.43***	-0.34***
	(0.11)	(0.11)	(0.13)
SUM(tight)^{^b}	-0.98	0.85	-0.27
SUM(easy)^{^c}	0.50	-0.79	-0.38
tight = 0^{^d}	22.5***	15.47***	1.89
easy = 0^{^e}	21.38***	21.54***	3.80
easy = tight^{^f}	5.51*	8.65**	0.09

^{^a}Regressions include a constant and three lags of the dependent variable.

^{^b}Sum of the tight coefficients.

^{^c}Sum of the easy coefficients.

^{^d}F-statistic of the hypothesis that the coefficients on tight are jointly zero.

^{^e}F-statistic of the hypothesis that the coefficients on easy are jointly zero.

^{^F}F-statistic of the hypothesis that the sum of the coefficients on tight equals the sum of the coefficients on easy.

^{Table 5.}

Job Flows Estimate 972:2–1988:4—Joint Estimates

(All firms, maximum likelihood)^{^a}

	NET	JD	JC
tight_t-1	-1.02***	0.75***	-0.20
	(0.22)	(0.21)	(0.16)
tight_t-2	0.04	0.10	-0.06
	(0.23)	(0.21)	(0.17)
easy_t-1	1.48***	-1.21***	0.08
	(036)	(0.32)	(0.2)
easy_t-2	—0.97***	0.41*	-0.39**
	(0.38)	(0.34)	(0.42)
F_t-1	0.99***	0.98***	0.91***
	(0.04)	(0.04)	(0.04)
G_t	0.49***	0.55***	0.56***
	(0.11)	(0.12)	(0.15)
π_t	0.61***	0.52***	0.47***
	(0.11)	(0.11)	(0.13)
π_t-1	-0.55***	-0.43***	-0.34***
	(0.11)	(0.11)	(0.13)
SUM(tight)^{^b}	-0.98	0.85	-0.27
SUM(easy)^{^c}	0.50	-0.79	-0.38
tight = 0^{^d}	22.5***	15.47***	1.89
easy = 0^{^e}	21.38***	21.54***	3.80
easy = tight^{^f}	5.51*	8.65**	0.09

^{^a}Regressions include a constant and three lags of the dependent variable.

^{^b}Sum of the tight coefficients.

^{^c}Sum of the easy coefficients.

^{^d}F-statistic of the hypothesis that the coefficients on tight are jointly zero.

^{^e}F-statistic of the hypothesis that the coefficients on easy are jointly zero.

^{^F}F-statistic of the hypothesis that the sum of the coefficients on tight equals the sum of the coefficients on easy.

^{Table 6.}

Job Flows Estimate 1972:2–1988:4—Estimates with Three Lags

(All firms, maximum likelihood,^{^a} three lags in tight and easy)

Source: Author’s calculations.Notes: NET, JD, and JC are net employment, job destruction, and job creation, respectively. Standard errors are in parentheses. One, two, and three asterisks indicate significance at the 10, 5, and 1 percent level, respectively.

^{^a}Regressions include a constant and three lags of the dependent variable.

^{^b}Sum of the tight coefficients.

^{^c}Sum of the easy coefficients.

^{^d}F-statistic of the hypothesis that the coefficients on tight are jointly zero.

^{^e}F-statistic of the hypothesis that the coefficients on easy are jointly zero.

^{^f}F-statistic of the hypothesis that the sum of the coefficients on tight equals the sum of the coefficients on easy.

^{Table 6.}

Job Flows Estimate 1972:2–1988:4—Estimates with Three Lags

(All firms, maximum likelihood,^{^a} three lags in tight and easy)

	NET	JD	JC
tight_t-1	-1.03***	0.73***	-0.17
	(0.04)	(0.19)	(-1.04)
tight_t-2	-0.23	0.06	-0.16
	(0.27)	(0.29)	(0.17)
tight_t-3	0.02	-0.07	-0.03
	(0.07)	(0.22)	(0.14)
easy_t-1	1.12***	-1.34***	-0.07
	(0.36)	(0.22)	(0.32)
easy_t-2	-0.83**	0.45*	-0.31
	(0.34)	(0.31)	(0.23)
easy_t-3	-0.16	-0.26*	-0.31
	(0.4)	(0.35)	(0.23)
F_t-1	0.96***	0.99***	0.93***
	(0.04)	(0.04)	(0.04)
G_t	0.57***	0.50***	0.57***
	(0.13)	(0.11)	(0.13)
π_t	0.43***	0.53***	0.47***
	(0.09)	(0.11)	(0.13)
π_t-1	-0.30**	-0.43***	-0.27**
	(0.15)	(0.12)	(0.13)
SUM(tight)^{^b}	-1.25	0.72	-0.38
SUM(easy)^{^c}	0.12	-0.15	-0.70
tight = 0^{^d}	17.5***	13.53***	2.51
easy = 0^{^e}	24.38***	18.32***	2.86
easy = tight^{^f}	2.76	7.67**	0.11

^{^a}Regressions include a constant and three lags of the dependent variable.

