Macroeconomic Dynamics Of Labor Market Policies

>> John Taylor: We're delighted to have Eric Hurst, getting macroeconomic dynamics of labor market policies with Eleanor, who's here. Is Patrick here?

>> Erik Hurst: Patrick Kehoe.

>> John Taylor: Okay, anyway. Go for it.

>> Erik Hurst: The title of the paper is shifted a little, but the content of the paper has not, we're just trying to broaden it a little bit.

So let me tell you a little of what we're thinking about, and then we could kind of go into the detail. So my introduction is very short, that's three slides. So there's been a lot of discussion recently about market power on the part of firms. And then potentially, if such power exists, what are some policies that we might want to consider to potentially correct that market power?

That's part one. Part two, those policies you hear a lot about recently, about raising the minimum wage potentially nationally to some very large levels or expanding the scope of other policies like the EITC. Simultaneously, we know there's a large amount of heterogeneity across workers in the data. We know that from the wages they earn, and that even hurts within education groups, there's just a huge amount of variation of types.

Third thing we kind of know is if we're gonna wanna think about these policies that are gonna target certain types of workers with the minimum wage or the EITC that hits some parts of the distribution. We might then care about the extent to which the firms might be willing to substitute away from those workers if they become more expensive, so that elasticity of substitution across workers is gonna be a key component of the potential efficacy of some of these policies.

And when I say across workers, it could be between workers and machines or a whole bunch of other types of input substitutability. And so the last thing I'm gonna say is there's a tension in the labor literature to the extent to which firms are willing to substitute across input at different horizons.

So what do I mean by that? So many of you heard of papers by Larry Katz and Kevin Murphy. That's just one of the many papers that show that there is a very high elasticity of substitution over longer periods of time across workers from different education groups, that firms are willing to substitute low educated workers for high educated workers with some sort of elasticity.

And that elasticity is about 1.5. I cite that paper, but there's been a thousand papers since then that find estimates in that range. We also have within education groups some work by David Card with Tom Lemieux, and then again, a handful of other sense that show that within education groups, workers are even more substitutable.

So if John and I are roughly similar and John becomes more expensive, they might switch away from John towards me with an elasticity of some number that's pretty big, like four or five. So these medium to longer run elasticity is a substitution are rather large, at least from estimates of the labor literature.

But we also know the labor literature estimates very short run elasticities to some policies, like when the minimum wage changes. And so there is a bulk of the minimum wage literature finds that upon impact of a minimum wage change, and we'll talk about some of this today, that employment doesn't move too much.

Maybe it goes up a little, maybe it goes down a little. And there's a discussion of whether it's down a little or up a little. But most of the bulk of the literature is in relatively small, short run responses in employment to the minimum wage.

>> Steve: Between those two literatures, bigger than you suggest, Katz and Murphy is about annual fluctuations around a trend, right?

That's kind of I wouldn't call that long.

>> Erik Hurst: It may be five year averages. I mean, so Katz and Murphy, it's not annual, they're pooling over the five year kind of horizon.

>> Steve: I misremembered them.

>> Erik Hurst: Yeah, and Card-Lemieux is very much in that similar bank, yeah.

>> Speaker 5: You look at a huge or small changes, for small changes, you find very little big changes.

 

>> Erik Hurst: We're gonna talk about exactly that non-monotonicity today and some of this stuff. So let me kind of now get into what the paper is gonna be about. So what I'm gonna be doing in the paper is gonna be developing a framework to assess the distributional impact of various labor market policies.

And we're gonna focus most on the minimum wage today and some on the EITC. The framework is gonna include a reason for those policies to exist. We're gonna have some monopsony power on the part of firms in the labor market. We're gonna have rich worker heterogeneity, and we're gonna have some adjustment cost to firm input.

We're gonna model it in a specific way, which is gonna allow for short run elasticities to be small. But because of those adjustment costs over time, those elasticities could be longer. We're gonna parameterize the framework using kind of data from the US labor market. And then, as I said before, we're gonna then evaluate both qualitatively and somewhat quantitatively, the effects of various labor market policies, like a $15 minimum wage or an expansion of the EITC, John.

Yes,

>> John Cochrane: my two cent in the EITC is less about offsetting monopsony power, but offsetting the disincentives of social programs. If you're gonna take away a dollar of benefits for every dollar people earn, well, let's subsidize earning to try to get them back to work. Does that matter or anything you're doing?

 

>> Erik Hurst: I mean, yes and no. So there's John saying there's a lot of other policies out there that are interacting with these two policies. We're gonna allow these two policies to interact, but not some of the others.

>> John Cochrane: The employment margin involves.

>> Erik Hurst: As we go through, you're gonna see.

So to answer John's question, yes, we're gonna have all of those incentives in an optimizing way that are gonna be there. I'm not gonna horse race it against 72 policies, but I'm gonna only have these two policies. I'm not gonna have a means tested TANF program or something in the background as well that we could think about, but I really wanna highlight the dynamics.

And so this is what the paper is gonna be about. We're gonna think about how the adjustment dynamics are gonna be important for studying these distributional effects to different policies. And I'm gonna highlight, and it's kind of not unrelated to what John said, but I'll be more precise as we go through that the transition dynamics to different policies are gonna differ from each other.

And I'm gonna show you that the transition dynamics of the minimum wage are going to be different than the transition dynamics for the EITC for economically fundamental reasons. And we'll kind of highlight what those are. I'm gonna show that for large changes in the minimum wage, low wage workers are gonna be helped significantly in the short run.

But as firms start adjusting their input mix, they're gonna be hurt in the long run. And for a large expansion of the EITC, low wage workers are still gonna be helped significantly in the short run. But I'm gonna show you the benefits are going to be evenly larger in the long run.

So the minimum wage long run is gonna hurt some of the low skilled workers. The minimum wage of the EITC in the long run, it's gonna reinforce it. And part of the reason is gonna relate it gently to what John was saying, but we'll be more precise. It's gonna matter whether the firm or the government is paying the marginal cost of these transfers to the workers, whether the firms face the incentive to substitute away or not.

I'm then going to kind of talk about what we were doing, the model is gonna highlight a lot of non monotonicities in policy changes. Small changes and large changes are gonna be very different from each other. And because the small policy changes might just reduce the monopsony distortion, but the bigger you get, terms are then gonna start facing the incentive to substitute away.

And I'm gonna show you that kind of concretely as we go through our framework. And then I'm gonna have just a couple more bullet points, and then we're gonna get into the model. But a lot of what we're doing is trying to inform the large empirical literature that often find well identified policy changes that are small.

Small in the data and they look at short run responses and then they wanna make predictions about large changes and long run. And I'm gonna show we have to be cautious about making such predictions from the data that we tend to identify and that both on the size of the policy, small versus large.

And the timing of the policy, short run versus long run. And I think a lot of what we're gonna do is kind of highlight the key theoretical forces through the lens of the model that are gonna determine when the short run responses of small policies is informative about long or large, it depends upon some parameters.

And so we're gonna try to be very specific about those economic mechanisms as we go through.

>> Speaker 7: Elasticities are not constant.

>> Erik Hurst: Elasticities are not constant either time or size.

>> Speaker 7: For the constant elasticity, it remains the case that a small input gives us a small output. So you're not really not giving sufficient credit to the importance of this finding?

 

>> Erik Hurst: Well, I want

>> Speaker 7: things like this is a big issue in trade theory, I should say elasticity is remarkably higher.

>> Erik Hurst: That's what I'm gonna try to say. Maybe I didn't say this close here, but the whole talk is gonna be about the timing of the elasticity, when the elasticity is gonna depend upon the time horizon.

Which is gonna be how much time it takes to adjust inputs. That's one of the things in that last bullet point, but also the extent of monopsony power, the degree of input substitutability. All of those are going to determine what makes the elasticity state, entire policy size and time specific, yeah.

 

>> Speaker 8: Measure monopsony power.

>> Erik Hurst: I'm going to take from the literature a range of estimates, and I'm gonna show you how, from the literature in these range of estimates, how the results change across these range from the ditch.

>> Sebastian: There's another dimension to this elasticity question, which is the horizon as a shock.

So here it looks like you're thinking about permanent shocks, so say I'll permanently change policy.

>> Erik Hurst: I'm going to show you permanent versus transitory as we go through, but all my shocks are gonna be MIT shocks. Can I say that in this room, which are basically gonna be unexpected, and then they're gonna be permanent or temporary, but they're gonna be in that dimension as temporary.

 

>> John Taylor: Chair of the meeting, I'm gonna rule out a questions about what are you going to do

>> Speaker 10: exactly. I'm off with that.

