Rising U.S. Wage Inequality: Whither the Middle Class?

Robert Valletta discusses rising U.S. wage inequality and how it has affected the American middle class (video, 27:01).

Robert Valletta discusses rising U.S. wage inequality and how it has affected the American middle class. Robert also discusses potential causes of rising wage inequality, including the roles of education, technological change, trade and immigration, and institutional factors such as union activity and pay practices.

Download the presentation slides (pdf, 448 kb)

This video is also available on our YouTube channel.

Video Chapters

Hide this section
  1. Introduction & Talk Overview
    0:00 - 3:17
  2. Trends in U.S. Wage Inequality and Why do we Care?
    3:18 - 9:23
  3. Rising U.S. Wage Inequality: Patterns
    9:24 – 14:52
  4. Rising U.S. Wage Inequality: Sources
    14:53 - 25:21
  5. Summary: The Middle-Class Squeeze
    25:22 -27:01


Hide this section

00:00:01     Good afternoon.  Thank you all for coming.  I’m happy to be here to talk with you about this particular topic.  The title of my talk – following on Brad’s fine presentation – is Rising U.S. Wage Inequality: Whither the Middle Class.  Now, you’ll notice that that’s a slight change in subtitle from the original one which was Extent and Causes.  That’s for two reasons; one, Extent and Causes is really boring;  and two, as I looked more into the relevant research, I realized the story that comes out that’s probably the most interesting one to tell in regard to wage inequality, is what’s been happening to the middle class,  and what’s likely to happen to them going forward as well.  So, rising wage inequality in general. First thing to note, it’s not just an issue of the one percent versus the 99 percent.  That particular cut on wage inequality and income inequality has gotten a tremendous amount of attention over the last year or so, with the Occupy Wall Street movement, and so on.

00:00:59     And I think that based certainly on what Brad presented, that that attention is quite justified;  it raises important issues.  On the other hand, the wage distribution, the distribution of earnings by individuals at their jobs, has been getting pulled apart for the last three decades.  Not just with the very top pulling apart from everybody else, but even within narrower groups.  Sort of middle earners versus top earners,  and bottom versus middle, there have been changes over the last three decades that have

00:01:30     implications for all 100 percent of the population, not just the 1 percent.  I think it’s important to see what those are,  and the trend is going to emerge here, and it’s not overpowering at this point, but I think you’ll see it in some of the displays that I’ll show, is that there’s a middle class squeeze going on.  And I use that in a somewhat loose sense.  It’s in the sense that people who were formerly regarded – families and individuals formerly regarded as being in the middle class – are first of all seeing their incomes either decline in absolute terms – that is in inflation-adjusted terms.  Also seeing them decline relative to higher income people, so they’re not keeping up with the highest income people.  And also, in some cases, seeing their incomes erode relative to people who are lower down on the income distribution,  and that’s the

00:02:19     sense in which the middle class is getting squeezed.  So, to just outline the talk briefly, I’m going to start with some basic facts and definitions just so that we’re speaking in common terms.  And I’m going to raise the question and try to answer, why do we care? Because this question comes up quite frequently when inequality and rising inequality is discussed in the public arena.  Then I’ll talk about the patterns of rising wage inequality, and I’ll finish with a fairly lengthy discussion of the sources, focusing to a large

00:02:52     degree on the changing technology of production in the workplace, but also talking about some factors that are related to that, and may be independent as well.  So, I’m going to try to give you a partial accounting for what’s been causing these changes in wage inequality.  Now, I apologize for this table of numbers.  I don’t expect you to absorb all of these numbers.  But just to give you a sense of what we’re talking about, this is a listing of household income in the left-hand column and typical annual earnings for people.  The household income data is direct from the Census Bureau based on an annual publication they put out.  Earnings are tabulated from the earnings records that are available in some of the data that comes from the Census and the BLS.  And I’m going to be dividing up the wage and income distribution into what