^{^b}Sum of the tight coefficients.

^{^c}Sum of the easy coefficients.

^{^d}F-statistic of the hypothesis that the coefficients on tight are jointly zero.

^{^e}F-statistic of the hypothesis that the coefficients on easy are jointly zero.

^{^f}F-statistic of the hypothesis that the sum of the coefficients on tight equals the sum of the coefficients on easy.

^{Table 6.}

Job Flows Estimate 1972:2–1988:4—Estimates with Three Lags

(All firms, maximum likelihood,^{^a} three lags in tight and easy)

	NET	JD	JC
tight_t-1	-1.03***	0.73***	-0.17
	(0.04)	(0.19)	(-1.04)
tight_t-2	-0.23	0.06	-0.16
	(0.27)	(0.29)	(0.17)
tight_t-3	0.02	-0.07	-0.03
	(0.07)	(0.22)	(0.14)
easy_t-1	1.12***	-1.34***	-0.07
	(0.36)	(0.22)	(0.32)
easy_t-2	-0.83**	0.45*	-0.31
	(0.34)	(0.31)	(0.23)
easy_t-3	-0.16	-0.26*	-0.31
	(0.4)	(0.35)	(0.23)
F_t-1	0.96***	0.99***	0.93***
	(0.04)	(0.04)	(0.04)
G_t	0.57***	0.50***	0.57***
	(0.13)	(0.11)	(0.13)
π_t	0.43***	0.53***	0.47***
	(0.09)	(0.11)	(0.13)
π_t-1	-0.30**	-0.43***	-0.27**
	(0.15)	(0.12)	(0.13)
SUM(tight)^{^b}	-1.25	0.72	-0.38
SUM(easy)^{^c}	0.12	-0.15	-0.70
tight = 0^{^d}	17.5***	13.53***	2.51
easy = 0^{^e}	24.38***	18.32***	2.86
easy = tight^{^f}	2.76	7.67**	0.11

^{^a}Regressions include a constant and three lags of the dependent variable.

^{^b}Sum of the tight coefficients.

^{^c}Sum of the easy coefficients.

^{^d}F-statistic of the hypothesis that the coefficients on tight are jointly zero.

^{^e}F-statistic of the hypothesis that the coefficients on easy are jointly zero.

^{^f}F-statistic of the hypothesis that the sum of the coefficients on tight equals the sum of the coefficients on easy.

Table 7 estimates the system of equations (23) and (26) using two lags of the federal fund rate, two lags of GDP growth, and two lags of the inflation rate. However, as the number of lags in equation (23) increases, the maximum likelihood estimator has problem of convergence. As an alternative, Table 7 employs OLS estimates.^³ Even though the F-statistics in Table 7 are somewhat smaller than the statistics in Table 3, the overall evidence on asymmetry is still remarkable.

^{Table 7.}

Job Flows Estimate 1972:2–1988:4—Two Lags of Federal Rate

(All firms, two-stage ordinary least squares,^{^a} two lags in F, G, and π)

^{^a}Regressions include a constant and two lags of the dependent variable.

^{^b}Sum of the tight coefficients.

^{^c}Sum of the easy coefficients.

^{^d}F-statistic of the hypothesis that the coefficients on tight are jointly zero.

^{^e}F-statistic of the hypothesis that the coefficients on easy are jointly zero.

^{^f}F-statistic of the hypothesis that the sum of the coefficients on tight equals the sum of the coefficients on easy.

^{Table 7.}

Job Flows Estimate 1972:2–1988:4—Two Lags of Federal Rate

(All firms, two-stage ordinary least squares,^{^a} two lags in F, G, and π)

First-stage regression
F_t-1	0.85***
	(0.12)
F_t-2	0.08
	(0.13)
G_t	0.45**
	(0.19)
G_t-1	0.34*
	(0.18)
G_t-2	0.07
	(0.16)
π_t	0.75***
	0.23)
π_t-1	-0.98***
	(0.40)
π_t-2	0.41*
	(0.23)

Second-stage regression
	NET	JD	JC
tight_t-1	-0.88***	0.70***	-0.13
	(0.27)	(0.21)	(-0.12)
tight_t-2	-0.45	0.38	-0.14*
	(0.29)	(0.22)	(0.12)
easy_t-1	0.49***	-1.05***	-0.02
	(0.36)	(0.28)	(0.16)
easy_t-2	-0.56**	0.21	-0.34
	(0.38)	(0.31)	(0.16)
SUM(tight)^{^b}	-1.33	1.08	-0.38
SUM(easy)^{^c}	0.37	-0.83	-0.70
tight = 0^{^d}	6.99**	6.92**	2.54
easy = 0^{^e}	3.71**	7.94***	1.47
easy = tight^{^f}	5.95**	11.35***	0.09

^{^a}Regressions include a constant and two lags of the dependent variable.