>> Erik Hurst: So now I'm gonna get into the model,

>> Speaker 11: I'm gonna start delaying too much.

>> Erik Hurst: Now I'm gonna get into the model now, that's kind of the background, so what I'm gonna do is going to set up the framework.

I really wanna think about the qualitativeness, of the mechanisms that were going to be key to the results I'm gonna show. And then I'm gonna talk briefly about the quantification, how I wanna take some stuff from the literature as kind of targets and then do some robustness around there.

And then I'm just gonna show you some results. And so I'm hoping that we could get to the results maybe in about 15 minutes, and then the whole rest of the talk is just gonna be about kind of the results in the economic mechanisms. But I do wanna get a little bit of the framework on the table.

So let's talk. So I'm gonna do three things, I'm gonna describe the heterogeneity on the worker side. I'm going to then describe the production structure both in the long run, and then how we get differences between the short and long run. And then I'm gonna talk about the monopsony part of our model, how we're going to introduce monopsony into this model, so we're gonna be in those three parts.

So let's talk about the worker heterogeneity first, this is the simplest. Households are going to be defined by the education group to which they are exogenously born, which could be high education, which think about as a bachelor's degree. Or more when I take it to the data, or low education, which is gonna be less than a bachelor's degree.

And then they're also conditional on being born into an education group, they're gonna have some latent ability z. And some of us, John and I, are both low educated, he might be more productive than me, even though he's going to have a higher z in terms of our notation than I do.

So an individual in our model is gonna be their late in productivity in the education group, which they're going for. The things I'm really gonna focus on today, you can just think about being a low education group, but we are allowing for some substitutability in the data between the two.

There's gonna be, each one of these individual eyes are gonna belong to a large family, we could talk about that later on. And then there's gonna be lots of large families we don't want the workers to have any of the power in this

>> Speaker 12: difference. If you don't explicitly have tangible capital as a potential substitute.

 

>> Erik Hurst: I'm gonna allow for it and I'm quantitatively not gonna be important, and I'll show you the production structure when we come through, and there's gonna be capital floating around too. In Mu, just so you know, that's gonna be something that's gonna tell us how many of each z type there is, And we're just gonna use Fitbat to match the wage distribution in the data

>> Erik Hurst: production structure.

I'm gonna break up the production structure in the long run, and then we'll talk about the short run production structure after that. So firms are gonna use heterogeneous labor, okay? By education group and z and capital in the production of a homogeneous output good. There's gonna be three margins of substitution, the first one I'm gonna talk about is by far the most quantitatively important in our analysis, I'm gonna spend the most time on that.

But I'm also gonna allow in the background for these other two margins of substitution. There's gonna be CES, aggregator of labor within each education group, so if John and I are low educated, there's gonna be some elasticity of substitution between us in the low educated group. And I'm gonna talk a lot about that on the next slide.

And then we're going to give me one second, Sebastian, and then I'm also gonna allow for elasticities of substitution across education groups. And I'm also gonna allow for elasticity of substitution between capital and a broad labor aggregate. And I'll show you all of those equations in general, but you could think about the three margins of substitution we're going to allow for is labor-labor substitutability within an education group.

Labor-labor substitutability across education groups in the substitutability between labor and capital, Sebastian.

>> Sebastian: So in the last slide, you mentioned that you use the Zs to match the distribution of income.

>> Erik Hurst: Yes.

>> Sebastian: But I wonder if this matters relative to an alternative where it reflects different monopsony power?

Some workers have lower bargaining power, that's why they get lower wages.

>> Erik Hurst: Yes,

>> Sebastian: Yeah, so.

>> Erik Hurst: I'm gonna tell you, I'm gonna make assumptions on what my monopsony power is gonna be when you see that coming up. My monopsony power is gonna be constant across workers, and all the wage variation is gonna then come from z.

I'm not gonna have heterogeneity, cuz the literature, as far as I could tell right now, doesn't have any sort of empirical estimates to say whether the low part of the distribution has more monopsony power or less monopsony power. In some sense, you might think they might be captured more by firms, but in the other sense, they switch jobs quite a lot, and I think from what the literature has shown so far, we have not found.

But if we ever found an estimate, we could embed that into our model, but right now, we don't have that, so. But once I show you the monoxony part, we could come back to this, but Sebastian's already foreshadowed. One assumption we're gonna have, we're gonna have a constant markdown of wages across workers of z types.

 

>> John Cochrane: My big understanding facts is there's a lot of wage variation just on where you are, which company you work for. If you're a janitor at Apple, you get paid a lot more than if you're a janitor somewhere else, that's gonna be greater skill in your model or.

 

>> Erik Hurst: Yeah. So, in the background, so let me start with the data first, and then we'll come back to the model. So, in the data, so the people who try to decompose how much wage variation is due to worker effects versus firm effects plus the sorting, the vast majority is worker effects.

And so almost like a ten to one ratio. Now, there's a sort of part in between, and then is that a worker effect or a firm effect? But in terms, I would say the consensus from this, David Cardware, Pat Clyne, John, Mark Robin have all tried to do these worker firm fixed effects.

So think in the back of your mind, about 60% of the wage variation. Is pure worker effect. About 5% of the variation is firm effect, maybe 5% to 10%. And then the rest of the assortative matches.

>> John Taylor: Wrong.

>> Erik Hurst: Yeah, you're just wrong. So.

>> Elena Pastorino: 70% according to recent.

 

>> Erik Hurst: It could be even higher. Those are the lower estimates. So let me talk to you now about this first labor substitutability. This is literally the model that card and Lemieux used. Two parts of this is gonna be equation one, and equation two are literally card and Lemieux.

There's just gonna be some elasticity of substitution fee that's gonna govern the substitutability of workers within an education group. So John's making $7 an hour. I'm making $10 an hour. If John's minimum wage goes up to eight, the workers are gonna substitute. Potential firms are gonna substitute towards me with some elasticity in the long run of feet.

 

>> Pete: So why do you even have the education groups?

>> Erik Hurst: Yeah.

>> Pete: Just so you can say, hey, we can fit their 1.5 for their aggregation. But more disaggregated, there's something you could have done quantitatively would have been similar.

>> Erik Hurst: Very similar again. So Pete's question is, why do we even have this other nest in equation two of allowing for two education groups?

So now I'm gonna allow for elasticity of substitution alpha between these two education groups. That is Katz and Murphy, and that is literally card and Lemieux when they estimated their model was equation 1 and 2. And so that allows us to map a little bit either. But it's adding nothing quantitatively to us to have this second education group, because when we talk about the minimum wage in the EITC, it is 99% of the workers treated are in the low education group.

And so in our model, when I even show you results later on, I'm never even going to show you the high skilled stuff because nothing interesting is going on there. The high education group, Chris.

>> Chris: This might be related. Is the fee different for the high and the low education group?

And do we-

>> Erik Hurst: It was not. So Cardinal Mew actually estimated the fee to be exactly the same between the two groups. Again, they have a range, but it's in the four to six range for both of them.

>> Chris: Thank you.

>> Erik Hurst: Okay, so equation 2 is gonna be exactly Pete's question.

We're gonna allow. It's not gonna be quantitatively important because nothing's really happening to the high educated group in this model. And then we're also gonna allow for some substitutability between broad labor and capital with some elasticity of substitution here. Rho, we're gonna assume Cobb doubles as our baseline.

Again, we're not getting a lot of traction empirically. When we start moving around with this, we do a lot of robustness. So that'll be in the background as well. But, almost all of the variation that we're gonna talk about is gonna be this fee, the labor, labor substitutability within an education.

But we allow these other margins as well.

>> Speaker 15: Why do I think about corve? You didn't go the corve route of having capital be complementary with skilled labor and substitutable unskilled.

>> Erik Hurst: The first version of the paper was all corve.

>> Speaker 15: Okay.

>> Erik Hurst: Quantitatively it just wasn't adding any extra kick in for the kind of the policies that we're doing.

So we stripped it down and simplified it. But we did have a core version of this. But quantity, even with their estimates, we just weren't getting an extra quantitative bite. It's really when a $7 worker has to be paid 14, they're willing to switch within there, but it's not so much with the capital.

Now, why do we have capital at all, you might ask? We might have, oops, I'm going the wrong way. Because we do want some adjustment costs in the background. And so now we are going to have our floor of adjustments. What's gonna happen? That's the long run production structure.

What's happened in the short run? There's a putty clay adjustment on the firm. So firms set up their production facility. There's a couple of cash registers. There's in the restaurant, there's a grille. There's some friars. And it just takes time to alter their mix of machine and workers such that the two match each other.