00:03:45     are called percentiles.  So, the 10th percentile for example, is the point at which for a household income, 10 percent of households have income below that, and 90 percent have income above it.  And you can see that in 2010, that was quite low.  That was about 12 thousand dollars.  By comparison, the 10th percentile – that is the 10 percentage point of the earnings distribution – is a little bit higher because a lot of low income households have very unstable work histories and may not, in general, have a full-time year-round earner in the household.  The P50, as I’ll be referring to it, is
00:04:22     just the median.  It’s the point right in the middle of the distribution.  You can see that was about 50 thousand dollars in household income.  And in the 90th  percentile is quite a bit higher. It’s almost double the typical annual earnings because high-income households tend to have multiple earners in many cases, and also they get income from other important sources other than their wage and salary earnings.  And the key thing I’m going to be focusing on to talk about inequality is these ratios of income at, for example, the 90th percentile relative to the 10th percentile.  That’s a way of giving you a sense of the overall spread of the distribution, how wide it is,  and that’s a nice metric, or way of measuring how it’s changed over time.  Everything I’m going to be saying is largely

00:05:11     invariant to using alternative measures of the dispersion, or gaps in income and wages across groups.  This is just the simplest one to use.  So, here’s what’s happened over the last 30 to 40 years.  This is a plot of the 90 to 10 ratio for household income in red, and wages in blue.  And you can see both of these have increased quite tremendously over the last three decades or so.  The wage data only start in 1979.  So, that’s why the plot starts there.  What you can see is, prior to that point, the household income distribution

00:05:49     was moving up and down, but was actually fairly flat.  And what happened is that income inequality – and you could also see this largely in Brad’s slides as well – income inequality in the United States, and wage inequality as well, started to pick up quite substantially around 1979 or so.  One of the important things about this chart is that the two series, they’re on different scales, but they’re moving largely in lock-step.  The increase in inequality in household income largely reflects the increase in inequality in wage income as well.  That’s important because when we think of the well-being of families and individuals in this country, we generally think of what their household income is.  That’s the key measure of just how well they’re doing in terms of their living standards.  When we see inequality in household

00:06:40     income, that raises the question of what’s causing these differences in the improvement in the living standards of different families,  and what this chart tells you is that it’s largely changes in the wages that are being earned for different types of workers in the labor force.  So, one question about this is, okay, I just showed you the gaps have grown quite substantially over the last three decades.  And a question often gets raises, why do we care?  In a market-based economy, it’s not surprising that wages and income

00:07:12     reflect effort and ability.  That’s a good thing in many ways.  That means that workers are receiving the correct incentives to exert effort.  And it’s a reflection of an efficient market-based economy.  Well, there are some reasons I would argue that we should care.  You may place varying amounts of weight on these different reasons, but they’ve all been raised in one context or another.  One issue is the general issue of social justice.  If some individuals in society are seeing their wages be largely flat, but other people earning a lot more, it may undermine their incentives to try to work hard and try to advance themselves.  And perhaps more importantly, it may have adverse political fallout.  Something like free trade for example, as I’ll talk about later, may arguably be a contributor to rising wage inequality.  And people who are hurt by

00:08:04     that free trade may start pushing for trade restrictions,  which many economists regard as being as detracting from the efficiency of the economy.  So, for those reasons we might care about people’s perceptions of where they are in the income distribution.  The other issue is the extent to which rising wage and income inequality might reflect market versus non-market forces.  Now, a typical economist would say, well, if it’s purely competitive market forces that are at work, then the differences in wages – that is,

00:08:40     some groups seeing much bigger gains than others – again, that’s just efficient market functioning.  But if there are non-market influences, either forms of government regulation that have changed, or unfair trade practices by other countries, then that may raise questions about potential policy remedies,  and just the overall economic efficiency of the changes that we’re observing.  But here’s the kicker, I think - and this is the one I’m going to lean on very heavily – if we’re seeing rising wage inequality, then chances are that not everyone is benefitting from economic growth. This is really very critical.  So, that leads to the next part of the presentation which is the patterns of rising wage inequality.  So, here’s a plot of the distribution of men’s wages.  And what leaps out at you is that for the median, that is the P50, the 50th

00:09:32     percentile and below, since 1979, real – that is inflation-adjusted – wages of men, have been flat to slightly down.  So, that middle and below, the bottom half of the male earnings distribution hasn’t really seen any advance in real wages over the last 32 years.  Now, on the other hand, you can see that those at the 90th and 95th percentile have seen quite large increases.  The value in 1979 is normalized to equal a hundred, so, all of these increases are percentage increases.  So, we’re seeing 30 to 42 percent increases

