I am dealing with a paper of Walsh & Ravenna.


I am kind of confused by the euqation (19) on page 33.

The market clearing condition goes as follows:

$$Y_{t} = C_{t} - w^{u}(1 - N_{t}) + \kappa\upsilon_{t}$$

Log linearization around the steady state yields

$$\hat{y}_{t} = \frac{\bar{C}}{\bar{Y}}\hat{c}_{t} - w^{u}\hat{n}_{t} + \left( \frac{\kappa\bar{\upsilon}}{\bar{Y}}\right)(\hat{\Theta}_{t} + \hat{u}_{t} ) \;\;\; \mathbf{(19)} $$

I don't know how this formula (19) is derived? Isn't there something missing? From my basic understanding of log linearization it ought to look like this:

$$\hat{y}_{t} = \frac{\bar{C}}{\bar{Y}}\hat{c}_{t} - \left( \mathbf{\frac{\bar{N}}{\bar{Y}}} \right) w^{u}\hat{n}_{t} + \left( \frac{\kappa\bar{\upsilon}}{\bar{Y}}\right)(\hat{\Theta}_{t} + \hat{u}_{t} )$$

with $$ Y_{t}\;\;...\;\; output $$ $$ C_{t}\;\;...\;\; consumption $$ $$ w^{u}\;\;...\;\; wage\;of\;unmatched\;workers $$ $$ 1-N_{t}\;\;...\;\; unmatched\;workers $$ $$ w^{u}(1-N_{t})\;\;...\;\;home\;production$$ $$ \kappa\;\;...\;\; cost\;of\;posting\;vacancy $$ $$ v_{t}\;\;...\;\; vacancies $$ $$ \hat{v}_{t} = (\hat{\Theta}_{t} + \hat{u}_{t} ) $$ $$ \omega = \frac{v_{t}}{u_{t}}\;\;...\;\;measure\;of\;labour\;market\;tightness$$ $$ \hat{\cdot}\;\;...\;\;log\;deviation\;from\;steady\;state\;value$$ $$ \bar{\cdot}\;\;...\;\;steady\;state\;value$$

Small letter with a hat: log deviation of a variable arount its steady state. Big letter with a bar: steady state value. K: cost of posting a job vancancy. w^u: "wage" of unemployed workers.

I've read the paper and the appendix aswell, read both the papers in the bibliography of this one aswell as later ones basing on this publication, but I could not find a helpful hint.

Is there any special relationship between N and Y in the steady state that explains why this whole term vanishs? Or do I have a wrong understanding of log-linearization?

I have to apologyze for my rusty English. I am already taking care of this problem. But for the above mentioned I would like to have your help. Does anyone has a decisive hint?

  • 1
    $\begingroup$ Could you please 1. also state the equation just above in your question (not numbered in the paper) and 2. a precise definition of the variables involved? $\endgroup$ May 21, 2016 at 17:23
  • $\begingroup$ @dugo: This way? $\endgroup$ May 22, 2016 at 11:59

1 Answer 1


We have:

$$y_t = c_t − w^u(1 − N_t) + \kappa v_t$$

We take the log of both sides:

$$\ln y_t = \ln \left[ c_t − w^u(1 − N_t) + \kappa v_t \right]$$

And then linearize around the steady states:

$$\begin{align} \ln \bar{y} + \frac{1}{\bar{y}}(y_t - \bar{y}) & = \ln \left[ \bar{c} - w^u(1-\bar{N}) + \kappa \bar{V} \right] + \frac{1}{\left[ \bar{c} - w^u(1-\bar{N}) + \kappa \bar{V} \right]}(c_t - \bar{c}) \\ & + \frac{w_u}{\left[ \bar{c} - w^u(1-\bar{N}) + \kappa \bar{V} \right]}(N_t - \bar{N}) \\ & + \frac{\kappa}{\left[ \bar{c} - w^u(1-\bar{N}) + \kappa \bar{V} \right]}(V_t - \bar{V}) \end{align} $$