So putty clay, what that means in our setup is that capital, in the long run is going to be just what I showed you, CES across all those components. But in the short run, it's essentially Leontief. It is fixed. Once you install your machines, it's matched to a worker.

And so if you want to adjust your worker types, that's fine, but you're gonna have to adjust your machines, and that happens relatively slowly over time.

>> Steve: As I understand the model, you haven't introduced a distinction between adjustments to a local, increase the minimum wage and, say, a national one, because you might think for McDonald's, they're gonna not care if one city raises its minimum wage.

 

>> Erik Hurst: Exactly. Yeah, so there's no regional. So Steve's 100% right. We thought about that, but we distracted from all of that for exactly your reason.

>> Steve: Well, I guess I can get to the reason. It seems to me it's probably important, but it's hard to get evidence on it, so it might be hard to monitor.

 

>> Erik Hurst: Let me just answer, Patrick, what I'm gonna think the question Steve should be asking, and it might be related to the one he's actually asking, which is, we are thinking about national minimum wage changes. There's no local anything in here. Now, part of the reason is a lot of the work on the minimum wage comes from the local components.

Yeah, almost all of it exactly in there. Think about you're in Idaho and you're at a Starbucks, and Idaho raises the minimum wage. Starbucks isn't gonna retool all of their production functions. So that's gonna bias again to smaller empirical responses in the short run. And so I'm gonna match empirical responses in the short run.

For us, it's gonna come from a lot of the putty clay structure. But Steve's right. If we're going to the data, a lot of that data might not be one for one matching with a model and the aggregate.

>> Speaker 16: Not for Starbucks, but other things. One of the responses is gonna be to move production across state lines.

 

>> Erik Hurst: So we thought about a version, not this paper, but we're thinking about what happens when the national minimum wage to the south, where you put a $15 minimum wage and there's a lot of production, they're just gonna move all production out of the south, and that's gonna be devastating in terms of some of

 

>> Speaker 16: the local minimum wage. So you're talking about long and short run elasticities. And this has got to, since the elasticities in the previous equations were all had nothing to do with time.

>> Erik Hurst: Yes.

>> Speaker 16: This is where the long and short run-

>> Erik Hurst: This is all the distinction and the parameter that governs that is basically the depreciation rate of your capital.

So as your capital starts depreciating away, you're gonna build in new capital, and as you bring in the new capital, it's gonna match whatever the mix of workers that you want. And so it's really the depreciation rate in the way. Now, on the good side, one parameter does all of our transition dynamics, which is tractable.

It's easy to discipline, in terms of the real world you could imagine the adjustment process is probably much more complex than a one parameter model. But this is gonna give you a flavor of how we could get the short run in the long run, in a very.

>> Erik Hurst: You could shut it down.

So they always have the option to shut it down.

>> John Taylor: Sorry, we have questions. Yeah.

>> Speaker 17: The minimum wage started out about 0.6% of manufacturing, it's been discounted. How do you adjust for these parameters that were fixed initially and were discounted? Are you gonna bring it back to the old established minimum wage?

 

>> Erik Hurst: Once I get to the quantification, well, let's talk a little bit about it now. Let's talk about the general model without a minimum wage right now, just to try to see the economic forces, and then we'll come back when we talk about quantification. The model is going to be real everywhere.

And so this kind of goes back to some of the questions we got earlier. So we're gonna have a real shock in the minimum wage when it goes up, because there's only real. But then I'm gonna allow for a temporary shock in the minimum wage, which happens if the nominal minimum wage goes up and then it erodes over time because of inflation.

I'm gonna show you all of those pictures, but I wanna get the model out there before we get to the policy evaluation stuff. Yeah.

>> Sebastian: You could do it multiple ways because the basic model is CES in the capital and the bundle and then the Leonchief. You could be capital.

Every single one.

>> Erik Hurst: We have two m labor factors, two m of one group, m of the other group, and one kind of capital. It's LTF across the two m plus. Plus one factor, joined for every single wire. Exactly, so capital.

>> Sebastian: Machine for one worker.

>> Erik Hurst: Exactly, yeah, think about a grill, is good for some types of workers, and then, computer is good for another type within the thing.

So that's exactly what it is.

>> Sebastian: Something a little bit, more within four.

>> Erik Hurst: You could do with. A thousand things differently. But that's not what we did.

>> Elena Pastorino: We didn't want additional parameters, Adrian. So we wanted this notion.

>> Sebastian: Still be one parameter.

>> Erik Hurst: Let me just kinda give a big picture to both of their, thank you both.

What it is, is what we really need is this putty clay does, is it gives us adjustment cost of both capital and labor simultaneously in a kind of joint way. I could put in, adjustment cost of capital in a complex way. I could put in adjustment cost of labor in a complex way.

I don't know how I would discipline them. This is a way, but the key thing is we need adjustment costs, to both capital and labor. And this is kind of a way, to get adjustment costs to both capital and labor and impersonal.

>> Speaker 18: So Eric has an example in the paper where if you have labor aggregates, but you have within the labor aggregate, it's perfectly fine and it's only the average.

You adjust like that, and it completely inconsistent with the data, so that's what it is really. Just shift. I'll go even further.

>> Erik Hurst: I think you've come to the conclusion that it's adjustment cost on labor, that is really the important thing that we kinda knew in a way that's kind of related the structure of the firm.

And that's kind of the key thing. And this kinda gives us that in a parse money switch. So let me give you the last part of the model on the table. I'm gonna then apologize for a couple of things that aren't in the model, and then we'll get to the results.

And so here's our household preferences. And this is where the monopsony is gonna come in. So you're gonna give me a couple of seconds to see how I'm gonna embed monopsony into our model. So, we're gonna have, individuals who are gonna have, for each of those families eyes, they're gonna have preferences over their consumption.

They're gonna have some disutility overworking. And then there's gonna be some search back in the background. I'll talk about that in a second. People are gonna have to find some jobs. That's also a way to close the model when there's rationing with the minimum wage. Ignore that part for now.

So, what is this disutility of what?

>> Speaker 19: When I see GHH I think this, is the last scoundrel of somebody who doesn't want any income effects on labor supply.

>> Erik Hurst: Or laziness, you got to give me laziness. Just tractability and laziness is another thing. So we have no income effects on labor supply.

 

>> Speaker 19: And that is important or it's just tractable.

>> Erik Hurst: Tractable is what we're doing. I don't know. On our list of things to think about is relaxing this part of the component it just makes attractive.

>> Erik Hurst: I mean, you have a large family that's relatively insured. So the question is, how we're gonna discipline the income effects on labor supply.

I'm gonna tell you, I could play around with our labor supply parameter. It is nowhere near as important as the firm substitutability. Once I get to the quantification you'll see where some of this, where some of the.

>> Speaker 20: Large family, is not a fact.

>> Erik Hurst: Not a fact, it's attractable way.

 

>> Speaker 20: Yeah, exactly.

>> Erik Hurst: It's a tractable way, exactly.

>> Sebastian: DHH also gives you this complementarity between labor and consumption. So, that could matter for the dynamics, okay?

>> Erik Hurst: So the question is.

>> Sebastian: Not just the discussioning of the income effect.

>> Erik Hurst: But I think it's gonna matter in a quantitative sense and not a qualitative.

And we get to the thing. I'm gonna show you some numbers, but I really wanna see the qualitative patterns, that really, I think are gonna help us, interpret the data from the empiricists that they give us. Now if I wanted to do a full quantitative and take my number of what is the optimal minimum wage, or the optimal EITC, thinking about these kinda forces is gonna be important.

That's not what my game I wanna play. I wanna play a game of what is the mechanisms that we need to think about, in a framework sense to kinda help interpret, a lot of the empirical literature that is coming to us. And let me show you a little bit.

And then we could come back at the end and say, here's all these things we might wanna do. And then we could talk about whether that's gonna change, the takeaways that I want to do. It certainly will matter quantitatively. I just don't think qualitatively for the points I'm trying to make, it's gonna be there.

Let me tell you now, I think this is something substantive.

>> Erik Hurst: A lot more bullets coming.

>> Speaker 7: Like so many labor economists. You really don't take unemployment seriously?

>> Erik Hurst: No, I do.

>> Speaker 7: Unemployment is the domain of.

>> Erik Hurst: You're saying things that are not true, cuz right now the whole model has a search component in the background.

I'm not gonna highlight this for again the points I'm making, but we have an unemployment versus out of labor force margin. That's where the search is coming in, in the background. I'm not gonna talk about that today, cuz it's irrelevant to the points I want to make.