00:10:08     for the 90th and 95th percentile.  So, it looks like there are distinct groups of men in the whole bottom half of the earnings distribution who are not seeing any increases in their actual wages.  How about for women?  So, women are doing somewhat better certainly than men are in terms of their earnings.  Their earnings at least are growing.  So, the P50, the median, you can see about a 30 percent increase over this entire period.  The P90 and P95, even larger increases than we saw for men.  Ninety-fifth percentile for women, almost a doubling.  It’s 175 percent of its 1979 value.  So, if that line got up to 200, that would be an actual doubling of their real income at that percentile.  So, all but the lowest paid women, those identified by that 10th percentile, have seen pretty substantial increases in their earnings.  Now, one thing to be concerned about

00:11:09     though – so, I just showed you the same chart but with a little bit of additional information – You can see that for women at the median, their wages have been flat over the last 10 years or so.  So, those gains that they saw – the typical women in the labor market up until about the year 2000 – those gains have stalled.  What does this all mean?  Well, now just in terms of overall economic well-being, I’m going to shift back just briefly to household income. And so, this is a plot of median household income since 1967.  And

00:11:47     what you can see is that it was sort of flat in the ‘70s.  That was not a very good decade for incomes in the United States.  But then after that, it started to take off and rose quite a bit.  But you can see that following that rise up until about the year 1999 or 2000, since then, there’s actually been a slight downward trend in median household income – the income of the household right in the middle of the distribution which you might consider the paradigmatic example of the middle class in this country.  Now, that downward trend has been certainly accelerated by the recent recession.  I have recession bars in this chart so that you can see that in recessions, typically, household income does go down.  But there already seems to be a slight downward trend even before this recent recession.  So, that’s one sense in which the middle class is

00:12:40     getting a bit squeezed.  Prior to 2000 or so, middle class incomes were going up largely not because men’s earnings were going up on average, but largely because of increased labor force attachment and work hours by women.  As of 2000 or so, that factor has ebbed, and is not making nearly the same contribution, and household incomes are just not seeing the same movement that they had in previous decades.  The other thing to keep in mind – and this will be very, very important when I start talking about

00:13:15     explanations,  and Brad emphasized this point as well – is that the United States compared to many other advanced economies has a higher level of inequality.  So, the US is the red line in this plot.  This is the P90 to P10 ratio for household income closely related to what I showed you earlier in the table except this is adjusted for household size, which tends to pull down the gaps quite a bit because the lowest income households tend to be small on average.  Many of them are single, individual households.  And also adjusted for taxes and transfers.  So, the net impact of that pulls down the 90-10 differential quite a bit.  But what you can see is that the US line is well above the line for most of the – well, for the other OECD countries that I’ve listed here in Europe, and also in North America.  And the black line at the far right shows

00:14:09     you the OECD average,  and you can see we’re quite a bit above that.  We have lower inequality than Mexico and Chile, which are also in the OECD, suggesting that more advanced countries in general have lower inequality than do less advanced countries.  But relative to other advanced countries, we have fairly high inequality.  And it’s generally been rising faster in the United States than it has been in these other countries over the last two to three decades.  Again, that’s just sort of the overall wage picture

00:14:37     that compliments what Bradley was showing you with the top part of the distribution.  So, what are some explanations?  What might be the sources of this rising inequality?  That’s the final part of this talk.  And what I’m going to focus on are things that might be causing a middle-class squeeze.  Because, as you’ll see shortly, that’s really the key thing that’s going to pop out here.  Now, one thing that Brad already mentioned is technological change.  And there’s a very simple concept here which is that new information technologies, and the way they get used in the workplace, are most advantageous for highly skilled individuals: those with the college education, those who have specialized skills in using that information technology.  And that caused their wages to rise more rapidly than the wages of lesser-skilled individuals.  I want

00:15:29     you to keep that in mind because there’s going to be a nuance to that that we’re going to see in a second.  Another possible source is globalization.  That is, the widening of markets internationally, and the more intense competition that’s coming from countries like China and India with respect to the kinds of middle-class jobs that they’re creating over there.  More generally, the resulting decline in manufacturing jobs.  And then also institutions,  either legal institutions, or cultural norms, that in one way or another

