A. Appendix
A.4. The Log-Linearized System of the DMP Model
60 A.3.16. The Taylor Rule
Using the properties of the natural logarithm, the Taylor rule described in equation 6.2-1 can be expressed as
ln(π π‘) β ln(π )
= ππ[ln(π π‘β1) β ln(π )]
+ (1 β ππ)[ππ(ln(ππ‘) β ln(π)) + ππ(ln(ππ‘) β ln(π))] + ππ‘.
A first order Taylor expansion around steady-state yields, after dropping constant terms, the form
π Μπ‘ = πππ Μπ‘β1+ (1 β ππ)[πππΜπ‘+ ππYΜt] + ππ‘. 0.a Consequently, the CEE variant described in section 3.8 has the log-linearized form
π Μπ‘= πππ Μπ‘β1+ (1 β ππ)[ππΞπ‘πΜπ‘+1+ ππYΜt] + ππ‘. 0.b A.3.17. Real Marginal Cost
As derived in appendix A.1, cost minimization in the intermediate goods sector implies that real marginal costs have the form of equation 3.5-2. Log-linearization yields
π Μπ‘= πΌπΜπ‘π+ (1 β πΌ)π€Μπ‘π Μπ‘. A.3.17 A.3.18. The Cobb-Douglas Production Function
Equation 5.4.2-1 describes the production function in the intermediate goods sector, which has the following log-linearized form:
π¦Μπ‘= πΌπΜπ‘+ (1 β πΌ)πΜπ‘. A.3.18 A.3.19. The Real Wage as Defined in CEE
CEE define the wage in their thirteen variable system as shown in equation A.3.3.i. For ease of reference, the corresponding log-linearized form, equation A.3.3.ii, is quoted again:
π€Μ Μπ‘ = π€Μπ‘β πΜπ‘. A.3.19
A.3.20. Real Cash Balances
Consistent with equation A.3.19, real cash balances in log-linearized form give
πΜ Μπ‘ = πΜπ‘β πΜπ‘. A.3.20
61 A.4.1. The Aggregate Resource Constraint
In addition to the standard resource constraint used in CEE, vacancy posting costs need to be taken into account. Moreover, the Search and Matching framework now determines the aggregate amount of the homogeneous labor input used in the intermediate production process.
A few considerations are needed to derive the final equation. As explained in section 3.5, capital and labor is rented in perfectly competitive factor markets. Consequently, each producer faces identical factor prices and constant returns to scale in the production function imply that there is no need to distinguish between firms. Thus, the unweighted average of output can be expressed as
πβ = β« ππππ = β« πππΌππ1βπΌππ = ππΌπ1βπΌ,
1
0 1
0
A.4.1.i
where π and π represent the aggregate amount of capital and of the homogenous labor input:54 π = β« ππππ
1
0
; π = β« ππππ.
1
0
A.4.1.ii
Time subscripts are omitted in order to be consistent with the derivation provided in the working paper of Christiano et al. (2001). Eventually, time subscripts will be reinserted to provide a correct representation of the log-linearized resource constraint. The authors list two reasons why this is not the best way to express the aggregate resource constraint. Firstly, total output should be related to total labor, calculated as in equation 5.5-1. This is preferable because ππ‘ characterizes labor as measured in the data, whereas ππ‘ is the amount of the homogenous labor input used in the intermediate goods production. Secondly, πβ does not have any meaningful economic interpretation, because it is simply a sum of differentiated intermediate goods. The latter problem can be solved by substituting optimal demand for intermediate good π, characterized by equation 3.4-2, into A.4.1.i:
πβ = β« ππππ = β« (π ππ)
ππ ππβ1
πππ = ππ
ππ ππβ1(πβ)
ππ 1βππ,
1
0 1
0
where πβ denotes the weighted average of all individual prices.55 Solving for π and substituting out πβ with the last term in equation A.4.1.i. gives
π = (πβ π)
ππ ππβ1
ππΌπ1βπΌ.
Instead of the unweighted average output, this equation now contains the aggregate output, which is allocated between consumption, investment, the resources used up in capital utilization
54 CEE use capital letters, whereas here lower case letters are used to avoid confusion with steady-state values.
55 Note that the weights are not identical to the ones used in equation 3.4-3 because πβ= [β« ππ
ππ 1βππ 1
0 ππ]
1βππ ππ
.
