The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Working Paper Series Technical Appendix for " Frictional Wage Dispersion in Search Models: A Quantitative Assessment" WP 06-08 Andreas Hornstein Federal Reserve Bank of Richmond Per Krusell Princeton University Giovanni L. Violante New York University This paper can be downloaded without charge from: http://www.richmondfed.org/publications/ Technical Appendix for “Frictional Wage Dispersion in Search Models: A Quantitative Assessment” ∗ Andreas Hornstein† Per Krusell‡ Giovanni L. Violante§ Federal Reserve Bank of Richmond Working Paper 2006-08 Abstract In this Technical Appendix to Hornstein, Krusell, and Violante (2006) (HKV, 2006, hereafter) we provide a detailed characterization of the search model with (1) wage shocks during employment and (2) on-the-job search outlined in Sections 6 and 7 of that paper, and we derive all of the results that are only stated in HKV (2006). In particular, we derive the expressions for our preferred measure of frictional wage inequality: the ratio of average wages to the reservation wage, or, the ‘mean-min’ wage ratio. Keywords: labor market, wage inequality, search frictions, job search JEL Classification: D83, E24, J31, J41, J63, J64 ∗ The views expressed here are those of the authors and do not necessarily reflect those of the Federal Reserve Bank of Richmond or the Federal Reserve System. Our e-mail addresses are: andreas.hornstein@rich.frb.org, pkrusell@princeton.edu, and gianluca.violante@nyu.edu † Federal Reserve Bank of Richmond ‡ Princeton University, CEPR, and NBER § New York University and CEPR 1 Wage shocks during employment In this section, we characterize the equilibrium of the search model with wage shocks while employed when wage shocks when employed and unemployed come from the same distribution. We first state the discrete-time approximation of the search model for a fixed and finite length of the time period. We then derive the continuous-time representation as the limit of the discrete-time approximation when the length of the time period becomes arbitrarily small. We show that the Bellman equations for the employment and unemployment values for the continuous- and discrete-time version are the same. Our derivation of the mean-min ratio for wage inequality is therefore independent of the time representation. We then show that in the discrete-time version the first-order autocorrelation coefficient of wages is one minus the arrival rate of wage changes. Finally, we consider a variation of the baseline model where wage shocks when employed come from a different distribution than wage shocks when unemployed. In particular, we study the Mortensen-Pissarides (1994) environment where unemployed workers on meeting a job always receive the highest wage. We describe the parameters of the environment in terms of the continuous-time framework. The economy is populated by ex-ante equal, risk-neutral, infinitely lived individuals who discount the future at rate r. Unemployed agents receive job offers at the instantaneous rate λu , and the wage of employed agents changes at the instantaneous rate δ. Conditionally on receiving a wage change, the wage is drawn from a well-behaved distribution function F (w) with upper support w max . Draws are i.i.d. over time and across agents. Note that employed and unemployed agents sample from the same wage distribution. If a job offer w is accepted, the worker is paid a wage w, until the next wage-change event occurs. While unemployed, the worker receives a utility flow b which includes unemployment benefits and a value of leisure and home production, net of search costs. We only consider steady-state allocations. 