00:16:02     guide or dictate pay practices.  Unions are one example of that.  I’ll show you a plot about that in a little bit,  and also pay practices more generally.  So, let’s talk first about skill-biased technological change which I’ll call SBTC for short.  There’s been a slight nudge to our conception of skill-biased technological change, which is called the polarization hypothesis.  And the polarization hypothesis basically states that new computer-related technologies tend to replace moderately skilled workers.  That is, workers who are doing repetitive, methodical tasks that could be easily mechanized, either in the services sector or the manufacturing sector – such as clerical work – they tend to get replaced or substituted for by the new technologies.  Whereas people at the bottom of the income distribution – janitors, security guards,

00:17:02     things like that – they can’t really be replaced by the new technologies.  So, the new technologies are tending to hit the middle class the hardest.  And you can see that in this chart.  Okay, so there are two components to this polarization hypothesis. One is that – so, the P90 over P50 in this chart is inequality in the top half of the distribution.  It’s exactly analogous to the P90 over P10, except I’m comparing the 90th percentile to the median.  And what you can see is that for men, this has been rising pretty

00:17:37     steadily since 1979.  And that reflects one of the overall impacts of the technological change, and its impact on the highly skilled workers.  But what’s been happening since the mid to late 1980s, so after an increase in lower-half inequality in the early to mid-1980s, since then, we’ve actually seen a decline in the wage gaps between the median and the 10th percentile.  This is really a squeeze on the middle class.  The notion that men who used to be the middle of the distribution and would see their wages rising over time relative to the bottom of the distribution, they’ve actually seen their wages declining in relative terms compared with people who are at the bottom of the distribution.  So, it’s not surprising that middle class families that rely heavily on male earners are feeling squeezed to some degree.  Now, the plot for

00:18:34     women, by the way, I should mention will look pretty similar to this except somewhat flat for the P50 over P10.  So, these forces have been affecting women less.  And in fact, there’s research suggesting that these new technologies aren’t replacing women in the middle of the distribution in the same way they’re replacing men.  Think about it to be a little bit stereotypical,  when we think of middle-class jobs for women historically, a lot of them have been nursing jobs, or teaching jobs.  Again, those are not jobs that are

00:19:08     readily replaced by technology.  In some cases, the technology may in fact enhance the work that those people are doing.  One of the net impacts of that over time, is you can see that the ratio – women’s to men’s wages – has been rising from about 65 percent in these data in 1979, up to almost 85 percent in 2011.  So, there is still a gap on average, but it’s been closed quite a bit.  So, it’s really been that the male side of the labor market that’s been affected most by these technologies.  The other thing – this is really striking, this is the great thing about having data through 2011.  We used to think about college-educated workers as being highly skilled and being in a good position to take advantage of new technologies.  What I’ve done here is I’ve divided up the college-educated into those with only a four-year college degree.  That’s

00:20:05     the red line.  And those with a graduate degree on top of that.  And what you can see is that since about 2000 – there’s that year popping up again – the return to a college degree relative to a high school degree has flattened out.  But what’s continued to rise over this entire time period, including in the last 10 years or so, is the return to a graduate degree.  So, now increasingly, it seems like to get your income above the median, a college degree isn’t enough.  you actually need a graduate degree,  and with the costs of college

00:20:38     and graduate school the way they are, that’s something that may be out of reach for a lot of middle-class families.  Now, there are some other factors at work here.  And I’m going to start to get a – I apologize – I’m going to get a little bit loosey-goosey because the literature is a little bit loosey-goosey about these issues as well.  Technology seems to be the main factor driving these changes,  but it then gets reflected in some other things we observe in the labor market.  One of them is the effect of globalization.  And related to that are changes in manufacturing activity in this country,  and also changes in union density and collective bargaining.  So, starting with globalization, for a long time the economics literature was finding that trade did not seem to affect inequality.  That, for example, industries that were most exposed to

00:21:28     rising imports were not seeing the biggest increases in inequality within those industries.  And so, the view was, well, trade probably isn’t a major contributor to rising inequality.  But more recent work has emphasized that we shouldn’t be looking at individual industries.  We should look at the workers who lose their jobs in industries that are subject to imports,  and follow them into to their new industries, and see what their wages are in those new industries.  And when you do that, what you find is that they