>> Speaker 7: Your idea is there's some determination of employment, and then there's a residual, which someone else can take care of.

 

>> Erik Hurst: We're taking care of it. I'm just not gonna talk to you about it today. What makes you say that.

>> Speaker 7: No, I'm just

>> Erik Hurst: search, work and stay at home. Okay, everybody else close your eyes for 1 second I hate doing this. I'm gonna go and in a little while there's gonna be a directed search framework in the background.

We're gonna have a whole bunch of things. Okay, everybody else open your eyes, we're back to this part. So let me talk to you now, about the monopsony part of the model, cuz that's the last, again, important component, in trying to think about the forces that I wanna highlight today.

So, I'm gonna have, preferences embedded in the model, that are gonna generate monopsony power. Let me give you the intuition and the metaphor of what we're doing, and I'm gonna tell you why this functional form, is a good metaphor for what we're trying to do. So the background concept, think about there is workers are gonna have idiosyncratic preferences, for the different types of workplaces that are out there.

Each of these workplaces, are ex-ante homogeneous, except they have the same production functions everything, except one's green and one's blue. John might like working at green firms, not like working at blue firms. Maybe the green firm's close to his house, and the blue firm's farther away. I might like working at blue firms and not green firms.

We have these preferences, for these jobs. As the green firm starts to expand, they're gonna basically hire John. But in order for them to hire me, in they got to compensate me for the disutility of coming into that firm. And so that means as the more the firm hires, the marginal cost of hiring workers is gonna increase, as they're bringing in people with less and less good fit for those jobs.

So that's kinda what we're talking about.

>> Speaker 21: And there's horizontal equity and pay?

>> Erik Hurst: And so firms, they can't price discriminate. So when they see a type, they can only see your highness, not your preference, who draws in the background. And so, they have to pay everybody the same.

And so, this preference structure I'm gonna do is exactly analogous, to Dick Stiglitz in the goods market. In terms of a love of variety sense, which generates monopolistic competition in the product market. I'm gonna use the same thing on the worker side, in terms of preferences for jobs.

That's gonna generate, monopsony power, among the firms. And so, this parameter omega, is then going to generate, from the workers perspective how substitutable those jobs are, with each other. And if that parameter goes to infinity, then there's no monopsony power. All the jobs are roughly the same. If, Omega is gonna be.

 

>> Steve: This part of the model, I have a hard time wrapping my head around for one, the first thing that comes to my mind, is in city. If you live in a city, you got all kinds of jobs to choose from. So, this effect seems like it's a lot less important than if you live in a rural area.

Where if you're 50 miles from the blue company and you, really like the blue company, you gotta work for the green company cuz you're so far away.

>> Erik Hurst: Steve is saying there's heterogeneity out there in dimensions, which the model does not have. We have a national market, and we are going to use national aggregates, from people who try to estimate the extent of wage markdowns, as our discipline.

And then, here's the key thing, Lucy, if you don't want any monopsony power, if you believe there's no monopsony power in the rural or high in the rural, we could just change our. Our parameter, and then I'm gonna show you a whole vector of parameters and then we can see how the model evolves.

Again, it's a framework paper allowing us to do these types of policies and if you want a different parameter, we could just put a different parameter.

>> Steve: Yeah, I don't know how are you gonna make strong welfare claims?

>> Erik Hurst: I'm going to show you qualitative welfare claims of how things change over different horizons.

 

>> Speaker 16: Just back to the equation a second. So every worker works a little bit of every job, is that what in J is?

>> Erik Hurst: Yes, every worker is a large family and you're aggregating over your family. So this is in a literal sense, there's lots of eyes and we all care, we're all pooling our I together and you go to the green firm, I go to the blue firm so.

 

>> Speaker 16: I as a family, J is a worker within the family,

>> Erik Hurst: the. I is a work family and a worker in a family and J is a firm. So J is green firm versus blue firms.

>> Speaker 16: IGT is what fraction of your family works at firm.

>> Erik Hurst: The green firms versus the blue firms.

 

>> Speaker 23: The families measure one cut up into people measure circuits.

>> Erik Hurst: And so the literature, again, in the aggregate data, there's a whole bunch of burgeoning literature on wage markdowns with a variety of different estimating strategies that have come out in the last four or five years. Magna Monstag has some stuff going to this Steve Student Claudia has some estimates on this.

Some of our, my former, our former, our current and former colleagues, David Berger and Simon Monge and Kyle Hercule have some estimates. They're all in this range of 0.65 to 0.85 I'm gonna take 0.75 as kind of my baseline, then I'm going to show you a whole bunch of other things.

 

>> John Cochrane: But back to the previous point. At least some of that work has got lots of spatial heterogeneity in the markdowns they estimate. Right?

>> Elena Pastorino: We think of it also in terms of other dimensions, other dimensions of resultant differentiation like amenities and match quality as reflected in non peculiar value of work.

 

>> John Cochrane: Yeah, I'm not adverse to the basic idea, it's just that the scope for those to matter seems to vary a lot in the spatial cross section. Maybe I'm thinking about it wrong

>> Erik Hurst: so let me show you what I'm gonna do, and then at the end let's come back and say, I would like more of this or less of that.

Once I tell you kind of what I'm doing with the model that I haven't I've kind of given you a roadmap. You haven't kind of seen the thing, but at the end of the day you might say, I think this whole thing is flawed because you don't have this kind of regional variation.

And I'll say, okay, then maybe we need to think about it, but I think we're gonna get pretty far with what I am going to do that then we could talk about how we're gonna on the edges.

>> Speaker 24: You are gonna do is a 25% markdown.

>> Erik Hurst: And one of mine, and then I'm gonna show you a 10% and I'm gonna show you a 30%.

Our baseline is gonna be 25%.

>> Speaker 24: We said 75, but we didn't say whether.

>> Erik Hurst: Yeah so I should be clear yeah. So it's basically a 25% markdown or 15% markdown or 35% markdown. So but our baseline is

>> Speaker 24: I get to choose

>> Erik Hurst: 25% below marginal product is our baseline in terms of our calibration.

 

>> Speaker 24: Baseline.

>> Erik Hurst: I'm gonna show you a lot of different ranges.

>> Speaker 24: The very next thing you'll do is 15.

>> Erik Hurst: Yes I'll show you some pictures with 35 yes, I'll show you those pictures because really I wanna show the framework in how these parameters move things around.

And so I'm gonna show you, I'm gonna move lots of things around because I really wanna get the economic mechanism of where these trade offs, why these dynamics could be big or small and I'm gonna show you that.

>> Speaker 25: Toyed with the idea of that power being a function of your time horizon right I mean.

 

>> Erik Hurst: Yes I have not played with it that's a good question. I mean, there's no timing dimension in this part of the model all the time is on the firm side, I don't have any time on the preference side. And that could be interesting to think about I just don't have any way to think about disciplining that yet, but if I did, that would be, that's a limitation because all of my time is on the phone, not on the house okay.

 

>> Bob: Are you specifically charged of this within the team?

>> Erik Hurst: We I should be saying Bob has made it clear.

>> Bob: No, no, I just wanted to.

>> Erik Hurst: It is, we are a team it's not just an I. But Bob was making fun of me because I was ignoring my co authority.

 

>> Speaker 27: You should sit a little closer to.

>> Erik Hurst: So it is a week, It is a week. Okay so let me just now say two more I was one more thing and then I wanna get to some results, but just to make sure. So what is this gonna do?

So the monopsony power, as we start going forward, there's going to be a directed search part of the model that I'm not even gonna talk about. But the way I want to think about it is we've already had, it's in the paper, if you could read the paper if you want more details, I should say that.

 

>> Speaker 28: So what makes you choose?

>> Speaker 29: I mean, that's another one of these theoretical constructions which is very, very distant from findings. Directed search is there something about directed search as opposed to random search that.

>> Erik Hurst: None of it matters quantitatively for what we're gonna talk about today and so why do we do it?

I want something, again, for those who are interested, there's a search part of the model in the background. It's just a way to close the model when there's rationing in minimum wage, who gets allocated to which jobs?

>> Speaker 7: That's a labor market labor economist talking.

>> Erik Hurst: But I'm part of that but there's, quantitatively it's not gonna be so there's gonna be lots of things around there.

I've chosen a talk that's gonna show you the ones that at least quantitatively are gonna be the more important of the mechanisms I'm gonna show you today. And then there's other things, some of which I've included in the model that we could talk about after and some of the things I didn't, which I'll apologize about in a few slides.