1 1.1 Discrete time versus continuous time The discrete-time approximations of the Bellman equations for the value of employment, W (w), and unemployment, U , are Z −r∆ W (w) = w∆ + e δ∆ max {W (z), U } dF (z) + (1 − δ∆) W (w) , Z −r∆ U = b∆ + e λu ∆ max {W (z), U } dF (z) + (1 − λu ∆) U , (1) (2) where ∆ is the length of the time interval, and λu ∆ (δ∆) is the probability that an (un)employed worker receives a wage offer at the end of the interval ∆. Using the definition of the reservation wage, W (w ∗ ) = U , and rearranging terms the value equations can be rewritten as Z wmax −r∆ −r∆ ∗ 1−e W (w) = w∆ + e δ∆ [W (z) − W (w)] dF (z) − δ∆F (w ) [W (w) − U ] , w∗ Z wmax −r∆ −r∆ 1−e U = b∆ + e λu ∆ [W (z) − U ] dF (z) w∗ Dividing by the length of the time interval and taking the limit as ∆ → 0 we get the continuous-time Bellman equations Z wmax rW (w) = w + δ [W (z) − W (w)] dF (z) − δF (w ∗ ) [W (w) − U ] , ∗ w Z wmax rU = b + λu [W (z) − U ] dF (z) . (3) (4) w∗ Rather than studying the continuous-time limit of the search model with wage changes, we can just use the discrete-time approximation and consider a unit length interval, ∆ = 1. In this case we get the following expressions for the value functions of being employed and unemployed: Z wmax ∗ (1 − β) W (w) = w + β δ [W (z) − W (w)] dF (z) − δF (w ) [W (w) − U ] w∗ Z wmax (1 − β) U = b + βλu [W (z) − U ] dF (z) , w∗ where β ≡ e −r ≡ 1/ (1 + r̄). Note that we can rewrite these discrete-time value equations as r̄ W̄ (w) = w + δ Z r̄Ū = b + λu w max w∗ Z wmax w∗ W̄ (z) − W̄ (w) dF (z) − δF (w ∗ ) W̄ (w) − Ū W̄ (z) − Ū dF (z) , 2 (5) (6) where W̄ ≡ W/ (1 + r̄) and Ū ≡ U/ (1 + r̄). Expressions (5) and (6) are formally equivalent to the expressions (3) and (4) for the continuous-time value functions. Therefore, the results for the mean-min ratio also apply for the discrete-time version of the paper. 1.2 The reservation wage As a first step towards deriving the mean-min wage ratio for the continuous-time model we characterize the reservation wage. For this purpose, evaluate the employment value expression (3) at w ∗ and use the definition of the reservation wage, W (w ∗ ) = U , to get ∗ rU = w + δ Z w max [W (z) − W (w ∗ )] dF (z) . w∗ Now substitute the unemployment value expression (4) for the left-hand side and solve for the reservation wage ∗ w = b + (λu − δ) Z w max [W (z) − W (w ∗ )] dF (z) . (7) w∗ Integration by parts on the right-hand side yields Z wmax Z wmax w max ∗ ∗ [W (z) − W (w )] dF (z) = [(W (z) − W (w )) F (z)]w∗ − W 0 (z) F (z) dz ∗ ∗ w w Z wmax = W (w max ) − W (w ∗ ) − W 0 (z) F (z) dz w∗ Z wmax = W 0 (z) [1 − F (z)] dz. (8) w∗ From the employment value expression (3) it follows that W 0 (w) = 1 . r+δ Hence the reservation wage expression is Z max λu − δ w ∗ w = b+ [1 − F (z)] dz r + δ w∗ Z max (λu − δ) [1 − F ∗ ] w 1 F (z) = b+ − dz r+δ 1 − F∗ 1 − F∗ w∗ with F ∗ = F (w ∗ ). 3 (9) (10) 1.3 The equilibrium wage distribution We now construct the equilibrium wage distribution G (w) implied by the interaction of the wage-offer distribution and the reservation wage. The measure of agents with wage below w is (1 − u) G (w) . Agents leave this stock because their wage changes and their new wage is either less than the reservation wage or higher than the current wage. Agents enter this stock if they were unemployed and receive an acceptable wage offer below w, or if they were employed at wage above w and are forced to accept a lower wage; hence (1 − u) G (w) δ {F (w ∗ ) + [1 − F (w)]} = {uλu + (1 − u) [1 − G (w)] δ} [F (w) − F (w ∗ )] . (11) We can solve this expression for the equilibrium wage distribution as a function of the wage offer distribution: λu u G (w) = + 1 [F (w) − F (w ∗ )] . δ (1 − u) (12) In steady state, the inflows and outflows from employment balance: (1 − u) δF (w ∗ ) = uλu [1 − F (w ∗ )] . Using the expression for steady-state employment in (12) we get G (w) = F (w) − F (w ∗ ) . 