^{^b}Sum of the tight coefficients.

^{^c}Sum of the easy coefficients.

^{^d}F-statistic of the hypothesis that the coefficients on tight are jointly zero.

^{^e}F-statistic of the hypothesis that the coefficients on easy are jointly zero.

^{^f}F-statistic of the hypothesis that the sum of the coefficients on tight equals the sum of the coefficients on easy.

^{Table 7.}

Job Flows Estimate 1972:2–1988:4—Two Lags of Federal Rate

(All firms, two-stage ordinary least squares,^{^a} two lags in F, G, and π)

First-stage regression
F_t-1	0.85***
	(0.12)
F_t-2	0.08
	(0.13)
G_t	0.45**
	(0.19)
G_t-1	0.34*
	(0.18)
G_t-2	0.07
	(0.16)
π_t	0.75***
	0.23)
π_t-1	-0.98***
	(0.40)
π_t-2	0.41*
	(0.23)

Second-stage regression
	NET	JD	JC
tight_t-1	-0.88***	0.70***	-0.13
	(0.27)	(0.21)	(-0.12)
tight_t-2	-0.45	0.38	-0.14*
	(0.29)	(0.22)	(0.12)
easy_t-1	0.49***	-1.05***	-0.02
	(0.36)	(0.28)	(0.16)
easy_t-2	-0.56**	0.21	-0.34
	(0.38)	(0.31)	(0.16)
SUM(tight)^{^b}	-1.33	1.08	-0.38
SUM(easy)^{^c}	0.37	-0.83	-0.70
tight = 0^{^d}	6.99**	6.92**	2.54
easy = 0^{^e}	3.71**	7.94***	1.47
easy = tight^{^f}	5.95**	11.35***	0.09

^{^a}Regressions include a constant and two lags of the dependent variable.

^{^b}Sum of the tight coefficients.

^{^c}Sum of the easy coefficients.

^{^d}F-statistic of the hypothesis that the coefficients on tight are jointly zero.

^{^e}F-statistic of the hypothesis that the coefficients on easy are jointly zero.

^{^f}F-statistic of the hypothesis that the sum of the coefficients on tight equals the sum of the coefficients on easy.

Finally, we check whether the results depend on the particular variables used in the specification of the federal fund process. In what follows, to the variables already considered in equations (23) and (26), we add the dollar price of oil, a proxy for supply shocks, the growth of the money supply, and the government borrowing requirement in percentage of GDP. with the aim of capturing the fiscal policy stance. Table 8 shows that the evidence of the asymmetric effect of monetary policy on job flows does not depend on the particular specification of the system in equations (23) and (26).^⁴ The important result in Table 8 is the fact that the significance of policy shocks on job creation disappears when we include the new regressors. Conversely, tight and easy policy have a significant and asymmetric effect on job destruction and on net employment changes.

^{Table 8.}

Job Flows Estimate 1972:2–1988:4—Oil Price Included

(All firms, two-stage ordinary least squares, ^a including controls for oil prices, money growth, and government borrowing)

^{^a}Regressions include a constant and two lags of the dependent variable.

^{^b}Sum of the tight coefficients.

^{^c}Sum of the easy coefficients.

^{^d}F-statistic of the hypothesis that the coefficients on tight are jointly zero.

^{^e}F-statistic of the hypothesis that the coefficients on easy are jointly zero.

^{^f}F-statistic of the hypothesis that the sum of the coefficients on tight equals the sum of the coefficients on easy.

