How does rising income shift labor demand across skill levels?
Core argument: A shift in labor demand toward higher income-elastic sectors accounted for an estimated 90% of the rise in the wage-bill share of low-skill jobs and 35% of that for high-skill jobs between 1980 and 2016.
The US labor market has experienced polarization driven by a shift in labor demand towards higher income elastic sectors, affecting both high and low-skilled occupations. Between 1980-2016, this shift accounted for 90% and 35% of the increase in the wage-bill share for low- and high-skill jobs, respectively, and 64% and 28% of the rise in employment shares. As aggregate expenditure grows, demand shifts to high-income-elastic sectors, intensifying the need for high and low-skilled workers while reducing demand for middle-skilled roles. This reallocation has led to a decline in the wage bill for medium-skill occupations by 8 percentage points, while increasing it by 2 and nearly 6 percentage points for low- and high-skill jobs. The model predicts continued polarization, with significant changes in relative wages and hours worked, particularly for low-skill occupations, from 2016-2035. This trend underscores the asymmetric impact of income elasticity on labor demand across different skill levels.
Ed Comment:I don't understand this. (I haven't read the paper yet). Plz try to explain it to me in plain English as best you understand it. What is a wage bill? (I'm not saying it's not an interesting paper, because it sounds interesting. Our apple core showed little change in the distribution of wages except from an overall upward drift, significant change at the top 1%, and slight upward change in the relative wages of the top 10 or 20%. We only saw dramatic change like this if we defined the middle class with silly dollar bookends, like Pew. I don't see how the demand for labor can vary much from the supply other than demand for (real) talent because it's supply is constrained, especially in the short run. I get how (phoney) talent (ie unneeded college education) can appear to distort the elasticity in the demand for talent, i.e. make the demand for real talent appear to be more elastic than it is. I don't see how the market can demand much more than the supply of unconstrained labor (unless you run large trade surpluses by exporting savings) because supply (is what) creates (it's own) demand.
Steve Comment: Table7 “…. The key take away is that the neutral increase in expenditure was very relevant in the labor market polarization during both subperiods, however it was somewhat less important for low-skilled labor in the 2000-2016 period. More specifically, we find that the neutral increase in expenditure observed during the 1980-2000 period more than accounts for the evolution of relative wage and the share of hours for low-skill occupations, and it accounts for 70% and 61%, respectively, of the change in the model simulations for relative wages and the share in hours worked by high-skill occupations. In the 2000s, the contribution of the neutral expenditure decrease but it is still key for the model success in explaining the observed polarization of labor markets. It generates between 50% and 60% of the change in the relative wage, share of hours worked and wage bill for high and medium-skill occupations and around 35% for the low-skill occupations…”
Diego Comment:The distinction you make is important because changes in the relative wages reflect both demand and supply forces. The consensus in the literature is that because relative wages increased for high and low-skilled workers, the increase in the demand dominated the increase in the supply of hours worked by those workers. Hence, the increase in demand has been very significant because hours worked have also contributed to the changes in the wage bill shares.It is also important to realize that over long periods of time, labor supply is endogenous. Therefore, channels like ours or technology will affect how many high and low skilled workers are in the economy. That is why we model the supply too, and show that indeedmuch of the contribution to the polarization of wage bills through the income channel takes place through the endogenous changes in supply. You can see that in tables 7 and 8 of the paper.
Steve Comment: Diego, Do you have a version of the table of the wage bill evolution decomposed into the wages and (times) number of employees/workers bythe three skill segments over time? The changes in wages relative to one another would imply/show demand relative to supply.
Ed Comment:In particular, I am curious to see the wage bill decomposed into the wages and (times) number of employees over time. Seems odd that they don't show it. Relative changes in wages ought to indicate demand relative to supply. With the distribution of wages not changing significantly other than to shift upward, I don't think they've found nearly the insight they seem to have. The paper grossly exaggerated the hollowing out of the middle because the movement they find is predominantly (slightly) upward. Ho hum.I continue to think they make an interesting point: that demand has shifted to where ever it has shifted. But they way over attribute that driver. And I still think it's missing big pieces of logic. One the one hand, if skill (in general) constrains growth, then an increase in skill will be demanded no matter wether there is an increase in prosperity. Similarly, if IT adds value, then specific IT skills will be demanded, no matter the general increase in prosperity. If the supply of low-skilled hispanic immigrants increases, at the very least they will work for each other. More likely, prices will adjust to increase demand by other more skillful workers for them, no matter the general level of prosperity. On the other hand a more prosperous population will likely spend their incremental dollars on healthcare (or housing) all other things equal. Or perhaps on dinning out, landscaping, and construction if low skilled labor is cheaper/more abundant. Complicating matters, an increase in skill, breakthrough in IT and increase in the supply of low-skilled workers will increase prosperity. I think they grossly simplify how preferences change as prosperity increases. They assume it's all exogenous like healthcare. But prosperity increases for different reasons and demand is affected by changing prices, which in turn is affected by supply (of talent and labor), opportunity (of which the advent of IT opens a window) and changing preferences, which are partially driven by rising prosperity. I hardly think the story starts and ends with exogenous changing preferences (e.g., for healthcare as the population grows more prosperous). But it's a worthy point that needs to be incorporated into a comprehensive macro economic theory.