1 − F (w ∗ ) (13) Thus, the equilibrium wage distribution with and without wage shocks during employment are the same, namely the wage-offer distribution truncated at the reservation wage. 1.4 The mean-min ratio Based on the equilibrium wage distribution we can calculate the average wage of employed workers as w̄ = Z w max w∗ dF (w) w max − w ∗ F ∗ w = − 1 − F∗ 1 − F∗ Z w max w∗ F (z) dz. 1 − F∗ (14) Solving the average-wage expression (14) for the right-hand side integral term and substituting this term for the corresponding integral in the reservation-wage expression (10) 4 yields w ∗ Z wmax 1 w max − w ∗ F ∗ (λu − δ) [1 − F ∗ ] = b+ dz + w̄ − r+δ 1 − F∗ 1 − F∗ w∗ (λu − δ) [1 − F ∗ ] = b+ [w̄ − w ∗ ] . r+δ (15) Using the definition of the replacement rate, b = ρw̄, we can solve equation (15) for the reservation wage and obtain an expression for the mean-min ratio, that is, the ratio of average wages to the reservation wage, w̄ = w∗ (λu −δ)[1−F (w ∗ )] r+δ (λu −δ)[1−F (w ∗ )] r+δ +1 +ρ = λ∗u −δ+σ ∗ r+δ λ∗u −δ+σ ∗ r+δ +1 +ρ , (16) with λ∗u ≡ (1 − F ∗ ) λu and σ ∗ ≡ δF ∗ . This is equation (13) in Section 6 of HKV (2006). Note that as δ goes to infinity 1.5 w̄ w∗ goes to 1/ρ. Wage persistence in the discrete time model It is straightforward to show that the equilibrium wage distribution for the discrete-time and the continuous-time versions of the model are the same; we now work with the former. We need the expected value of the cross-product of today’s and tomorrow’s wage, conditional on being employed in both periods, to calculate the autocorrelation coefficient. We proceed in two steps: first, we obtain the value conditional on today’s wage, and then we integrate over today’s wage to get the unconditional expectation. E [w 0 w|w] = (1 − δ) w 2 + δwE [w̃|w̃ ≥ w ∗ ] = (1 − δ) w 2 + δww̄ E [w 0 w] = (1 − δ) E w 2 + δ w̄ 2 We can now define the first-order autocorrelation coefficient as (1 − δ) E [w 2 ] + δ w̄ 2 − w̄ 2 V ar (w) (1 − δ) (E [w 2 ] − w̄ 2 ) = V ar (w) = 1−δ ρ = 5 1.6 The mean-min ratio for a Mortensen-Pissarides (1994) environment Suppose now that an unemployed worker who receives a wage offer always receives the highest wage w max , whereas wage changes of employed workers continue to be drawn from the distribution F . We will show that the mean-min ratio that we previously derived for our baseline model, equation (16), represents an upper bound for the mean-min ratio in this Mortensen-Pissarides (1994) environment. The value function equation for an employed worker, (3), remains unchanged, but the value function equation for an unemployed worker is now rU = b + λu [W (w max ) − U ] . (17) Following the same steps as in Section 1.2 we derive the modified expression for the reservation wage: λu − δ max δ w =b+ [w − w∗] + r+δ r+δ ∗ Z w max F (z) dz. (18) w∗ The modified steady-state expression characterizing the equilibrium wage distribution is now (1 − u) G (w) δ {F (w ∗ ) + [1 − F (w)]} (19) = (1 − u) [1 − G (w)] δ [F (w) − F (w ∗ )] for w < w max . Note that there are no inflows from the pool of unemployed since all unemployed workers who receive wage offers receive the highest wage. Thus, the equilibrium wage distribution for w < w max is G (w) = F (w) − F (w ∗ ) . (20) Since all unemployed workers receive the highest wage with probability one there is now a mass point at w = w max , that is, the cumulative density function is discontinuous at w max . Integrating the wage with respect to the