And so I think, let me get a little further and I think some of this feedback will be even better as we get a little further. But the key thing that I wanna show why is the monopsony gonna be important in this model, is for the exact reason that Steve kind of highlighted already, is that when you wanna hire more workers, you got to pay all your old workers of the same type, the same wage.

And so that means you're gonna have these upward sloping marginal cost curves that are gonna be rising even faster than the average cost curve because you got to pay all these workers.

>> Speaker 30: We can all see worker types.

>> Erik Hurst: And we can see a worker I observable to everybody it's stamped on your head.

And so with this monopsony power, just in a classic undergraduate text with sense, that's gonna cause firms to pay lower wages relative to marginal product, and hire less workers than they would in a competitive model. And so let me just kind of give you the textbookness of this.

In a stripped down model, this isn't exactly our model because there's no worker heterogeneity, there's no search or anything, but this is what you would get in an undergraduate textbook, citria, with an upward sloping average cost curve for labor from the monopsony power. The competitive equilibrium is gonna be those WC and WN the competitive equilibrium, the CQ competitive, the firms are gonna price on their marginal cost curve, which is gonna be lower wages and higher lower employment at the same time.

Now, why is that important? Now we're getting to the nonlinear part of the model, if I start doing positive, like minimum wages, in the amount of forcing you to pay a little bit more for a little while, you're gonna get higher wages and higher employment. But if the minimum wage keeps going up, if you start paying, forcing people to pay higher than WC, firms are gonna do what any neoclassical firm is gonna do, they're just gonna hire less of those workers.

So you're gonna have this very non monotonic response to policy because of the monopsony power, and so that's gonna be in the background of some, a lot of what we're gonna talk about today. This non monetization is gonna be important at different horizons, even okay. Talk about that here's my apology slide there's a lot of other things.

Thanks, the regional one I should have added to this. This is all national coming through. I'm not gonna have endogenous firm entry. In this model, we could talk about how that might change as well, even though the policies are gonna start playing around with firm profits. And I'll show you how that's gonna happen.

That might be something we might want to consider. Workers are gonna be be born with their Z's, and their skills, their education groups. I'm not gonna allow for any upgrading, maybe that's important. The model is going to be completely real, so there's no real price response. But importantly there's not even heterogeneous response.

Different groups might consume different goods and different goods might be moved bound by the minimum wage. I have none of that in there. We could talk about qualitatively how that might change our results a little bit later on, but I just wanted to get it up. These are all questions that usually come up and I'm gonna say I have nothing to say about those today except some speculations I think.

 

>> John Taylor: There's a question, you're ready for this question.

>> Speaker 7: The monopsony power is the employer generates a lot of profit. What do you do to this? Profit doesn't exist in our normal thinking. How do you dissipate the profit?

>> Erik Hurst: So the profits go to the workers based upon their share of labor income.

And so, low wage workers are not getting any of those profits because they get very little of the labor income. The high wage workers are getting most of the profit because they make most of the income. So we keep the profit share in the model similar to the income share out of the total income wage bills.

 

>> Speaker 7: What do you mean the go to?

>> Erik Hurst: So the workers own the firms, the firms are earning profits. Those profits then dissipated back to the workers and their budget constraint. And the share they get rebated back is going to be to the share of their profits for laboring.

 

>> Speaker 31: I mean, all this machinery, this is what drives any benefit from any minimum wage, right? If it wasn't the case that there was monopsony power and it wasn't the case that the rents accrued from that didn't got unevenly distributed, then minimum wages would always be bad.

>> Erik Hurst: We're trying to give minimum wage a chance to do something and we're gonna try to learn, and then how much it does.

We could take the monopsony power to zero. I mean, make it perfectly competitive in the labor market, and that will always be bad. Dan's already forecast, you could already see, and I'm gonna show you some of that as we do some of the markdowns go close to zero.

 

>> Speaker 31: So the way to look at these- Welfare results are assuming that there is monopsony power, then-

>> Erik Hurst: Everything is dependent upon the parameters of the model, and we can change those parameters in any way we want.

>> Speaker 31: I understand, okay.

>> Erik Hurst: Okay, so, now, that's kind of the framework.

So, let me just summarize the key parts of this framework that we're gonna go going forward, and how I'm gonna discipline those key parts. And then there's a whole bunch of other stuff in there. But I think those are not going to be where the big value added is, or at least quantitatively, those other mechanisms for what I'm gonna show you.

So there's going to be the monopsony power, we can make that anything we want. We're gonna target the wage markdowns in the literature, but then we're also going as a baseline of a 25% markdown. People are paid 25 less than their marginal product as our baseline. We could take that people are paid their marginal product.

We could then take it people are paid 35% less than their marginal product. So we could play with that parameter. And I'm gonna show you different results. But in our baseline, I'm just gonna target the average of this range, the labor-labor substitutability. Just give me one second, Chris.

I'm just gonna take right from the lower bound of card and Lemieux, and I'm gonna show you lots of robustness. What happens if that number is higher or lower? We're gonna match the types of the muse, as we talked about earlier on, just to match the wage distribution in the data.

And then the depreciation rate, which determines the one parameter that determines all the timing. We're just gonna set that to 15% in our background so we could play with any of these. But these are the four things that are gonna determine the quantitative patterns I'm gonna show you coming upwards.

First Chris, and then I'll come in.

>> Chris: Clarification, when you change the monopsony power parameter, do you re estimate?

>> Erik Hurst: Yes, everything estimated joint. I should have been saying, these are kind of targeted stuff, but everything's estimated jointly, Jack, and then I'll come back to you.

>> Jack: That seems high to me, so is it that you think structures are probably not putty clay and equipment is, and equipment?

 

>> Erik Hurst: We've gone back between 10% and 15% all the time, one is in the appendix, one is in the paper. We're doing, again, trying to, because people have pushed back, maybe the structures isn't going to be the key component relative to the adjustment relative. We went a little bit lower, but we want to show robustness to all of these.

Again, it's a framework setting and we're using this as a baseline. But the other one, the 10% is how we had everything until about two weeks ago. And then we got comments the other way, maybe we should exclude structure part and go to the 15%, yep.

>> Speaker 33: You're assuming employee owned firms for the model?

 

>> Erik Hurst: Yeah, so the profits then go back to-

>> Speaker 33: But obviously they're distributional effects if it's not employee owned firms.

>> Erik Hurst: Yeah-

>> Speaker 33: That's the big argument in the monopsony area.

>> Erik Hurst: Yeah, so the key part is, at least for these policies targeting the low end of the distribution, they essentially, in our model, essentially own none of the profits anyways, because it's just epsilon.

So whether we make that zero or epsilon, it doesn't change any of the results in the paper. Now, if all the poor people owned all the firms, that would be something, but that's counterintuitive in the data, AJ.

>> AJ: You haven't talked much about the labor supply decision, but for the ITC, it's all about kind of marginal tax rates.

 

>> Erik Hurst: That's coming in 20 minutes. Or maybe we won't get to it. We have a fridge one in the background, we show you what happens when the freesia two or fish 0.5 and those quantitatively don't change any of kind of our results. Again, it moves things, but Epsilon relative to these other parameters.

 

>> AJ: The US tax code in the background.

>> Erik Hurst: So let me show you some of those as we get to those, let me kind of then here's my roadmap of what we're gonna do in the last half hour. Now, I'm gonna start by with the minimum wage. Everything's going to be a permanent change in the real minimum wage.

To start, we'll do temporary stuff. At the end, I'm gonna show you long run results. I'm gonna show you short run results. I'm gonna show you some transition paths, all for the minimum wage. And then I'm going to show you what happens when we go to the EITC.

I'll have to introduce the EITC. I'll show you how we could compare the EITC policy to the minimum wage. But the key part is I wanna highlight between these two policies, the transition dynamics are very different. And when we start integrating over periods of time, projecting from the short run.

It really matters whether you're doing the minimum wage or the EITC, because the pattern are not only quantitative, they're just qualitatively different. And I'll show you that when we get there. So that's kind of the structure of what's to come. So let's start now, God comes down. We start with the steady state of the world in a current $7.25% national minimum wage in the model, which is essentially binding on nobody.

And then we drop a minimum wage unexpectedly on this market. And we're gonna drop small minimum wages, medium minimum wages, large minimum wages unexpectedly on top of this market. And then see what happens to the employment, the earnings, and the welfare of all the different types of workers in the model.

And again, I'm just gonna focus on the low educated workers in what I'm going to show you today because the high educated aren't doing much. Okay, so this is the first set of effects. So I'm gonna go through these a little bit slowly. It's the non monotonicity that we were talking about at the beginning.