^{Table 8.}

Job Flows Estimate 1972:2–1988:4—Oil Price Included

(All firms, two-stage ordinary least squares, ^a including controls for oil prices, money growth, and government borrowing)

First-stage regression
F_t-1	0.83***	F_t-2	-0.08
	(0.12)		(0.14)
G_t	0.31*	G_t-1	0.21
	0.19		(0.18)
G_t-2	0.03	π_t	0.47**
	(0.15)		(0.24)
π_t-1	-0.81**	π_t-2	0.45***
	(0.39)		(0.22)
π_t-2	0.45***	bg_t	0.95***
	(0.22)		0.39
oil_t	0.05***	mπ_t	0.007
	(0.02)		(0.04)
m_t-1	0.02
	(0.04)

Second-stage regression
	NET	JD	JC
tight_t-1	-0.92**	0.72**	-0.17
	(0.31)	(0.24)	(0.21)
tight_t-2	-0.43	0.33	-0.16*
	(0.31)	(0.25)	(0.13)
easy_t-1	0,73 **	-0.78**	-0.07
	(0.44)	(0.35)	(0.19)
easy_t-2	-0.46	0.22	-0.30
	(0.44)	(0.35)	(0.19)
oilt_-1	0.015	-0.08	-0.01
	(0.07)	(0.05)	(0.03)
oilt_-1	-0.03	0.01	0.01
	(0.07)	(0.05)	(0.03)
SUM(tight)^{^b}	-1.35	1.06	-0.33
SUM(easy) ^{^c}	0.27	-0.56	-0.38
tight = 0 ^{^d}	5.44***	5.38***	1.47
easy = 0 ^{^e}	1.72	2.54	1.50
easy = tight^{^f}	3.68*	5.59**	0.019

^{^a}Regressions include a constant and two lags of the dependent variable.

^{^b}Sum of the tight coefficients.

^{^c}Sum of the easy coefficients.

^{^d}F-statistic of the hypothesis that the coefficients on tight are jointly zero.

^{^e}F-statistic of the hypothesis that the coefficients on easy are jointly zero.

^{^f}F-statistic of the hypothesis that the sum of the coefficients on tight equals the sum of the coefficients on easy.

^{Table 8.}

Job Flows Estimate 1972:2–1988:4—Oil Price Included

(All firms, two-stage ordinary least squares, ^a including controls for oil prices, money growth, and government borrowing)

First-stage regression
F_t-1	0.83***	F_t-2	-0.08
	(0.12)		(0.14)
G_t	0.31*	G_t-1	0.21
	0.19		(0.18)
G_t-2	0.03	π_t	0.47**
	(0.15)		(0.24)
π_t-1	-0.81**	π_t-2	0.45***
	(0.39)		(0.22)
π_t-2	0.45***	bg_t	0.95***
	(0.22)		0.39
oil_t	0.05***	mπ_t	0.007
	(0.02)		(0.04)
m_t-1	0.02
	(0.04)

Second-stage regression
	NET	JD	JC
tight_t-1	-0.92**	0.72**	-0.17
	(0.31)	(0.24)	(0.21)
tight_t-2	-0.43	0.33	-0.16*
	(0.31)	(0.25)	(0.13)
easy_t-1	0,73 **	-0.78**	-0.07
	(0.44)	(0.35)	(0.19)
easy_t-2	-0.46	0.22	-0.30
	(0.44)	(0.35)	(0.19)
oilt_-1	0.015	-0.08	-0.01
	(0.07)	(0.05)	(0.03)
oilt_-1	-0.03	0.01	0.01
	(0.07)	(0.05)	(0.03)
SUM(tight)^{^b}	-1.35	1.06	-0.33
SUM(easy) ^{^c}	0.27	-0.56	-0.38
tight = 0 ^{^d}	5.44***	5.38***	1.47
easy = 0 ^{^e}	1.72	2.54	1.50
easy = tight^{^f}	3.68*	5.59**	0.019

^{^a}Regressions include a constant and two lags of the dependent variable.

^{^b}Sum of the tight coefficients.

^{^c}Sum of the easy coefficients.

^{^d}F-statistic of the hypothesis that the coefficients on tight are jointly zero.

^{^e}F-statistic of the hypothesis that the coefficients on easy are jointly zero.

^{^f}F-statistic of the hypothesis that the sum of the coefficients on tight equals the sum of the coefficients on easy.

VI. Conclusions

This paper has presented theory and evidence on the asymmetric effects of monetary policy on job creation and destruction. Using the most recent matching literature developed by Mortensen and Pissarides (1994), the paper has shown that the theory predicts a clear asymmetry between the effects of interest rate changes on job creation and job destruction. In a model in which existing firms face idiosyncratic uncertainty and endogenously select the separation rate, a tightening of monetary policy, as described by an exogenous increase in interest rates, is immediately transmitted into higher job destruction. Conversely, easing monetary policy produces a slow effect on job creation and, in particular, does not produce the one-time jump in job creation that higher interest rates produce on job destruction. As a consequence, net employment change responds more to increases in interest rates than to reductions. However, as time goes on and the slow matching takes place, unemployment goes back to its steady-state level and the short-run asymmetry fades away.