equilibrium wage distribution then yields the average wage Z wmax Z max ∗ max ∗ max ∗ w̄ = wdF (w) + w F (w ) = w + F (w ) (w − w )− w∗ 6 w max F (w) dw. (21) w∗ Solving the average-wage expression (21) for the right-hand side integral term and substituting this term for the corresponding integral in the reservation-wage expression (18) yields λu − (1 − F ∗ ) δ δ λu + δF ∗ max ∗ 1+ w = ρ− w̄ + w r+δ r+δ r+δ λu − δ (1 − F ∗ ) w̄. > ρ+ r+δ (22) Note that the last inequality implies that the mean-min ratio for this setup is bounded above by that of the baseline economy with wage shocks given in (16). 2 On-the-job search We describe a general version of the on-the-job search model that includes forced jobto-job mobility and wage shocks (the models in Section 7 of HKV, 2006). The Bellman equations for the employment and unemployment values are Z rW (w) = w + λw max {W (z) − W (w) , 0} dF (z) − σ [W (w) − U ] Z +φ [max {W (z) , U } − W (w)] dF (z) Z rU = b + λu max {W (z) − U, 0} dF (z) The basic on-the-job search model in HKV (2006), Section 7.1, assumes that a worker receives outside wage offers at a rate λw . Without loss of generality, and motivated by what equilibrium firm behavior would dictate, we assume that the wage-offer distribution is such that unemployed workers accept all wage offers: F (w ∗ ) = 0. A worker can always reject a wage offer and keep the current wage. The worker may lose the current job at an exogenous separation rate, σ. On-the-job search with forced job-to-job mobility in HKV (2006), Section 7.2, assumes that for some wage offers, the worker just has to take the offer, even if the new job pays less than the current job. Forced job mobility is reflected in the parameter φ > 0. In the basic on-the-job search model φ = 0. On-the-job search with wage shocks in HKV (2006), Section 7.2, is essentially the same as forced job-to-job mobility. The only difference is in its implications for observable transitions. The arrival of a type φ wage offer now results 7 in a wage cut on the existing job, rather than a separation and transfer to a lower-paying job. Finally, we analyze the on-the-job search model of Christensen et al. (2005) where search effort is endogenous, as discussed in HKV (2006), Section 7.1. 2.1 The reservation wage Workers continue to follow reservation-wage strategies and the Bellman equations can be rewritten as Z w max rW (w) = w + λw [W (z) − W (w)] dF (z) − σ [W (w) − U ] w Z wmax +φ [W (z) − W (w)] dF (z) w∗ Z wmax = w + (λw + φ) [W (z) − W (w)] dF (z) − σ [W (w) − U ] w Z w +φ [W (w) − W (z)] dF (z) w∗ Z wmax rU = b + λu [W (z) − U ] dF (z) (23) (24) w∗ Evaluate the employment value equation (23) at w ∗ , using the reservation wage property, W (w ∗ ) = U , and the unemployment value expression (24) to obtain Z wmax ∗ rU = w + (λw + φ) [W (z) − W (w ∗ )] dF (z) ∗ w Z wmax = b + λu [W (z) − W (w ∗ )] dF (z) . w∗ We can solve this expression for the reservation wage Z wmax ∗ w = b + (λu − λw − φ) [W (z) − W (w ∗ )] dF (z) . (25) w∗ As with equation (7), we can integrate the right-hand side integral by parts, as in (8), and get the reservation-wage equation ∗ w = b + (λu − λw − φ) Z w max W 0 (z) [1 − F (z)] dz. (26) w∗ Note that differentiating the employment value equation (23) with respect to the current wage yields W 0 (w) = 1 . r + σ + φ + λw [1 − F (w)] 8 (27) Substituting (27) in (26) we can rewrite the reservation wage as ∗ w = b + (λu − λw − φ) 2.2 Z w max w∗ 1 − F (z) dz. r + σ + φ + λw [1 − F (z)] (28) The equilibrium wage distribution We now construct the equilibrium wage distribution G (w) implied by the interaction of the wage-offer distribution and the reservation wage. The measure of agents with