Each one of those lines is a different type of worker. So the blue line, that's one all the way off. To the left is gonna be a worker who was initially pre-minimum wage, making $7.50. So the bottom of the wage distribution. The purple line is a worker making initially $10 an hour.

And the red line is going to be a worker making dollar 13 an hour. So these are three different types of eyes that we're looking at. The minimum wage comes in and the x axis is the size of the minimum wage. So basically up to $7.25. There was no minimum wage.

That was an initial steady state. And then the minimum wage starts rising. And you can see that first worker, let's work on the left panel for Employment. That first worker, that $7.50 worker, their efficient wage, they're making a markdown of 25%. So their efficient wage is something like $9.50 or something.

So as the minimum wage rises to $9.50, they are starting to work more. The monopsony Power is declining. The firms are optimally choosing to hire more of these workers cuz they're still making some profits. And then eventually, as the minimum wage gets above $9.50, the firm says, hey, wait a second, I'm starting to lose money on this worker, and the firm is going to start switching away from them.

And so that's where you're gonna start getting this hump-shaped pattern. And then, again, there's a different level of the optimal minimum wage for different workers. So the peak of the worker initially making $10 an hour is something like 13. And so the first thing you wanna think about is the peak of this Laffer curve.

I call them a laffer curve. They're basically just a non-monotonic hump-shaped profile in the minimum wage in terms of employment depends upon the worker's type. And that means that the single minimum wage is just a very blunt policy tool to deal with monopsony distortions. In this model, where monopsony is constant across workers, when you start fixing the monopsony for the low wage worker, you're having no effect on the higher wage worker.

If you start going to the higher wage worker, you're starting to basically kill the lower wage workers in the long run. I'm not going to talk a lot about numbers in the paper, but if you look at the x axis or the y axis and you go to the right picture, that's labor income.

The peak of the labor income is just higher than the peak of the employment, why? Because they're getting the minimum wage effect directly. So their wage is going up and their employment's going up. So these workers really get better off and then you start switching away, Steve.

>> Steve: Some of the studies in the literature exploit national minimum wage changes in the heterogeneity across localities and how much it bites.

You would expect if your model is kind of basically on the right track that you'd see evidence of these-

>> Erik Hurst: Even longer, if we went out to the long run, we would.

>> Steve: Yeah, so, well, is there any hint of this kind of thing in the literature already?

 

>> Erik Hurst: Yes, cuz we match the short run exactly. Now when I show you the short run pictures-

>> Steve: No, I'm talking about the evidence of the non-monotonicity.

>> Erik Hurst: The non-monotonicity is the long run.

>> Steve: I know, I know, so you wanna look at a period where the wage change and nothing else happened for a couple years or so.

 

>> Speaker 35: 20, you need 20.

>> Erik Hurst: I mean, let me tell you. I mean, these are-

>> Erik Hurst: I'll show you all the time dynamics. Exactly, okay, I'll show you all that. I'm gonna show you Steve's picture to match the literature when I get to the short run in the transition dynamics.

Cuz that's where the literature really comes into these two year, three year, five years.

>> Steve: Even in the short run, some of these guys in Mississippi, if the national minimum wage goes up, they're already on the downward sloping part, aren't they? I wouldn't say the evidence is that people whose productivity is $2 an hour are not employed.

They're out there.

>> Erik Hurst: Yeah.

>> Steve: Right?

>> Erik Hurst: Let me give-

>> Steve: Shouldn't lose sight of that kind of thing.

>> Erik Hurst: Put this together in an order that I think we're gonna get to that soon. Let me just kinda do a couple more things, and then we're gonna come and I wanna talk about the literature.

And I'm gonna show you, focus in on where the period of time when the literature estimates, and I'll show you where that is in our transition dynamics.

>> Elena Pastorino: But Steve, there are, me and others have found out that three to five to seven years rate, employment difference, change in sign and size.

So this switches from positive to negative.

>> Speaker 37: So one quick question on this, following up on Steve. So many of these people are going to be on other government programs. So how are you netting all that out faster? Many people are in three, four, five.

>> Erik Hurst: So on this part, I'm just doing the minimum wage, holding everything else with no other policy.

Then we put the two together. I'm gonna have a tax and transfer system that looks like the US, and then I'll put minimum wage, and you'll see there's some interactions. I might not show you much of that today, but in the paper we do a lot. But right now, think about the minimum wage, just to get the fundamentals of the mechanisms, and then we'll try to put them together with a tax.

 

>> John Cochrane: I think where this comment comes in is that people's labor supply lasts are different. If your alternative is social programs with 100% marginal tax rate, that gives you a very different labor supply elasticity than our alternative, which is just leisure.

>> Sebastian: What's the tax code here?

>> Erik Hurst: None on this part.

 

>> Speaker 38: No government, they all funded minimum wage.

>> Erik Hurst: Yep.

>> Speaker 39: And we had a thing in the when we put it on top of that.

>> Elena Pastorino: Minimum wage, you get to see-

>> Speaker 39: So this is the purifier,

>> Erik Hurst: cuz another fact here is the very low labor supply, low labor force participation of low income people compared to very high income people, which I think-

 

>> Speaker 39: Yeah, we match that. Those are targets that we're gonna be matching

>> Erik Hurst: yep, exactly. So we're matching the employment rates and the wage distribution in our initial equilibrium. Okay, so this is now, instead of worker by worker, this is going to be all of non college workers together.

And again, going back, here's the non monotonicity that if we're going to look at non college employment, not the 750 person or the $15 person, put them all together. You see that hump shaped pattern that we were talking about early on, small minimum wage changes could actually have a positive effect.

The larger you get the minimum wage, the more negative the total effect will be on this group. And then the labor income peaks a little later, why? Because even those who aren't working, those who are working are getting the big wage increase. So there's a direct effect of the minimum wage in increasing earnings for those who are working.

So the peak is higher and a little bit later than it would be with the employment. Okay, we already talked about that. So here is now, suppose we put an $8.50 minimum wage. So we're starting with 7.25. We just increased the minimum wage by $1.25. What happens to different types of workers with this policy?

And in this policy, the lowest paid workers, like 7.50, that's where the minimum wage is currently binding, or some low wage worker. An $8.50, all it does is reduce the monopsony distortion. And you get positive effects on net, even in the long run, by reducing the monopsony distortion in this model, they're not huge.

 

>> Bob: Does anyone remember Stigler's famous paper on this

>> Erik Hurst: why don't you tell me a little bit?

>> Bob: Just, I give it exact replica of what you just said.

>> Erik Hurst: Okay, perfect. But the key is then when we start doing policy, let me show you what happens to a $15 minimum wage in the long run, which is well outside of those bands.

And so now we put a national $15 minimum wage on. And what is happening is you're killing the bottom part of the distribution. And so by a huge amount, it's basically firms are substituting away from these workers. If you take those cardinal mu estimates seriously, which is really the key part here.

That is saying that when those workers making 7.25, you have to pay them $15, the firms are gonna say, I don't need any of those cuz their marginal product is just so low. And you can see that unemployment is just plummeting for these workers in the right panel.

The middle is labor income. Even labor income in the long run is plummeting for these workers because no one wants to hire them. Those who do get hired are making more, but that's relatively small. And then the last part is welfare. What's the difference between income and welfare?

There's gonna be some disutility of labor. Supply in the background. So if I force you to work, you don't like that as much. And so that's where the labor supply elasticity quantitatively is making the difference between the middle picture and the end picture.

>> Bob: Do I remember correctly that you have a little on 100 diagram?

Why there's the Stigler Effect disappears at a certain point taken by anti-

>> Erik Hurst: Yes. So that was kind of the reduced form of stuff I showed you in the textbook case. The Econ One case is eventually you're going to hit the monopsony power and then whatever that monopsony is, once you extract that, you're not going to get any of the benefits anymore.

And then you're only going to get the neoclassical effects in the long run. And that's kind of here is all just the neoclassical effect. People are, firms are just substituting away by the card and the MU estimate of the labor substitutability.

>> Speaker 40: Even emphasize that seems like-

>> Erik Hurst: That's what I really wanted.

Those are the old for I should have maybe go back. Maybe I should've did this one more time because maybe I didn't say it and Patrick's getting mad because when I go backwards. But let me just kind of say these hump shapes. To be clear, the upward sloping part of the hump shape is the monopsony power.

So that's Omega. The downward sloping part of these profiles is the card in the MU. Labor substitutability and the fee. So the upside is omega. The downside is the fee. And those are the two parameters. So if the omega was basically no wage markdown, you'd only get the down.