62 as well as the total vacancy posting costs. Consequently, the model economy is facing the following modified resource constraint:
π + π + π(π’)πΜ + π π£ β€ (πβ π)
ππ ππβ1
ππΌπ1βπΌ.
The last modification needed regards the total labor force. Substituting equation 5.1-2, using definitions 5.5-1 and A.4.1.ii, gives
π + π + π(π’)πΜ + π π£ β€ (πβ π)
ππ ππβ1
ππΌ(π΄π)1βπΌ, A.4.1.iii
which corresponds to the sought-after resource constraint of equation 5.5-2. CEE state that the term in front of the production function is similar to the so called Solow residual, which is essentially the TFP part, denoted as π§π‘, of the production function described in equation 6.5-1.
Yun (1996) showed that in a first order approximation, as used here, this βSolow residualβ is a constant. As CEE highlight, this result can be found by obtaining expressions for πβ/π and πβ/π respectively.
Equivalent to the way equation 3.4-3 was rewritten in the form of equation 3.5-7, ππ‘β and , ππ‘βcan be expressed as
ππ‘β = [(1 β ππ)πΜπ‘
ππ 1βππ
+ ππ(ππ‘β1ππ‘β1β )
ππ 1βππ]
1βππ ππ
.
Dividing this term by ππ‘ yields
ππ‘β= [(1 β ππ)πΜπ‘
ππ 1βππ
+ ππ(ππ‘β1ππ‘β1β )
ππ 1βππ]
1βππ ππ
, A.4.1.iv
where πβ= πβ/ππ‘. Log-linearizing this expression around the steady-state results in πΜπ‘β = (1 β ππ)πΜΜπ‘+ ππ(πΜπ‘β1β πΜπ‘+ πΜπ‘β1β ).
After substituting equation 3.5-8 into the log-linearized expression one obtains πΜπ‘β = πππΜπ‘β1β .
Letβs supddpose πΜ0β = 0. Therefore, πΜπ‘β = 0 for all future time periods and consequently, ππ‘β = ππ‘ for all possible realizations of t. Since the almost identical steps apply for ππ‘β and this part is not required in the augmented model, this derivation is left out here.56 During log-linearization, the transformations above can be used to rationalize considering πβ/π as constant
56 CEE provide this part in their working paper appendix (2001).
63 equal to unity. For this reason, the following steps obtain the log-linearized resource constraint of equation A.4.1.iii. Using the properties of the natural logarithm, this constraint can be written as
ln(π(π’)πΜ + π + π + π π£) = ππ
ππβ 1ln (πβ
π) + (1 β πΌ) ln(π΄) + (1 β πΌ) ln(π) + πΌ ln(π), where the inequality sign was switched to an equality sign because at the optimum all output is used up. After reinserting the omitted time subscripts, a linear first order Taylor expansion yields,
ln(π(π)πΎΜ + πΆ + πΌ + π π) +1
π(πβ²(π’)πΎΜ ππ’Μπ‘+ π(π)πΎΜ πΜ Μπ‘+ ππΜπ‘+ ππΜπ‘+ π ππ£Μπ‘)
= (1 β πΌ) ln(π΄) + (1 β πΌ)π΄Μπ‘+ (1 β πΌ) ln(π) + (1 β πΌ)πΜπ‘+ πΌ ln(πΎ) + πΌπΜπ‘ Using the relationship ππ‘ = π’π‘πΜ π‘ and equation A.3.A.3.5.i, this can be rewritten as
πβ²(π’)πΎ
ππ’Μπ‘+π ππΜπ‘+ π
ππΜπ‘+π π
π = (1 β πΌ)(π΄Μπ‘+ πΜπ‘) + πΌπΜπ‘,
where the steady-state terms of the log-linearized resource constraint are already cancelled out. It helps to define
π π = π/π, π π = π/π = πΏ πΎ/π, π π = πΎ/π, π ππ£ = ππ/π.