wage below w is (1 − u) G (w) . Agents leave this stock because (1) they get separated at rate σ, (2) they receive an outside offer which they accept at rate λw [1 − F (w)], or (3) they are forced to leave at rate φ but are lucky enough to get an offer above w. Workers enter this stock if (1) they were unemployed and receive a wage offer below w, or (2) they were employed at wage above w and are forced to accept a lower wage. In a steady state the inflows and outflows balance: (1 − u) G (w) {σ + (λw + φ) [1 − F (w)]} (29) = uλu F (w) + φ (1 − u) [1 − G (w)] F (w) . We can solve this expression for the equilibrium wage distribution as a function of the wage offer distribution: G (w) = λu u + φ (1 − u) F (w) · 1−u σ + φ + λw [1 − F (w)] (30) In steady state, if all the job offers are above w ∗ so that F (w ∗ ) = 0, uλu = (1 − u) σ. Hence G (w) = (σ + φ) F (w) σ + φ + λw [1 − F (w)] (31) and σ + φ + λw [1 − F (w)] σ + φ + λw [1 − F (w)] r + σ + φ + λw ' [1 − F (w)] r + σ + φ + λw [1 − F (w)] 1 − G (w) = 9 (32) 2.3 The mean-min ratio The average wage is Z w̄ = w max max [wG (w)]w w∗ Z w max wdG (z) = − w∗ Z wmax = w max − G (z) dz w∗ Z wmax max ∗ ∗ = [w −w ]+w − G (z) dz w∗ Z wmax ∗ = w + [1 − G (z)] dz. w∗ G (z) dz (33) w∗ Solve the wage distribution expression (32) for 1 − F and use it in the reservation-wage expression (28) to get λu − λ w − φ w 'b+ r + σ + φ + λw ∗ Z w max [1 − G (z)] dz. w∗ Finally substituting for the integral term from the average-wage equation (33) we can solve for the mean-min ratio: λu − λ w − φ (w̄ − w ∗ ) r + σ + φ + λw λu −λw −φ +1 r+σ+φ+λw . λu −λw −φ +ρ r+σ+φ+λw w ∗ ' ρw̄ + Mm ' ⇒ (34) Equation (34) corresponds to equations (17) and (20) in Section 7 of HKV (2006). 2.4 Turnover rates in the basic on-the-job search model In the basic on-the-job search model there are exogenous separations, σ > 0, but no forced wage changes, φ = 0. The average completed job tenure in the model is Z wmax dG (w) τ= . σ + λw [1 − F (w)] w∗ (35) Recall the equilibrium wage distribution from (31) and solve that expression for the wageoffer distribution 1 − F (w) = σ [1 − G (w)] . σ + λw G (w) 10 Substitute this expression in the expression for average job tenure (35) and we get τ = = Note that Thus Z Z dG (w) σ+ w∗ 1 σ (λw + σ) w max w∗ w max = w max [σ + λw G (w)] dG (w) σ (λw + σ) G (w) dG (w) . Z σ[1−G(w)] λw σ+λ w∗ w G(w) Z wmax σ + λw w∗ wmax G (w) dG (w) = G (w) w∗ − Z 2 Z (36) w max G (w) dG (w) . w∗ w max G (w) dG (w) = 1/2. w∗ Hence, the average job tenure is σ + λw /2 τ = ∈ σ (λw + σ) 1 1 , 2σ σ . (37) The (total) separation rate for employed workers, the sum of exogenous separations and job-to-job transitions, is sep = σ + λw Z w max [1 − F (w)] dG (w) . (38) w∗ Hence, the share of separations attributable to job-to-job separations is R wmax λw w∗ [1 − F (w)] dG (w) jjsep = . R wmax sep σ + λw ∗ [1 − F (w)] dG (w) (39) w Following Nagypal (2005), and integrating jjsep by parts, yields jjsep = λw Z w max [1 − F (w)] dG (w) Z wmax = λw − λw F (w) dG (w) w∗ w∗ max (w) G (w)]w w∗ Z w max = λw − λw [F + λw G (w) dF (w) w∗ Z wmax = λw G (w) dF (w) w∗ Z wmax F (w) = λw σ dF (w) . σ + λw [1 − F (w)] w∗ 11 (40) The last step uses equation (31) for G. Now change the variable of integration to z = F (w), and we get Z 1 0 λw {z}10 + (λw + σ) {log [σ + λw (1 − z)]}10 z dz = − σ + λw [1 − z] λ2w λw + (λw + σ) (log σ − log (σ + λw )) = − λ2w w (λw + σ) log σ+λ 1 σ = − . λ2w λw (41) Hence, the job-to-job separation rate is σ (λw + σ) log jjsep = λw σ+λw σ −σ (42) and the total separation rate is σ (λw + σ) log sep = jjsep + σ = λw σ+λw σ . (43) The share of separations attributable to job-to-job transitions is then jjsep λw / (σ + λw ) =1− . sep