Empirically, this paper has implemented a standard econometric technique for identifying the stance of monetary policy and has shown that the empirical implications of the model are broadly supported by the data. Increases in interest rates significantly affect job destruction, while reductions in interest rates fail to stimulate job creation. Using quarterly data for the U.S. manufacturing data, there appears to be a clear asymmetric effect of interest rate changes on the dynamics of job creation and destruction. Moreover, the asymmetric effects of interest rate changes seem to hold also in the long run, in addition to the short-run effect pointed out in the theoretical analysis. Future research should investigate the temporary and permanent effects of interest rate changes on job creation and destruction.

Throughout the paper, the role of monetary policy has been collapsed to an exogenous change in interest rate and the paper was silent on the choice of monetary policy at various stages of the cycle. Caplin and Leahy (1996) propose a policy game between two investors and a policymaker uncertain about the state of the world, and they show that gradual reductions in inter rates provide little stimulus to an economy in recession. Caplin and Leahy provide some insight into the apparent failure of monetary policy to end recessions, but their paper does not help explain why tight policy produces significant reduction in economic activity. Future research should try to endogenize monetary policy in a theoretical framework that implies the asymmetric effect of monetary policy on job flows.

APPENDIX I

Some Comparative Static Results

In the steady-state model, the reservation productivity reads

\begin{matrix} ϵ_{d} = b - \frac{λ}{σ + r} \int_{ϵ_{d}}^{ϵ_{u}} [1 - F (y)] d y . & (A1) \end{matrix}

The effect of the interest rate is $\frac{\partial \in_{d}}{\partial r} > 0$ . Differentiating equation (A1) with respect to r and rearranging yields

\begin{matrix} \frac{\partial ϵ_{d}}{\partial r} [r + λ F (ϵ_{d})] = \frac{λ \int_{ϵ_{d}}^{ϵ_{u}} [1 - F (y)] d y}{r + λ} . & (A2) \end{matrix}

Since the integral on the right-hand side of equation (A2) is positive, it follows that $\frac{\partial \in_{d}}{\partial r} > 0.$

The effect of λ is $\frac{\partial \in_{d}}{\partial r} > 0$ . Differentiating equation (A1) with respect to λ and rearranging yields

\begin{matrix} \frac{\partial ϵ_{d}}{\partial λ} [r + λ F (ϵ_{d})] = \frac{- (r + λ) \int_{ϵ_{d}}^{ϵ_{u}} [1 - F (y)] d y - \int_{ϵ_{d}}^{ϵ_{u}} [1 - F (y)] d y}{{(r + λ)}^{2}} . & (A3) \end{matrix}

Since the expression on the right-hand side of equation (A3) is negative, it follows that The $\frac{\partial \in_{d}}{\partial λ} < 0.$

job creation condition is

\begin{matrix} \frac{ϵ_{u} - ϵ_{d}}{r + λ} = \frac{γ}{q (θ)} . & (A4) \end{matrix}

The effect of r is $\frac{\partial θ}{\partial r} < 0$ . Differentiating equation (A4) with respect to θ yields

\begin{matrix} \frac{- \frac{\partial_{ϵ_{d}}}{\partial_{r}} (r + λ) - (ϵ_{u} - ϵ_{d})}{{(r + λ)}^{2}} = \frac{q^{'} (θ}{q (θ} & (A5) \end{matrix}

Since q2032;(θ) < 0, from equation (A5), making use of the result in equation (A2), it follows that $\frac{\partial θ}{\partial r} < 0.$

Equilibrium unemployment reads

\begin{matrix} u = \frac{λ F (ϵ_{d})}{λ F (ϵ_{d}) + θ q (θ)} . & (A6) \end{matrix}

The effect of the interest rate is $\frac{\partial u}{\partial r} > 0$ .

Differentiating equation (A6) with respect to r and rearranging yields

\begin{matrix} \frac{\partial u}{\partial r} = \frac{λ f (ϵ_{d}) \frac{\partial_{ϵ_{d}}}{\partial_{r}} (θ) - q (θ) [1 - η (θ)] λ F (ϵ_{d}) \frac{\partial θ}{\partial r}}{{[θ_{q} (θ) + λ F (ϵ_{d})]}^{2}}, & (A7) \end{matrix}

where η(θ) is the elasticity of the matching function with respect to θ. With a constant return technology η(θ) < 1 and, since $\frac{\partial θ}{\partial r} < 0$ , the second term in the numerator is positive overall. Furthermore, making use of the fact that $\frac{\partial \in_{d}}{\partial r} > 0$ , the overall expression is unambiguously positive.