Thus, in combination with equation A.3.A.3.5.ii, the log-linearized resource constraint can be written in the final version, which has the following form:
[1
π½β (1 β πΏ)]π π
π ππ’Μπ‘+ πΜπ‘+πΏπ π
π π πΜπ‘+π π π£
π π π£Μπ‘ = πΌ
π ππΜπ‘+1 β πΌ
π π (π΄Μπ‘+ πΜπ‘). A.4.1 A.4.2. The Labor Augmenting Technology Shock
Log-linearizing the labor augmenting technology shock of equation 5.1-3 yields
π΄Μπ‘ = πΏπ΄Μπ‘β1 β ππ‘. A.4.2 A.4.3. The Price Charged for the Aggregate Labor Input
Equation 5.2-9 in log-linearized form yields
πΜπ‘π€ = ππΜπ‘π€ A.4.3
A.4.4. The DMP Production Function
The homogenous aggregate labor input is produced according to equation 5.1-2. Using equation A.4.1.ii one obtains the following log-linearized form:
πΜπ‘ = π΄Μπ‘+πΜπ‘. A.4.4
64 A.4.5. The Law of Motion for Employment
The employment accumulation equation is described by equation 4-9, which has the log-linearized form
ππΜπ‘= (1 β π)ππΜπ‘β1+ ππΜπ‘β1. A.4.5 A.4.6. The Unemployment Equation
Log-linearizing equation 4-1 results in
(π/π)πΜπ‘= βπ’Μπ‘. A.4.5
A.4.7. The Matching Function
Since the matching function defined in equation 4-2 has the form of a standard Cobb-Douglas function, the log-linearization was already described in section A.2 of the appendix. Applied to the matching function, log-linearization yields
πΜπ‘= πΎΜπ‘+ ππ’Μπ‘+ (1 β π)π£Μπ‘ A.4.7 A.4.8. Labor Market Tightness
Equation 4-3 has the following log-linearized form:
πΜπ‘= π£Μπ‘β π’Μπ‘ A.4.8
A.4.9. The Wage Equation
As already noted in section 4.3, the equilibrium real wage π€π‘ is obtained by maximizing the generalized Nash product denoted by
maxπ€π‘
(π½π‘β ππ‘)1βπ(π»π‘β πΌπ‘)π. For convenience, all relevant equations are stated again:
ο· Value of employment 4.1-1: π»π‘= π€π‘+ Ξπ‘π½[(1 β π)π»π‘+1+ ππΌπ‘+1]
ο· Value of unemployment 4.1-2: πΌπ‘ = π + Ξπ‘π½[ππ‘π»π‘+1+ (1 β ππ‘)πΌπ‘+1]
ο· Value of a vacancy 4.2-1: ππ‘ = βπ + Ξπ‘π½[π(ππ‘)π½π‘+1+ (1 β π(ππ‘))ππ‘+1]
ο· Value of a job 4.2-2: π½π‘ = π΄π‘β π€π‘+ Ξπ‘π½[(1 β π)π½π‘+1+ πππ‘+1]
ο· Free entry of firms: ππ‘ = 0
Since the free entry condition is imposed, the value of a job equation reduces to the relationship π½π‘ = π΄π‘β π€π‘+ Ξπ‘π½(1 β π)π½π‘+1. Therefore, it becomes evident that equation 5.2-8 is almost identical to equation 4.2-2. The only difference is the additional marginal cost term as well as the stochastic discount factor, which links the NK model to the DMP model, as described in section 5.2.
65 For ease of reference, equation 5.2-8 is quoted again:
π½π‘= πππ‘π€π΄π‘β π€π‘+ (1 β π)πΈπ‘π½ππ,π‘+1
ππ,π‘ π½π‘+1.
In the Nash product above, taking the first order condition with respect to the real wage π€π‘ yields
(1 β π)(β1)(π½π‘β ππ‘)βπ(π»π‘β πΌπ‘)π+ π(π½π‘β ππ‘)1βπ(π»π‘β πΌπ‘)πβ1= 0.
This expression can be simplified to
π(π½π‘β ππ‘) = (1 β π)(π»π‘β πΌπ‘). A.4.9.i Imposing the free entry condition and rearranging yields the form
π½π‘ = 1 β π
π (π»π‘β πΌπ‘), A.4.9.ii
which is identical to the Nash Sharing Search specification in Christiano et al. (2013). Taking the stochastic discount factor into account, inserting the value functions into A.4.9.i results in
π (πππ‘π€π΄π‘β π€π‘+ Ξπ‘π½ππ,π‘+1
ππ,π‘ [(1 β π)π½π‘+1])
= (1 β π) (π€π‘+ Ξπ‘π½ππ,π‘+1
ππ,π‘ [(1 β π)π»π‘+1+ ππΌπ‘+1] β π
β Ξπ‘π½ππ,π‘+1
ππ,π‘ [ππ‘π»π‘+1+ (1 β ππ‘)πΌπ‘+1]).
