How to think about the claim by Justin Wolfers that “the income of the average American will double approximately every 39 years”?

Paul Campos forwards along this quote from University of Michigan economist Justin Wolfers:

The income of the average American will double approximately every 39 years. And so when my kids are my age, average income will be roughly double what it is today. Far from being fearful for my kids, I’m envious of the extraordinary riches their generation will enjoy.

I don’t know where to begin with this one! OK, let me begin with what Campos reports: “a quick glance at the government’s historical income tables shows me that the 20th percentile of household income is currently $30,000, while it was $24,000 39 years ago (constant dollars obvi) which is . . . far from doubling.”

This got me curious so I googled *government historical income tables*. Google isn’t entirely broken: the very first link was right here from the U.S. Census. . . . Scrolling down, it looks like we want “Table H-3. Mean Household Income Received by Each Fifth and Top 5 Percent,” which gives a convenient Excel file. Wolfers was writing about “the income of the average American,” which I guess is shorthand for the middle fifth of income. Household income for that category is recorded as $74,730 in 2022 and $55,390 (in 2022 dollars) in 1983, so . . . yeah, not doubling.

On the other hand, Wolfers is talking about his kids, and that’s a different story. They’re at the 99th percentile, not the 50th percentile. And the 99th percentile has done pretty well during the past 39 years! How well? I’m not quite sure. The Census page doesn’t have anything on the top 1%. They do have data on the top 5%, though. Average income in this group was 499,900 in 2022 and 230,600 (in 2022 dollars) . . . hey, that is pretty close to that doubling reported by Wolfers.

But this won’t quite work for Wolfers’s kids either, because regression to the mean. If you’re at the top 1%, your kids are likely to be lower on the relative income ladder than you. I’m sure Wolfers’s kids will do just fine. But maybe they won’t see a doubling of income, compared to what they’re growing up with.

OK, that’s household income. The Census also has a page with trends of family (rather than household) income. Let’s again go to the middle quantile, which is our closest to “the average American”: It’s $93,130 in 2022 and $65,280 (in 2022 dollars) in 1983. Again, not a doubling.

I did some searching and couldn’t find any Census tables for quantiles of individual income, so I guess maybe that’s what doubled in the past 39 years? I’m skeptical, but Wolfers is the economist, and I’m pretty sure his calculations are based on some hard numbers.

Beyond all that, though, there are two other things that bother me about Wolfers’s quote:

1. “Approximately every 39 years”: what kind of ridiculous hyper-precision is this? Incomes go up and down, there are booms and recessions, not to mention inflation and currency fluctuations. In what sense could you possibly make a statement with this sort of precision?

I’m reminded of the story about the tour guide who told people that the Grand Canyon was 5,000,036 years old. When asked how he came up with the number, the grizzled guide replied that, when he started in the job they told him the canyon was 5 million years old, and that was 36 years ago.

2. “The income of the average American will . . .”: Hey, you’re talkin bout the future here. Show some respect for uncertainty! Economists know about that, right? Do you want to preface that sentence with, “If current trends continue” or “According to our models” or . . . something like that?

I guess we can check back in 39 years.

I’m kinda surprised that an economist would write an article, even for a popular audience, that would imply that future income growth is a known quantity. Especially given that elsewhere he argues, or at least implies, that there are major economic consequences depending on which party is in power. If we don’t even know who’s gonna win the upcoming election, and that can affect the economy, how can we possibly know what will happen in the next 39 years?

And here’s another, where Wolfers reports, “For the first time in forever, real wage gains are going to those who need them most.” If something as important as this is happening “for the first time in forever,” then, again, how can you know what will happen four decades from now? All sorts of policy changes might occur, right?

That said, I agree with Campos that Wolfers’s column has value. It’s interesting to read the column in conjunction with the accompanying newspaper comments, as this gives some sense of the differences between averages and individual experiences. Wolfers’s column is all about how individuals can feel one way even as averages are going in a different direction, and that’s interesting. I will say that the even the comments that are negative about the economy are much less negative than you’d see in a recession. In a recession, you see comments about people losing everything; here, the comments are more along the lines of, “It’s supposed to be an economic boom, but we’re still just getting by.” But, sure, if there’s an average growth of 2%, say, then (a) 2% isn’t that much, especially if you have a child or you just bought a new house or car, and (b) not everybody’s gonna be at that +2%: this is the average and roughly half the people doing worse than that. The point is that most people are under economic constraints, and there are all sorts of things that will make people feel these constraints—including things like spending more money, which from an economic consumption standpoint is a plus, but also means you have less cash on hand.