Cancel $\ln \bar{y}$ and $\ln \left[ \bar{c} - w^u(1-\bar{N}) + \kappa \bar{V} \right]$

$$\begin{align} \frac{1}{\bar{y}}(y_t - \bar{y}) & = \frac{1}{\left[ \bar{c} - w^u(1-\bar{N}) + \kappa \bar{V} \right]}(c_t - \bar{c}) \\ & + \frac{w_u}{\left[ \bar{c} - w^u(1-\bar{N}) + \kappa \bar{V} \right]}(N_t - \bar{N}) \\ & + \frac{\kappa}{\left[ \bar{c} - w^u(1-\bar{N}) + \kappa \bar{V} \right]}(V_t - \bar{V}) \end{align} $$

Multiply the first term on the right hand side by $\frac{\bar{c}}{\bar{c}}$, and similarly for the other terms:

$$\begin{align} \frac{1}{\bar{y}}(y_t - \bar{y}) & = \frac{\bar{c}}{\left[ \bar{c} - w^u(1-\bar{N}) + \kappa \bar{V} \right]}\frac{(c_t - \bar{c})}{\bar{c}} \\ & + \frac{w_u \bar{N}}{\left[ \bar{c} - w^u(1-\bar{N}) + \kappa \bar{V} \right]}\frac{(N_t - \bar{N})}{\bar{N}} \\ & + \frac{\kappa \bar{V}}{\left[ \bar{c} - w^u(1-\bar{N}) + \kappa \bar{V} \right]}\frac{(V_t - \bar{V})}{\bar{V}} \end{align} $$

And we simplify, taking advantage of the fact that $\hat{v}_t = \hat{\theta}_t + \hat{u}_t$, to get what you have:

$$\hat{y}_t = \frac{\bar{C}}{\bar{Y}}\hat{c}_t + w^u \frac{\bar{N}}{\bar{Y}} \hat{n}_t + \left( \frac{K\bar{V}}{\bar{Y}} \right) (\hat{\theta}_t + \hat{u}_t) $$

I don't think this derivation is wrong. But you can see that $y_t$ and $N_t$ are directly proportional to each other. It's definitely plausible that $\frac{\bar{N}}{\bar{Y}} = 1$, but I've pored over the equations in the setup for 3 hours to no avail at showing that rigorously. It's probably something silly that I don't understand. I'll look back later and if nothing else, I'll turn this answer into a community wiki so anyone who has the right idea can edit it.

  • $\begingroup$ Thank you for your effort. I thought this through - again and again. Maybe I overread it or the authors do not point this one out, but could it be, that the technology has been left our of consideration? In this case $$y_{t}=n_{t} + a_{t} $$ would simplify to the identity $$y_{t}=n_{t}$$. This would lead to the log-linearized equation as mentioned in the paper, but I am not sure whether this simplification is feasible or not. $\endgroup$ May 22, 2016 at 12:15
  • $\begingroup$ There's an $a_t$ in this model?? Or do you mean $z_t$. I thought about using $y_t = n_t + z_t$, but it's used in the next step rather than here, and I couldn't juut say $z_t$ was zero. (Not actually sure what $z_t$ is either.) $\endgroup$
    – Kitsune Cavalry
    May 22, 2016 at 16:08
  • $\begingroup$ No, your right: Technology is not mentioned in the paper. I just thought that $y_{t} = n_{t} + a_{t}$ depicts one of the key elements of equilibirum and thus holds for every DSGE model... $Z_{t}$ is defined as an exogenous stationary productivity shock common to all firms. $\endgroup$ May 22, 2016 at 16:30
  • $\begingroup$ I forgot to mention that $\hat{z}_{t}$ is a stationary exogenous shock hitting all firms equally. $\endgroup$ Jun 8, 2016 at 17:25
  • $\begingroup$ $\bar{z}_t $ is set 0 in the steady state. $\endgroup$ Jun 11, 2016 at 11:23

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.