That was Dan's part early on. If you had less labor, labor substitutability, that downslope is going to be a lot flatter. And that's what I'm going to show you next. And so to Bob's point, I promised. The paper has lots of pictures like this. I'm going to show you one.

That middle box is kind of our baseline parameterization card in the MU substitutability of 4, the wage markdown of 25%. That's basically saying to a $15 minimum wage, the non college labor group on average would be in the long run, declining employment by about 12%. If you basically go to the right to make the markdown even smaller, those numbers get bigger.

So the more, this is Dan's comment. If you take monopsony, I didn't take it all the way to 0, but I'm taking it close to 0. And what you're getting there is only the negative effects.

>> Bob: On the top line minuscule 2, you get 0.75 and 0.9. The Stigler Effect disappears completely at a markdown of years.

 

>> Erik Hurst: So markdown of 40% and no labor, labor substitutability. It means even a $15 minimum wage, in the long run, has essentially no negative effect to non college workers as a whole. There might be some in the distribution sense that are made worse off the bottom part. But as a whole, this is, you know, this is what I want to say.

The parameters that matter. Does this Stigler kind of the Stigler Effect? If the substitutability is really low and the markdowns are huge, you could even handle more higher minimum wages. Now, you can't go to $100 an hour, but in this calibration, you can go to $15 an hour.

If the markdown was large enough and the substitutability was small enough and you'd have no negative effects in the long run.

>> Bob: I was asking about the other end, where there's no stable effect, because there's no markdown going over here.

>> Erik Hurst: And so here,

>> Speaker 42: if you go that way, they just get killed.

 

>> Bob: The question is why is it 6.3 and 6.4? There's almost no difference between 0.75 and 0.9 even though that number must go to 0 if it's 1.0. Plus, I've missing something.

>> Erik Hurst: I have to go back and see what is the quantification, why those numbers don't change.

But the key part is, if there's no substitutability, it doesn't matter. Suppose we're in a world where I needed all John productivities. John is like, you know, I of type something that gave him 750 in the world. If I had a production function where I needed John in that production, it doesn't matter what I have to pay them.

 

>> Speaker 43: If you go Leontief, you keep going flat until you shut the whole thing down.

>> Erik Hurst: Until you shut the whole thing down, there's going to be some point we're going to show down.

>> Speaker 43: So make fee go to 0. You go Leontief until you say it's not worth operating this firm, you just shut it down.

So be 000 till it went to like 35 in our model. Then the economy shut down.

>> Speaker 44: Wait, wait, let's keep going. That's not a big deal. What you're talking about

>> Speaker 45: is the monopsony there to explain effect in the data or just to make the minimum wage?

Do something.

>> Erik Hurst: You want to give them the minimum wage, a chance to do something good. Now it's also matched. The data says people are estimating something, that it could be there. But in the model, it's a way to make the minimum, give the minimum wage or any policy a fighting chance to have positive welfare.

 

>> Speaker 44: Since 1923, Joan Robinson argued that's why you should have a minimum wage. Every single pro minimum wage paper has monopsony. We wanted to be balanced and say.

>> Speaker 46: It's not a set of facts.

>> Speaker 44: I'm just telling you.

>> Erik Hurst: You were trying to match set of facts that other people.

Yeah.

>> Speaker 46: Okay, good. That's what I wanted to know.

>> Erik Hurst: Okay, so let's go now to the short run because I want to do a few more things before in the last 15 minutes. So what I'm going to do now is kind of give you the whole transition dynamics.

And so this kind of goes to Steve's question. The lines on the left, let's start there are employment. The Ryan line, the right are going to be labor income. And I have three lines there, and those are for different sized minimum wage changes. So this is all non college workers we're looking at now, pooled together.

I'll show you different types of workers in the next slides. And the black line is what happens if we put a $15 minimum wage. The red line is what happens when we put like a, I think I'm going to use a $10 minimum wage. And I think the blue line is going to be the $8.50 minimum wage.

And so let's start with the top two lines. And you can see from those top two lines of that, the short run and the long run kind of look the same in the sense that they are so small, the minimum wage, that it reduces the markdown. It creates a little bit of a positive effect on workers, even in the short run, small.

And then over time, firms don't want to substitute away from those workers. They're still making profits on those workers. So the long run effects and the short run effects are relatively identical. And if anything, the short run effects are small relative to long run effects because the workers might reallocate some of their production towards those workers because they're hiring more of those workers in the long run.

And so you get a little extra kick. That extra kick is quantitatively small. But the key thing is if we have small minimum wage changes, the short run response is not a bad approximation for the long run response. However, with a large minimum wage change, the short run response is a little negative.

But the longer you go over time firms are going to start substituting more and more away from that, and then the short run response is going to be completely uninformative about the eventual long run response. The shaded area is where. I don't know if you guys had David Newmark out recently, but David Newmark has a paper summarizing 100 papers.

Mostly all from the regional stuff, as Steve said, which basically had a range of estimates from the literature about. And all of this literature was about two years out. Now, there's some papers now that do a little bit longer than two years, but the bulk of the literature looks two years post a minimum wage change.

That's going to be that shaded thing. You could have a small minimum wage change, a medium minimum wage change, or even a very large minimum wage change, and they're all in the same kind of ballpark of relatively small estimates in the short run. And so this is my answer to Steve's early on.

The longer you go out, though, you should start seeing more differences between these policies. And some papers have done longer time horizons for smaller changes, I don't have a pointer, but for smaller changes, again, the medium run and the long run and the short run are roughly the same.

Yeah.

>> Steve: What's the evidentiary foundation that led you to structure your model in such a way. That long run substitution from low skilled labor to capital is small, is really small.

>> Erik Hurst: So we did, we took three estimates. We did Cobb Douglas as the baseline. But others, we've tried to put elasticity, a substitution of two instead of one and we did it.

 

>> Steve: Do you have any direct evidence on this?

>> Erik Hurst: People have estimated Juan Carlos left, but Juan Carlos has a paper on trying to estimate that says it's not too crazy different from Cobb Douglas. Core has a different kind of structure.

>> Erik Hurst: Yeah. And so we've tried to play around with some of those.

We tried to hardwire. We thought that when we originally wrote the paper, we thought that's where the biggest kick was going to be.

>> Steve: Yeah, that seemed like a

>> Erik Hurst: quantitatively it was all the labor. Labor substitutability was much more quantitatively important, at least given because the card in the ME were estimating things like four to six.

The people doing the capital subsidies sometimes get 1.2 or 1.3. And so they're just quantitatively small relative to this.

>> Speaker 47: So what's building over time if the capital labor substitution, what's building so much over time and the dotted line on the left?

>> Erik Hurst: It's the production plans. The capital is changing, but it's changing for peat worker versus Patrick worker.

 

>> Speaker 48: So what is the capital labor substitution. With the high five

>> Speaker 49: and the work

>> Speaker 50: different than the user type?

>> Speaker 51: Yeah.

>> Speaker 49: It's a different.

>> Erik Hurst: A lot of experiments including some at the Minneapolis fed, look at hours effects. Do you ever get it?

>> Speaker 52: We don't have no hours.

 

>> John Taylor: Margin was a time.

>> Erik Hurst: Yeah, I'm almost done. Okay, let me, boom. This is all the things I said in words. Let me kind of show you now two types of workers, the transition path. So on the left is going to be a small minimum wage change and the two lines are initial worker making 750 an hour and the red line is going to be a worker making $12 an hour.

When you put an 850 minimum wage on this, it hardly has any effect on the worker who's making $12 an hour. That's what the red line is, dollar twelve. Now the worker making dollar seven to dollar 50 gets a little better off. Like when you give them a 15% minimum wage, their wage increases by about 15% and then as firms start adjusting their capital towards that worker because they're hiring more of that worker, they get a little extra kick going forward.

Whereas the large minimum wage change, and this is the key part here for a large minimum wage change. Take a look at that 750 worker. They earn huge amounts of gains in the short run. Why? Because firms aren't substituted away from them and their minimum wage doubled. They were earning 750 and now they're being paid $15.

So they get 100% increase in their labor earnings in the short run. But over time, firms start substituting away from them. So when we do the. If we want to evaluate this policy, you got to integrate over the short run gains in the long run costs. And those could be very different from each other for big minimum wage changes for certain types of workers.

Now notice the $12 worker. When you go to 15, you're actually getting closer to their efficient level. And so they get a jump in the short run, and then they get even more of a gain in the long run as firms start switching towards them.

>> Speaker 53: What's the main reason that there's this lagged adjustment of common firms.