log [(σ + λw ) /σ] 2.5 (44) Wage cuts in the model with on-the-job wage changes In this version of the model, not only does a worker receive outside wage offers at the rate λw , but the wage on the current job is also changing with arrival rate φ. If the new wage is less than the current wage, the worker has to accept a pay cut. The average rate at which workers are forced to take a wage cut is Z wmax f cut = φ F (w) dG (w) w∗ Z wmax w max = φ [F (w) G (w)]w∗ − φ G (w) dF (w) w∗ Z wmax = φ−φ G (w) dF (w) . (45) w∗ Now substitute for the equilibrium wage distribution from (31) and we get Z wmax F (w) dF (w) , f cut = φ − φσ̂ σ̂ + λw [1 − F (w)] w∗ 12 (46) where σ̂ ≡ σ + φ. We have previously derived the right-hand side integral: expressions (40) and (41). Use equation (41) for the integral and we get " # w (λw + σ̂) log σ̂+λ 1 σ̂ f cut = φ − φσ̂ − λ2w λw " # w σ̂ (λw + σ̂) log σ̂+λ σ̂ σ̂ = φ 1+ − . λw λ2w (47) With f cut being the average arrival rate of a wage cut, the fraction of employed workers that receive a wage cut in a unit time period is 1 − exp (−f cut). 2.6 Endogenous effort choice in the model with on-the-job search In the Christensen et al. (2005) model, the utility cost to attain a contact rate λ is c(λ), with c0 > 0 and c00 > 0. Thus, optimal search choice will deliver a policy function λ(w), where we use the notation λ∗ ≡ λ(w ∗ ). The first-order condition at the reservation wage w ∗ reads Z wmax 1 − F (z) 0 ∗ c (λ ) = dz. r + σ + λ(z)(1 − F (z)) w∗ In steady state, flows for the workers employed at wage w or lower satisfy Z w (1 − u)G(w)σ + (1 − u) [1 − F (w)] λ(z)dG(z) = uλ∗ F (w) (48) (49) w∗ and for the unemployment pool the flows satisfy the relation uλ∗ = (1 − u)σ. (50) Combining the last two equations we obtain σF (w) = G(w)σ + [1 − F (w)] Z w λ(z)dG(z). w∗ Because λ(z) is decreasing, we have σF (w) ≤ G(w) [σ + λ∗ (1 − F (w))] , so G(w) ≥ σF (w) , σ + λ∗ (1 − F (w)) 13 or 1 − G(w) ≤ (σ + λ∗ )(1 − F (w)) . σ + λ∗ (1 − F (w)) This implies 1 − F (w) 1 − G(w) σ + λ∗ (1 − F (w)) ≥ . r + σ + λ∗ (1 − F (w)) r + σ + λ∗ (1 − F (w)) σ + λ∗ (51) Since r is small relative to σ, we have r + σ + λ∗ (1 − F (w)) σ + λ∗ (1 − F (w)) ' , σ + λ∗ r + σ + λ∗ and hence, from the inequality in (51) 1 − G(w) 1 − F (w) ≥ ∗ r + σ + λ (1 − F (w)) r + σ + λ∗ (52) almost holds, for all w. Going back to the FOC for effort (48), using the fact that λ(z) is decreasing, we have 0 ∗ c (λ ) ≥ Z w max w∗ 1 − F (z) dz, r + σ + λ∗ (1 − F (z)) and then, using the inequality in (52), we obtain that an approximate inequality for c0 (λ∗ ) is given by 0 ∗ c (λ ) ≥ Z w max w∗ 1 − G(z) dz = r + σ + λ∗ R wmax w∗ (1 − G(z))dz w̄ − w ∗ = . r + σ + λ∗ r + σ + λ∗ Assuming, as done in Christensen et al. (2005), that c(λ) = λγ , we have that c0 (λ) = γc(λ)/λ so that c(λ∗ ) ≥ λ∗ (w̄ − w ∗ ) γ(r + σ + λ∗ ) which yields the lower bound for search effort during unemployment discussed in the main text. 14 References [1] Christensen, B.J., R. Lentz, D.T. Mortensen, G.R. Neumann, and A. Werwatz (2005). On-the-Job Search and the Wage Distribution, Journal of Labor Economics, vol. 25(1), 31-58. [2] Hornstein, Andreas, Per Krusell, and Giovanni L. Violante (2006). Frictional Wage Dispersion in Search Models: A Quantitative Assessment, Federal Reserve Bank of Richmond Working Paper 2006-07, available at http://www.richmondfed.org/publications/economic research/working papers/index.cfm. [3] Mortensen and Pissarides (1994). Job Creation and Job Destruction in the Theory of Unemployment, Review of Economic Studies, vol. 61(3), 397-415. [4] Nagypal, E. (2005). Worker Reallocation Over the Business Cycle: the Importance of Job-to-Job Transitions, mimeo, Northwestern University. 15