APPENDIX II

Distribution of Employment

To actually simulate the model’s dynamics, it is first necessary to discretize time and make some assumptions on the behavior of ϵ_d and θ. The behavior of firms is fully characterized by the reservation productivity ϵ_d and the market tightness θ. When there is a change in policy the variables will immediately jump to the new value. Thus, in what follows, we discretize time and assume that uncertainty is completely resolved at the beginning of the period and constant throughout. Furthermore, we have to keep track of the distribution of employment over the job-specific productivity. Let us discretize the support of the distribution F over K at distinct points a₁ … a_k with a₁ <a_K, and let n_t(a_k − a_{k + 1}) represent the measure of workers at jobs with productivity within the interval a_k − a_{k + 1}. As all existing jobs flow into this set at the rate λ[F(a_{k + 1}) − F(a_k)] and flow out at a rate equal to the arrival rate of the productivity shock λ, if a_k > ϵ_d, and if a_k − a_{k + 1} is sufficiently small, the law of motion of the distribution of employment can be written as

\begin{array}{l} \begin{matrix} n_{t + 1} (a_{k + 1} - a_{k}) = (1 - λ) n_{t} + λ [F (a_{k + 1}) - F (a_{k})] [N_{t} - \int_{ϵ_{1}}^{ϵ_{d}^{i}} n_{t} (x) d x] & (A8) \end{matrix} \\ if ϵ > a_{k + 1} and a_{k} > ϵ_{d} \end{array}

\begin{matrix} n_{t + 1} = 0 if a_{k} < ϵ_{d} . & (A9) \end{matrix}

In equations (A8) and (A9), N_t represents total employment at the beginning of period t and $\in_{d}^{i}$ is the reservation productivity at timer t. Note that the most productive jobs, those for which ϵ = ϵ_u, are excluded. The latter jobs are obtained as the difference between N_t and the integral $\int_{\in_{I}}^{\in_{N}} n_{t} (x) d x$ .

Job creation is identical to the rate at which vacant jobs are matched with unemployed workers in the model. Given that every unemployed worker finds a vacancy with probability $θ_{t}^{i} q (θ_{t}^{i})$ , where θⁱ is the market tightness at time t, total job creation at time is

\begin{matrix} C_{t} = θ_{t}^{i} q (θ_{t}^{i}) (1 - N_{t}) . & (A10) \end{matrix}

If the matching function assumed is log-linear with search input elasticity α, making use of equation (6). the matching rate reads

\begin{matrix} θ^{i} q (θ^{i}) = k J^{i} {(ϵ_{u})}^{\frac{α}{1 - α}} . & (A11) \end{matrix}

A job is destroyed for one of two reasons. Either there is policy switch and the previous job-specific component ϵ is now below the new cut-off value or a new job-specific component falls below the existing component. It follows that

\begin{matrix} D_{t} = \int_{ϵ_{t}}^{ϵ_{d}^{i}} n_{t} (y) d y + λ F (ϵ_{d}^{i}) [N_{t} - \int_{ϵ_{1}}^{ϵ_{d}^{i}} n_{t} (y) d y] . & (A12) \end{matrix}