Multiplying out the terms that involve π€π‘ gives ππππ‘π€π΄π‘β ππ€π‘+ Ξ·Ξπ‘π½ππ,π‘+1
ππ,π‘ [(1 β π)π½π‘+1]
= (1 β π)π€π‘
+ (1 β π) {Ξπ‘π½ππ,π‘+1
ππ,π‘ [(1 β π)π»π‘+1+ ππΌπ‘+1] β π
β Ξπ‘π½ππ,π‘+1
ππ,π‘ [ππ‘π»π‘+1+ (1 β ππ‘)πΌπ‘+1]}.
Solving for π€π‘ yields
π€π‘= ππππ‘π€π΄π‘+ Ξ·Ξπ‘π½ππ,π‘+1
ππ,π‘ [(1 β π)π½π‘+1] + (1 β π)π
β (1 β π)Ξπ‘π½ππ,π‘+1
ππ,π‘ [(1 β π)π»π‘+1+ ππΌπ‘+1] + (1 β Ξ·)Ξπ‘π½ππ,π‘+1
ππ,π‘ [ππ‘πΏπ‘+1+ (1 β ππ‘)πΌπ‘+1].
66 Note that after imposing the free entry condition and forwarding by one period, equation A.4.9.i has the form
πΞπ‘π½π‘+1 = (1 β π)Ξπ‘(π»π‘+1β πΌπ‘+1). A.4.9.iii The wage equation can be simplified using A.4.9.iii, together with equation 5.2-7, reformulated
as π
Ξπ‘π½ππ,π‘+1
ππ,π‘ π½π‘+1= ππ‘. Using the fact that ππ‘
ππ‘= ππ‘, the equations can be combined to the form π ππ‘ = ππ‘Ξπ‘π½ππ,π‘+1
ππ,π‘ π½π‘+1. A.4.9.iv
Using equation A.4.9.iii, one can obtain πΞπ‘π½ππ,π‘+1
ππ,π‘ (1 β π)π½π‘+1 = (1 β π)Ξπ‘π½ππ,π‘+1
ππ,π‘ (1 β π)(π»π‘+1β πΌπ‘+1), which can be rewritten as
Ξ·Ξπ‘π½ππ,π‘+1
ππ,π‘ (1 β π)π½π‘+1= (1 β π)Ξπ‘π½ππ,π‘+1
ππ,π‘ ((1 β π)π»π‘+1β (1 β π)πΌπ‘+1), which in turn is identical to
Ξ·Ξπ‘π½ππ,π‘+1
ππ,π‘ (1 β π)π½π‘+1
= (1 β π)Ξπ‘π½ππ,π‘+1
ππ,π‘ ((1 β π)π»π‘+1+ ππΌπ‘+1) β (1 β π)Ξπ‘π½ππ,π‘+1 ππ,π‘ πΌπ‘+1. Likewise, (1 β π)Ξπ‘π½ππ,π‘+1
ππ,π‘ [ππ‘π»π‘+1+ (1 β ππ‘)πΌπ‘+1] can be written as (1 β π)Ξπ‘π½ππ,π‘+1
ππ,π‘ [ππ‘π»π‘+1β ππ‘πΌπ‘+1] + (1 β π)Ξπ‘π½ππ,π‘+1 ππ,π‘ πΌπ‘+1 . Thus, the wage equation has the form
π€π‘= ππππ‘π€π΄π‘+ (1 β π)Ξπ‘π½ππ,π‘+1
ππ,π‘ [(1 β π)π»π‘+1+ ππΌπ‘+1] β (1 β π)Ξπ‘π½ππ,π‘+1 ππ,π‘ πΌπ‘+1
β(1 β π)Ξπ‘π½ππ,π‘+1
ππ,π‘ [(1 β π)π»π‘+1+ ππΌπ‘+1]+(1 β π)π + (1 β Ξ·)Ξπ‘π½ππ,π‘+1
ππ,π‘ [ππ‘π»π‘+1β ππ‘πΌπ‘+1] + (1 β π)Ξπ‘π½ππ,π‘+1 ππ,π‘ πΌπ‘+1 The terms (1 β π)Ξπ‘π½ππ,π‘+1
ππ,π‘ πΌπ‘+1 as well as (1 β π)Ξπ‘π½ππ,π‘+1
ππ,π‘ [(1 β π)π»π‘+1+ ππΌπ‘+1] cancel.