00:22:00     typically have much lower wages,  and it’s largely workers in that middle portion of the distribution that are getting affected by the growing manufacturing that’s occurring in countries like China and others, with the rise of the moderately skilled workers in those countries.  Then more generally we’ve seen for related reasons, a decline in middle-class jobs.  And that relates very closely to the decline of manufacturing, the decline of unions in this country.  So, I’m just showing you a long-term plot of the decline in the manufacturing share of total employment in the United States, and also the decline in union membership.  Manufacturing – Glenn actually mentioned this this morning – is actually a little bit under 10 percent of the workforce at this point.  It’s declined by about two-thirds relative to its level in 1973.  Union membership has

00:23:00     moved almost in lock-step.  Both of those have contributed to the erosion of middle class jobs.  Now, some people might argue, well, the decline in unions is just because manufacturing is disappearing.  But research suggests that in fact, the decline in unions was hastened to a large degree by some changes in legislation at the federal and state level in the 1980s.  So, that’s an institutional factor that’s also been contributing to the erosion of middle-class jobs.  The research is quite definitive on this point.

00:23:30     Compared with non-union workers, if you look at the decline in unionization over time, a big portion of the hollowing out of the middle – the undercutting of middle-class wages – can be attributed to that decline in unionization.  Now, one last factor I’d like to mention is changing pay practices.  And I should say – I’m going to mention this quickly because I think it’s more relevant for what Brad was talking about, about the top end of the distribution.  But there is recent research suggesting that rising reliance on bonuses and other forms of incentive pay – telling workers that if you perform in such and such a way, your earnings will go up, will be higher by a specific amount – that can actually account directly for a pretty big portion of that growing gap in the top half of the wage distribution.  So, that’s been a change that’s occurred.  Also,

00:24:26     there’s increasing competition for jobs at the top end.  There’s more ferocious competition, perhaps because international – now we’re competing with people all over the world in some sense for jobs.  Now, some of this may relate to technological change in the sense that if it’s easier to track performance because of new information technologies, it may also be easier to base pay on specific measures of performance.  So, it’s hard to separate these things out from technology.  But arguably, as Brad emphasized, a lot of

00:25:00     the technology is changing in other countries as well,  and we’re not seeing the same increase and inequality in some of those countries where the technology has changed in the same way.  So, there seems to be something cultural going on as well.  So, to summarize, it appears that middle-class workers are getting squeezed, and middle-class families, due to changing technology and the leveling of the global economic landscape.  Increasingly, education is important.  But even a college education is getting to the point where it’s not enough to see your incomes rise,  and maintain that sort of living standard that people expect.  Now, certainly globalization has contributed, there’s some inklings that maybe the adverse impact of globalization on the middle class may be easing up a little bit because costs are rising in China.  One of

00:25:53     the benefits of development in China is that wages are rising quite rapidly.  And there’s even been a little bit of shift back towards the United States in terms of manufacturing,  and that may help the middle class going forward,  but it’s far too early to tell whether that’s going to be a big impact.  And finally, what I’d like to emphasize as well, is that in addition to the changing technology and globalization, there have been some institutional and cultural factors related to the role of unions, related to decisions about pay

00:26:23     practices that are made in this country, that are affecting wages throughout the distribution, not just at the top of the distribution. That is basically the story I wanted to tell.  Thanks to all for listening.  I appreciate it.

00:26:42     [End of recording]

About the Speaker

Robert Valletta is a Research Advisor within the Microeconomic Research group at the Federal Reserve Bank of San Francisco. His research at the Bank has a dual focus on national labor markets and general economic activity in the 12th Federal Reserve District. In addition, his scholarly articles and ongoing research are primarily in the field of labor economics, including topics such as job mobility and job security, unemployment, income inequality and poverty, and the effects of employer-provided health insurance on labor market outcomes. Robert has also investigated the potential impacts of climate change on regional economic activity. See his research page.


Valletta, Robert and David Neumark. "Worker Skills and Job Quality." FRBSF Economic Letter (April 30, 2012).

Valletta, Robert and Katherine Kuang. "Why is Unemployment Duration So Long?" FRBSF Economic Letter (January 30, 2012).

Valletta, Robert and Katherine Kuang. "Recent Layoffs in a Fragile Labor Market." FRBSF Economic Letter (September 26, 2011).