Steve Comment:They are using an older Autor skill level classification that I wasn’t familiar with (it’s in Appendix B) Autor had classified occupations into the three skill categories (High, Middle, Low) according to those occupations average wage in 1980, so from my read they are effectively using occupational wage in 1980 as a proxy for skill: “….Our occupation classification is also taken from Acemoglu and Autor (2011). They divide the 382 original occupations into 4 broader categories that are characterized by their skill level: (1) managerial, professional and technical occupations; (2) sales, clerical and administrative support occupations; (3) production, craft, repair and operative occupations; and (4) service occupations. The first group is characterized by high skill occupations, the second and third groups are characterized by middle skill occupations and the last group is characterized by low skill occupations. They measure skill by the average hourly wage of individuals in the occupation in 1980 where the mean wage in each occupation is calculated using workers’ hours of annual labor supply times the Census sampling weights..”
Diego Comin Comment:Indeed, Steven. One relatively surprising finding we have is that how income is distributed across the population does not seem to matter much for the income driven channel.
Steve Comment:Thank you very much for the wage bill table, very helpful. The market income story pre-tax transfers is a much more interesting story than post taxes/transfers. Got it. In terms of netting out the top/bottom 5%, I guess including those groups (mismeasurement issues aside) wouldn't change the story in the paper if the recent Auten-Splinter and Smith/Zidar/Zwick work is directionally correct.
Ed Comment:Add to the data base and mark important. It’s hard for me to interpret their answer without having the questions you asked him. The distribution of wages has scarcely changed, except at the 1% tail. They argue that demand shifted (nonhomotheticly) away from medium-skilled work. That should have flattened out the distribution. But we didn’t see a significant flattening. Pew found an 11 percentage point hollowing out of the number of middle class households by using dollar bookend (which I think are silly): 7 percentage points up, 4 down, with 3 of the 4 down from an increase in low-skilled Hispanics between 1971 and 2017. Between 1980 and 2016, the authors find the total dollars of wages to have shifted upward 19 percentage point and down 2. It’s also hard for me to understand why the changes in the demand for skill seems to correspond so closely with changes in the supply of skills. College hollowed out some of the middle skilled; and low-skilled immigration added to the supply low-skilled workers, and that’s exactly what we see. Do they decompose the total wage bill for each skill level into: wages (times) the number of workers? It’s not clear to me how they defined skill levels. How do they define skill levels?
Herewages are before taxes, so all of our analysis is better interpreted as describing the evolution of labor markets before any redistributive policy is implemented.With respect to the top 1%, the data at the CPS and ACS is top-coded and additionally, we eliminate the top and bottom 5% of earners. Therefore, we cannot say much about the evolution of wage bills once those earners are considered. However, as you can see from the table, there has been very significant reallocation of labor income below the very top with a mass of medium-skilled workers that in 1980 occupied 65% of the labor force, and that has experienced very significant reductions in both hours worked and wages relative to the other occupations.On who receives capital income, we have shown that it does not matter much for the paper’s conclusions. In the baseline we distribute capital income evenly across households, but we also explore the more realistic scenario where the high-skilled get all capital income. The results are similar.In our analysis, wages are endogenous and equal to the marginal product of labor. So, we already consider the effect that lower number of middle skilled workers will have on the middle skilled relative wages. In spite of the reduction in middle skilled hours worked, their relative wages declined very significantly which suggests that the demand for these workers declined very significantly.I hope this clarifies the paper and answers your questions. I’d be more than happy to talk over the phone if you have any question.