Finally, to close the model and obtaining N_t+1, we make use of the identity

\begin{matrix} N_{t + 1} = N_{t} + C_{t} - D_{t} . & (A13) \end{matrix}

REFERENCES

Ball, Laurence, and N. Gregory Mankiw, 1994, “Asymmetric Price Adjustment and Economic Fluctuations.” Economic Journal, Vol. 104 (March), pp. 247–61.
- Search Google Scholar
- Export Citation
Barro, Robert J., 1977, “Unanticipated Money Growth and Unemployment in the United States,” American Economic Review, Vol. 67 (March), pp. 101–115.
- Search Google Scholar
- Export Citation
Barro, Robert J., 1978, “Unanticipated Money, Output, and the Price Level in the United States,” Journal of Political Economy. Vol. 86 (August), pp. 549–80.
- Search Google Scholar
- Export Citation
Caballero, Ricardo J., and Eduardo Engel, 1992, “Price Rigidities, Asymmetries, and Output Fluctuations,” NBER Working Paper No. 4091 (Cambridge, Massachusetts: National Bureau of Economic Research, June).
- Search Google Scholar
- Export Citation
Caballero, Ricardo J., and Hammour, Mohamad L., 1995, “The Cleansing Effect of Recessions,” American Economic Review, Vol. 84 (December), pp. 1350–68.
- Search Google Scholar
- Export Citation
Caplin, Andrew, and John Leahy, 1991, “State-Dependent Pricing and the Dynamics of Money and Output,” Quarterly Journal of Economics, Vol. 106 (August), pp. 683–708.
- Search Google Scholar
- Export Citation
Caplin, Andrew, and John Leahy, 1996, “Monetary Policy as a Process of Search,” American Economic Review, Vol. 86 (September), pp. 689–702.
- Search Google Scholar
- Export Citation
Cover, James Peery, 1992, “Asymmetric Effects of Positive and Negative Money-Supply Shocks,” Quarterly Journal of Economics, Vol. 107 (November), pp. 1261–82.
- Search Google Scholar
- Export Citation
Davis, Steven J., and John Haltiwanger, 1990, “Gross Job Creation and Destruction: Microeconomic Evidence and Macroeconomic Implications,” in NBER Macroeconomics Annual 1990 (Cambridge, Massachusetts: MIT Press), pp. 123–68.
- Search Google Scholar
- Export Citation
Davis, Steven J., and John Haltiwanger, 1992, “Gross Job Creation, Gross Job Destruction and Employment Reallocation,” Quarterly Journal of Economics, Vol. 107 (August), pp. 819–63.
- Search Google Scholar
- Export Citation
Davis, Steven J., and John Haltiwanger, and Scott Shuh, 1996, Job Creation and Destruction (Cambridge, Massachusetts: MIT Press).
- Search Google Scholar
- Export Citation
De Long, J. Bradford, and Lawrence H. Summers, 1988, “How Does Macroeconomic Policy Affect Output?” Brookings Papers on Economic Activity: 2, pp. 433–80.
- Search Google Scholar
- Export Citation
Diamond, Peter A., 1971, “A Model of Price Adjustment,” Journal of Economic Theory, Vol. 3 (June), pp. 156–68.
- Search Google Scholar
- Export Citation
Diamond, Peter A., 1982, “Wage Determination and Efficiency in Search Equilibrium,” Review of Economic Studies, Vol. 49 (April), pp. 217–27.
- Search Google Scholar
- Export Citation
Garibaldi, Pietro, 1997, “Job Flow Dynamics and Firing Restrictions,” European Economic Review, Vol. 41, forthcoming.
- Search Google Scholar
- Export Citation
Jackman, Richard, and John Sutton, 1982, “Imperfect Capital Markets and the Monetarist Black Box: Liquidity Constraints, Inflation and the Asymmetric Effects of Interest Rate Policy,” Economic Journal, Vol. 92 (March), pp. 108–28.
- Search Google Scholar
- Export Citation
Johnson, Harry G., 1962, “Monetary Theory and Policy,” American Economic Review, Vol. 53 (June), pp. 335–84.
- Search Google Scholar
- Export Citation
Karras, Georgios, 1996, “Are the Output Effects of Monetary Policy Asymmetric? Evidence from a Sample of European Countries,” Oxford Bulletin of Economics and Statistics, Vol. 58 (May), pp. 267–68.
- Search Google Scholar
- Export Citation
Millard, Stephen P., and Dale T. Mortensen, 1994, “The Unemployment and Welfare Effects of Labour Market Policy: A Comparison of the U.S. and the U.K.” (unpublished; London, Bank of England).
- Search Google Scholar
- Export Citation
Mishkin, Frederic S., 1982, “Does Anticipated Policy Matter? An Econometric Investigation,” Journal of Political Economy, Vol. 90 (February), pp. 22–51.
- Search Google Scholar
- Export Citation
Morgan, Donald P., 1993, “Asymmetric Effects of Monetary Policy,” Federal Reserve Bank of Kansas City Economic Review, Vol. 78, No. 2, pp. 21–33.
- Search Google Scholar
- Export Citation
Mortensen, Dale T., 1970, “A Theory of Employment and Wage Dynamics,” in Microeconomic Foundations of Employment and Inflation Theory, ed. by Edmund S. Phelps et al. (New York: Norton), pp. 167–211.
- Search Google Scholar
- Export Citation
Mortensen, Dale T., 1994, “The Cyclical Behaviour of Job and Worker Flows,” Journal of Economic Dynamics and Control, Vol. 18 (November), pp. 1121–42.
- Search Google Scholar
- Export Citation
Mortensen, Dale T., and Christopher A. Pissarides, 1994, “Job Creation and Job Destruction in the Theory of Unemployment,” Review of Economic Studies, Vol. 61 (July), pp. 397–115.
- Search Google Scholar
- Export Citation
Pissarides, Christopher A., 1990, Equilibrium Unemployment Theory (Oxford: Basil Blackwell).
- Search Google Scholar
- Export Citation

Pietro Garibaldi is an Economist in the IMF’s Research Department. He holds a Ph.D. from the London School of Economics. He benefited from comments of Alberto Martini, Julian Berengaut, Giuseppe Bertola, John Leahy, Marcello Estevao, Peter Hole, and Jorge Márquez-Ruarte. He is particularly indebted to Humberto Lopez, and thanks seminar participants at the IMF, at the Board of Governors of the Federal Reserve, and at the 1997 European Economic Association Meeting in Toulouse.