So, lots of paradoxes here at the intersection of politics, economics, and psychology: some of the components of economic growth can make people feel strapped—if they’re focusing on resource constraints rather than consumption.

Aaaaand, the response!

I sent the above to Wolfers, who first shared this note that someone sent to him:

It seems like you were getting a lot of hate in the comments section from people who thought you had too positive a view of the economy—which seemed to just further your point: people feel like the economy is doing awful even though it really isn’t.

I agree. That’s one of the things I was trying to get at in my long paragraph above. People have this idea that a positive economy would imply that their economic constraints will go away—but even if there really is a 2% growth, that’s still only 2%, and you can easily see that 2% disappear cos you spent it on something. From the economist’s perspective, if you just spent $2000 on something nice, that’s an economic plus for you, but from the consumer’s point of view, spending that $2000 took away their financial cushion. The whole thing is confusing, and I think it reflects some interaction between averages, variations, and people’s imagination of what economic growth is supposed to imply for them.

Next came the measurement issue. Wolfers wrote:

But let’s get to the issue you raised, which is how to think about real income growth. Lemme start by being clear that my comment was about average incomes, rather than incomes at any given percentile. After all, if I’m trying to speak to all Americans, I should use an income measure that reflects all Americans. As you know, there are many different income concepts, but I wanted to: a) Use the simplest, most transparent measure; and b) broadest possible income concept; which was c) Not distorted either by changing household composition, or changing distribution. And so that led me to real GDP per capita. (Yes, you might be used to thinking of GDP as a measure of total output, but as you likely know, it’s also a measure of total income… This isn’t an equilibrium condition, but an accounting identity.)

My guess is that if you were trained as a macroeconomist, your starting point for long-run income growth trends would have been to look at GDP per capita.

And indeed, that’s where I started. There’s a stylized fact in macro—which I suspect was first popularized by Bob Lucas many moons ago—that the US economy seems to persistently grow at around 2% per year on a per capita basis, no matter what shocks hit the economy. I went back and updated the data, here, and it’s a shame that I didn’t have the space to include it. The red line shows the trend from regressing log(GDP per capita) on time, and it yields a coefficient of 0.018, which is the source for my claim that the economy tends to grow at this rate. (My numbers are comparable to—but a bit higher than—CBO’s long-term projections, which shows GDP per person growing at 1.3% from 2024-2054.) Then it’s just a bit of arithmetic to figure out that the time it takes to double is every 39 years.

You suggest that saying that it’ll double “approximately every 39 years,” is a bit too precise. I agree! I wish we had better language conventions here, and would love to hear your suggestion. For instance, I was raised to understand that saying inflation is 2-1/2 percent was a useful way of showing imprecision, relative to writing that it’s 2.5%. But we don’t have similar linguistic terms for whole numbers. I could have written “every 40 years,” but then any reader who understands compounding would have been confused as to why I wrote 40 when I meant 39. So we added the “approximately” to give a sense of uncertainty, while reporting the (admittedly rough) point estimate directly.

Let’s pan back to the bigger picture. Yes, there’s uncertainty about the growth of average real incomes. And while we could quibble about the likely growth rate of the US economy over the next several decades, I think that for nearly every reasonable scenario, I’m still going to end up thinking about my kids and being “envious of the extraordinary riches their generation will enjoy.” That’s the thing about compounding growth — it delivers really big gains! Moreover, I think this is a point that too few understand. After all, according to one recent survey, 72 percent of Americans believe that their children will be worse off financially than they are. If you think about historical rates of GDP growth, and the narrow corridor within which we’ve observed decade-to-decade growth rates, it’s almost implausible that this could happen, even if inequality were to rise!

I replied:

Regarding what you said in your column: you didn’t say “average income”; you specifically said, “The income of the average American.” And if you’re gonna be referring to “the average American,” I think it does refer to the 50% percentile or something like that.