0.09
0.42
0.49
2016
0.07
0.51
0.42
2000
0.07
0.63
0.30
1980
0.07
0.74
0.19
1950
Low
Medium
High
Diego Comin Comment: First let me apologize for the delay in our response and thank you for your interest.Let me start by correcting a bit the characterization of our story. Our argument is not just that the expenses of the rich are lifting the poor (and that the reduction in expending in middle skill intensive goods and services are depressing the livelihood of the middle class). This effect, though part of the story is not the bulk of it. What we have documented is that the spending patterns are common across income groups. Any household, rich or poor, changes their pattern of consumption away from income inelastic goods and services and towards income elastic ones as it experiences an increase in income. Therefore, it is not the rich that are spending more in health, education and restaurants. It is all that are experiencing increase sin income over the last 40 years (actually we go back to 1950, so the last 70 years). Hence, the effects on polarization are much larger than if the income channel was driven just by the rich. I paste next the table that I think you as for. This just shows the share of the wage bill accrued by high, middle and low skilled occupations at different points in time.
Steve Comment:I just read your "Income-driven Labor Market Polarization" and had a few questions that I was hoping you could answer or point me in the right direction. Your findings are intuitive to me, higher incomes creates higher demand for lower skilled (for lack of better phrase) services (restaurants, landscaping) and higher end work as talented workers increasing service each other (Bloomberg workers provide data for bankers) creating a self-reinforcing feedback loop that excludes middle skilled work. I was hoping you might have the absolute share of income going to your three groups overtime across sectors? (Ie in 1980 high skilled workers earned X% of total wage bill, middle skilled earned X%, and low skilled earned X% sum =100%. in 2000....2016...) so not the relative change but just a snapshot in time? My understand is most of the cutting edge research has convincingly showed the only major change in distribution of market income was only at the very top (top.1-.01%)How do you think about (and maybe I missed this in paper failed to understand it ) craving out productivity/marginal product of labor from relative labor supply? For example, less supply will raise wages (ie shadow prices) by (necessarily) pruning demand for jobs with lower marginal products of labor, which in turn raises the marginal product of labor. We can only measure productivity from wages, which can only equal the marginal product of labor no?From my read you attribute all capital income to the high skilled, I assume this wage share measure is gross (ie it's pre-tax compensation) Thank you and congratulations on the paper!
Ed Comment:The sum of the sectors has to add back to the Apple core, i.e., no significant change in the distribution of wages. Again, because the supply of labor creates its own demand, if you add 35million low-skilled Hispanics, you are going to see growth in unskilled employment. Similarly, when talent constrains growth, you will see an expansion in the employment of talent as the supply grows. People will eat in restaurants and buy more landscaping, rather than tutor their children if that's who is available to hire. Prices will adjust to equilibrate supply and demand, in large part by raising prices (on skill) to reduce demand. So prices for real skill will rise enough relative to the price of others to employ all the low-skilled Hispanics. As I have said in my book and chapter: if high-skilled (or whites) can gain more income from hiring and working for high-skilled, it will create a self-reinforcing feedback loop that excludes others. But the low-skilled waiters will then wait on each other. The wage that matches that marginal product of labor will ensure they all get hired. What you won't get is a wage (ie shadow price) that could have been higher if their supply had been more constrained. It's not obvious to me how to distinguish their productivity/marginal product of labor from their wages and relative supply. They are all entangled. For example, less supply will raise wages (ie shadow prices) by (necessarily) pruning demand for jobs with lower marginal products of labor, which in turn raises the marginal product of labor. We can only measure productivity from wages, which can only equal the marginal product of labor (unless their is cronyism). Then the low wage workers will resent the fact that you didn't support policies that would have given them higher wages and vote for guys like Trump or the Dems.
Steve Comment:I wasn’t familiar with the termwage billeither. It’s the total sum of the gross wages of an establishment, here they are using it as the “sectorial wage bill of each occupation” My read of their finding is as national income grew in both the US and Europe there was more demand (so higher elasticity) for both high skilled and low skilled labor, and that effect is what has been driving the income polarization that they find. Their high skilled makes an effort to capture talent (or Benzell/Brynjolfsson’s G share) by assigned all capital income to workers in high skilled occupations so the talent constraint is somewhat captured. They are using a skill measure based off Autor’s work that classifies occupations into three skill categories based off average wage in 1980 so the skill is implied. In terms of reconciling this with the apple core will have to spend a little more time with it. We carved out the apple core to show absolute amounts this is attempting to measure relative changes. They don’t have a great table assigning the absolute wage bill by sector overtime which would make it very clear (at least to me). Need to reread.