The framework set forth by Mortensen and Pissarides has been recently expanded by Mortensen (1994), Millard and Mortensen (1994), and Garibaldi (1997).

Making use of equation (1), it is easy to derive that

\int_{ϵ_{d}}^{ϵ_{u}} J (x) d F (x) = J (ϵ_{u}) - \frac{1}{r + λ} \int_{ϵ_{d}}^{ϵ_{u}} F (x) .

The maximum likelihood estimates are very similar but the convergence takes place only with a 1 percent level of tolerance.

Table 8 reports the results for all the variables together, but regressions with each variable were also tried and this did not change the results.

Other Publishers

Save
Cite
Email this content

Share Link

Copy this link, or click below to email it to a friend
Email this content
or copy the link directly:

https://www.elibrary.imf.org/view/journals/024/1997/004/article-A007-en.xml
The link was not copied. Your current browser may not support copying via this button.

Link copied successfully

IMF Staff papers: Volume 44 No. 4

Author: International Monetary Fund. Research Dept.

Volume/Issue:: Volume 1997: Issue 004

Publisher:: International Monetary Fund

ISBN:: 9781451930962

ISSN:: 1020-7635

Pages:: 204

DOI:: https://doi.org/10.5089/9781451930962.024

Keywords
APPENDIX I
- Some Comparative Static Results
APPENDIX II
- Distribution of Employment
REFERENCES

View raw image

Net Employment Change in Response to Changes in Policy

View raw image

Gross Job Flows in Response to Changes in Policy

View raw image

Net and Cross Job Flows in Manufacturing. 1972:2 to 1988:3

^{Table 8.}

Job Flows Estimate 1972:2–1988:4—Oil Price Included

(All firms, two-stage ordinary least squares, ^a including controls for oil prices, money growth, and government borrowing)

First-stage regression
F_t-1	0.83***	F_t-2	-0.08
	(0.12)		(0.14)
G_t	0.31*	G_t-1	0.21
	0.19		(0.18)
G_t-2	0.03	π_t	0.47**
	(0.15)		(0.24)
π_t-1	-0.81**	π_t-2	0.45***
	(0.39)		(0.22)
π_t-2	0.45***	bg_t	0.95***
	(0.22)		0.39
oil_t	0.05***	mπ_t	0.007
	(0.02)		(0.04)
m_t-1	0.02
	(0.04)

Second-stage regression
	NET	JD	JC
tight_t-1	-0.92**	0.72**	-0.17
	(0.31)	(0.24)	(0.21)
tight_t-2	-0.43	0.33	-0.16*
	(0.31)	(0.25)	(0.13)
easy_t-1	0,73 **	-0.78**	-0.07
	(0.44)	(0.35)	(0.19)
easy_t-2	-0.46	0.22	-0.30
	(0.44)	(0.35)	(0.19)
oilt_-1	0.015	-0.08	-0.01
	(0.07)	(0.05)	(0.03)
oilt_-1	-0.03	0.01	0.01
	(0.07)	(0.05)	(0.03)
SUM(tight)^{^b}	-1.35	1.06	-0.33
SUM(easy) ^{^c}	0.27	-0.56	-0.38
tight = 0 ^{^d}	5.44***	5.38***	1.47
easy = 0 ^{^e}	1.72	2.54	1.50
easy = tight^{^f}	3.68*	5.59**	0.019

^{^a}Regressions include a constant and two lags of the dependent variable.

^{^b}Sum of the tight coefficients.

^{^c}Sum of the easy coefficients.

^{^d}F-statistic of the hypothesis that the coefficients on tight are jointly zero.

^{^e}F-statistic of the hypothesis that the coefficients on easy are jointly zero.

^{^f}F-statistic of the hypothesis that the sum of the coefficients on tight equals the sum of the coefficients on easy.

Topics

Countries

Series

About

Resources

Abstract

I. Existing Empirical Evidence

II. Concepts and Notations

III. A Minimalist Model: Steady State

IV. The Model with Tight and Easy Policy

Model

Discussion and Simulation

V. Empirical Evidence

Data

Evidence of Asymmetry

Robustness Checks

VI. Conclusions

APPENDIX I

Some Comparative Static Results

APPENDIX II

Distribution of Employment

REFERENCES

Same Series

Other IMF Content

Other Publishers

Asian Development Bank

Inter-American Development Bank

International Labour Organization

The World Bank

Share Link

Table of Contents

Headings

Figures

Metrics