 

>> Erik Hurst: It slowly adjust their capital stock. So the putty clay nature. So think about it as way they have this capital there. It's already installed, so they can make a choice to either use it or shut it down. And then they're buying new capital over time, and they're going to buy new capital to fit whatever the production mix of the workers they want.

And so Elena uses a certain type of capital as a low skilled worker. I use a different type of capital as a high skilled worker and as things start changing over time, we're going to start making the capital. If I hire more of Elena, I want to have more machines that go with Elena and that increases her marginal product a little over time.

 

>> John Cochrane: This is really weird though. Sorry, go ahead.

>> Bob: You said a little.

>> Erik Hurst: I'm just saying in this example, if she was the $12 worker, it would be a little on the right hand side when we go to $15.

>> Elena Pastorino: But if it was $10 an hour, you would do it.

 

>> John Cochrane: The government goes and makes Elena more expensive and your reaction is, well, I'll buy more capital. That's specific to Elena. Now I understand what's bigger.

>> Erik Hurst: Substituting away from the ones that became more expensive. You're only substituting Elena when she is still profitable to you.

>> John Cochrane: Right.

 

>> Erik Hurst: So Elena is still profitable. You made Elena a little bit more expensive, but she's still profitable to you. But to all the other people below Elena, you've swept away from and so now you're switching to all other workers.

>> Speaker 55: Remember to static monopoly. As long as she for you specifically, you're below your peak, you're less than loosely 25% over what you're making.

You want to go up because you wiped out your monopsony power. And when you hit the peak, then you go straight competitive logic down.

>> Erik Hurst: So let's take John's question with the right hand picture. You can see they're substituting away from that $750 worker

>> Speaker 55: because you're over the peak,

 

>> Erik Hurst: you're over the peak, and now they're substituting towards all other workers. And one of those is going to be towards Elena and so she's getting, you're making some capital to go with. Even though she's a little more expensive, you're still going to willing to use her, as opposed to the people you're switching away from completely.

So you're just basically killing the bottom of the distribution, moving up the distribution. And now if we went to too high of a wage, you'd start switching away from Eleanor.

>> Elena Pastorino: You see it overnight.

>> Speaker 58: But this turns very much on this very high adjustment cost of capital.

 

>> Erik Hurst: But there's no cost. I mean, the way we model, it's a putty clay structure. It is.

>> Speaker 58: You just can't do it.

>> Speaker 56: You don't burn up real reason.

>> Speaker 57: Infinite cost.

>> Speaker 58: Like we've done a lot of the minimum wage.

>> Erik Hurst: But the key thing is, so there's two parts of this.

There's on impact, which allows us to match the data from the minimum wage, which basically, no matter what you do in the minimum wage, in the short run, people are getting relatively small response on impact. And so we needed some mechanism to match that. And then we have this long run adjustment that other people have estimated and so we anchor that.

Now, how you get from the long run to the short run, how long that time is, we have a specific structure to do that, but that won't change the short run effect. There's no effect going on. The long run effect seems to be going on and then we've just had the putty play structure to get us from the short run to long run.

So you might say, maybe we could get to the long run quicker. Well, then we'll just play with the disappreciation rate or you get the long run slower. We could just play with the depreciation rate. But the key thing is the short run, nothing's happening because firms can't adjust their production mix on impact.

We have that. We know in the long run, this other literature makes them adjust. We have that. And then the speed of adjustment is really just the depreciation.

>> Steve: Question about modeling. So you did not include a mechanism used to get some attention that increases the minimum wage crowd out general human capital accumulation on the job.

Is that because you just don't want that complexity or you're kind of persuaded by the literature that's not a very important mechanism.

>> Erik Hurst: We blah, blah, blah about the skill upgrading part. It is an additional layer of complexity in an already complex model.

>> Steve: Because that could reverse some of your long run results.

 

>> Erik Hurst: And then Bob's force could enhance some of our things if the firm start not entering, and then we have a lost set of firms to go through. So there are two forces, these two big forces that are in the background that we could sign, which way they go, and then those are not the forces we're highlighting here.

Let me do three things quick before I'm just going to show you three. One is how does the transition affect the welfare dynamics relative to the steady state? Let's go to the $15 minimum wage, which is going to be on the right side of this picture. The black, the dotted line is the steady state long run effect that I showed you before.

A dollar 15 minimum wage makes certain workers worse off in the long run, others better off. Some workers no effect. The solid line is what would happen if I integrate over the whole dynamics and not just look at the steady state and you can see the gains to workers at the bottom are going to be higher, not quite as bad when you incorporate all those short run gains.

And it could be quantitatively large in terms of doing this. So here, even in our model, a $15 minimum wage helps a chunk of workers over the long run. Now, it still has a negative effect on the bottom, but not quite as negative you would get just from the pure neoclassical story, just with the long run.

Let me kind of show you a temporary change in the minimum wage. Let's go to both pictures. The left is all non college employment. The right is going to be just for a worker who's making 750. And the key things I'm showing you here is just what happens if we have a transitory minimum wage.

The red is large, $15. The blue is small. The key thing is, from temporary changes in minimum wage, you only get the gains and you have none of the costs. So suppose you increased once the nominal minimum wage to $15 and let it deflate away by 5% a year over time.

Firms aren't going to respond to the temporary change because workers tomorrow are gonna be relatively cheap again, and so they're not gonna adjust their production structure. So temporary changes in the minimum wage, you get all the gains without the firms substituting away, yeah.

>> Speaker 59: Basically, it is unindexed.

 

>> Erik Hurst: Now, suppose you did that every year, And you don't do it again. Then you do it again, and you don't do it again.

>> Erik Hurst: Yeah, and so then the last result I'm gonna show you now is what happens when we add an EITC, just as a comparison.

And so for those who are unfamiliar, so the EITC gives you kind of a subsidy to work a little bit, then it kind of plateaus for some range, and then the subsidy starts phasing out. And so you're starting to face kind of a negative tax rate or early on, a positive tax rate later on.

And so what we're going to do is we're going to expand the minimum EITC in our model. How are we gonna do that? We're going to fund it by taxing profits, and we're going to fund it by taxing profits by exactly the same amount of loss of profits from a $15 minimum wage.

So this is apples to apples in terms of profit losses from a $15 minimum wage and the expansion of the EITC. And so now, we're gonna ask what happens when we expand the EITC in this model by playing around with these marginal tax rates, making them negative at the bottom and then phasing them out over time.

And there's some degrees of freedom of how big the peak is or how high the slope is. But all of that is gonna be second order to the point we're gonna make. Is look at the welfare to low skilled workers on employment on the left panel, income in the middle panel, and welfare on the last panel, between the minimum wage in solid lines and the EIT expansion in the red lines.

And you could see in the long-run, the EITC actually makes low-waged workers better off relative to the minimum wage. If you wanna help low-wage workers, expanding the EITC is orders of magnitude better than the minimum wage. Now, why is that? Because when the government pays the marginal cost, the firms aren't paying any of the additional costs to these workers, so they don't face the incentive to substitute away from them.

And if anything, this is my last picture, and then I'll conclude on the next slide. If anything, this is the time path of the dynamics of an EITC for, again, a 750 worker and then some higher paid worker in the red line. But the key thing is look at the blue line, the worker 750, they get a gain on impact, but the gain grows over time for exactly the same reason we were talking about earlier on, is that firms are going to hire more of these workers, and then they're going to start adjusting their capital stock to those types of workers.

They're not paying the cost of making them more expensive, but they are hiring more of them. And so the time dynamics actually go the other direction relative to the minimum wage. So part of our paper is we have a framework that kinda thinks about these time dynamics, and different policies go in even different direction.

The minimum wage makes low wage workers really better off in the short run, but worse off in the long run. The EITC makes them kind of better off in the short run, but even better off in the long run because they face different incentives to adjust their input mix.

And so that's kinda what the goal of the paper, and this is my last slide, just to conclude, is thinking that these transition dynamics seriously are gonna be important when we're thinking about policy. And in particular, a lot of our empirical work is on small policies in the short run, whether it be minimum wage or EITC.

Using those empirical estimates to forecast the dynamics in the long-run are fraught with potential errors. And how big those errors are depend upon the monopsony power, the degree of substitutability, and the speed at which firms adjust.

>> Speaker 60: Have you thought about what a wage subsidy rather than either an EITC or minimum wage-

 

>> Erik Hurst: Yes. Yes, we started with that as the first best policy. If we could basically have a wage subsidy that targets the markdown, that's the first best policy, with some other tax funded, depending upon how it gets funded with

>> John Taylor: As organizer, I have to say, Perfect. Those who want to go are free to go, thank you very much.