Moreover, using A.4.9.iii and A.4.9.iv results in (1 β Ξ·)Ξπ‘π½ππ,π‘+1
ππ,π‘ [ππ‘π»π‘+1β ππ‘πΌπ‘+1] = ππ ππ‘.
67 Combining these considerations, the final solution to the Nash bargaining process is
π€π‘= π(πππ‘π€π΄π‘+ π ππ‘) + (1 β π)π, A.4.9.v which is identical to equation 5.3-1.
Taking logs results in
ln(π€π‘) = ln( π(πππ‘π€π΄π‘+ π ππ‘) + (1 β π)π), which after a linear first order Taylor expansion yields
ln(π) + 1
π(π€π‘β π€)
= ln(π(ππΆπ€ β π΄ + π π) + (1 β π)π) + 1
π(πππΆπ€ β π΄π΄Μπ‘+ π΄ β ππΆπ€ππΜπ‘π€+ π ππΜπ‘).
Canceling the steady-state terms and rearranging results in the final solution
ππ€Μπ‘= πππΆπ€ β π΄(π΄Μπ‘+ ππΜπ‘π€) + ππππΜπ‘. A.4.9 A.4.10. The Value of a Job Equation
For ease of reference, equation 5.2-8 is quoted again:
π½π‘ = πππ‘π€π΄π‘β π€π‘+ (1 β π)πΈπ‘π½ππ,π‘+1 ππ,π‘ π½π‘+1. Taking the natural logarithm results in
ln (π½π‘) = ln (πππ‘π€π΄π‘β π€π‘+ (1 β π)πΈπ‘π½ππ,π‘+1
ππ,π‘ π½π‘+1 ), which after a linear first order Taylor expansion yields
ln(π½) + π½Μπ‘ = ln (ππΆπ‘π€π΄ β π + (1 β π)πΈπ‘π½ππ πππ½) +1
π½(ππΆ β π΄(π΄Μπ‘+ ππΜπ‘π€) + ππ€Μπ‘+ (1 β π)πΈπ‘π½1
ππ½ππ(ππ,π‘+1β ππ ππ ) + (1 β π)πΈπ‘π½ππ
ππ2π½(ππ,π‘β ππ) + (1 β π)πΈπ‘π½ππ
πππ½ (π½π‘+1β π½ π½ )).
Cancelling steady-state terms and combining terms results in the final equation
π½π½Μπ‘ = ππΆ β π΄(π΄Μπ‘+ ππΜπ‘π€) + ππ€Μπ‘+ (1 β π)πΈπ‘π½π½(πΜπ,π‘+1β πΜπ + π½Μπ‘+1) A.4.10
68 A.4.11. The Job Creation Condition
Equation 5.2-7 states
π
π(ππ‘)= πΈπ‘π½ππ,π‘+1 ππ,π‘ π½π‘+1. Taking logs results in
ln(π ) β ln(π(ππ‘)) = ln(π½) + ln(ππ,π‘+1) β ln(ππ,π‘) + ln (π½π‘+1) Linearization yields
ln(π ) β [ln(π(π)) + 1
π(π)πβ²(π)π (ππ‘β π π )]
= ln(π½) + ln(ππ) + 1
ππ(ππ,π‘+1β ππ) β [ln(ππ) + 1
ππ(ππ,π‘β ππ)] + ln(π½) +1
π½(π½π‘+1β π½).
Dropping steady-state terms and using the elasticity of the job filling rate with respect to the labor market tightness, defined as β π =ππ(π)
ππ π
π(π), the function can be simplified to
ππΜπ‘ = πΜπ,π‘+1β πΜπ,π‘+ π½Μπ‘+1. A.4.11