Regarding the specifics, if the CBO’s long-term projections are 1.3% growth per year, then you get doubling after 54 years, not 39 years. So I guess you might say something like, “Current projections are for a doubling of GDP over the next 50 years or so.”

Justin responded:

1. I understand that there’s a meaning of “average” that incorporate many measures of central tendency (mean, median, mode, etc), but it’s also often used specifically to refer to a “mean.” See below, for the top google search. Given this, I’m not a fan of using the word “average” ever to refer to a “median” (unless there was some supporting verbiage to describe it more).

2. On 1.3% v. 1.8%: Even at 1.3% growth, in 39 years time, average income will be 65% higher. Point is, that’s “a lot” (as is 100% higher). Also, here’s the graph:

My summary

– If you say “average income,” I agree that this refers to the mean. If you say “average American” . . . well, that’s not clearly defined, as there is no such thing as the “average American,” but if you’re talking about income and you say “average American,” that does sound like the 50% percentile to me.

– I was giving Wolfers a hard time about the making a deterministic statement about the future, but, given the above graph, sure, I guess he has some justification!

– I think there is a slightly cleaner phrasing that would allows him to avoid overprecision and determinism: “Current projections are for a doubling of GDP over the next 50 years or so.” Or, “If trends of the past decades continue, we can expect the income of the average American to double by the end of the century.”

– Yes, I know that this is me being picky. But I’m a statistician: being picky is part of my job! You wouldn’t want it any other way.

84 thoughts on “How to think about the claim by Justin Wolfers that “the income of the average American will double approximately every 39 years”?”

  1. Maybe he meant mean income? Which doesn’t correspond to any percentile. But of course an increase in mean income could come about by a large increase in a single individual’s income (or the income of a very small group)–which is not very useful for the “average” (i.e., typical) family.

    At least he specified he meant real income in the middle of your post. With annual inflation at 2-3%, a doubling of nominal income every 40 years is pretty much guaranteed, so such a prediction would be really boring.

    • Your point about nominal vs real is very illuminating. As an economist myself, I never even considered the possibility that he would be referring to nominal income here. To my ears, everything in the original quote is trivial and unobjectionable, which is why I find Andrew’s response to be so jarring. What Wolfers and I are missing is that the jargon and shorthand of economics often conflict with the actual definitions of the words being used.

  2. Unless I’m really mistaken, it is entirely possible for mean income to increase in the way Wolfers claims it will, but that 72% of current children will end up financially worse off that their parents. This is where I very much agree with Andrew and think that using mean in talking about both “income” and “American”, ascribing mean income to some “average” American is pretty shoddy by Wolfers. This is pretty clearly not an accurate representation how of income is distributed. Any time any economist talks about a “macro identity” or an “accounting identity” I think it’s reasonable to read this as “we have made a definition of this term that may be entirely independent of important underlying structure of what we purportedly describing”.

    • Yes, and I think this is the key point that is distorting the analysis. GDP per capita is an ensemble average, ie a single slice in time and over some population. Because the distribution of income (and wealth) is shifting, GDP per capita (as an ensemble average) will diverge from the median (it will also diverge from almost every individual’s trajectory in the long run). With worsening income inequality, Wolfers’s argument doesn’t hold any weight.

  3. Cause and effect is tricky and far be it for me to take things personally, but I noticed that the blue line on the graph (real GDP per capita) takes a sharp dip exactly at the year in which I was born (mid 1930s). It did take a while, but by the time I left grad school (early 1960s), the USA managed to get ahead of the (straight line red) curve and stay there. And, the greatest positive vertical distance between the blue line and the red line seems to be when I retired. This summer, I am slated to have minor surgery and hope this does not effect the blue line too much.

    • If I remember correctly his calculation would say that if you bought a stock for $10 and it paid $1 (inflation-adjusted) in dividends every year for a century before going bankrupt the return of that stock is -100%.

        • And if you put the dividends into an envelope looking to leave it to your grand-kids you end up with $550 and a calculation that is just as accurate tells you that the return is +5450%.

          So maybe the problem would be _your_ reinvesting. If you’re buying more shares from someone else you’re saving them from sharing your fate. The -100% return stock calculation may be “accurate” for you but may not be a universal truth.

          The fact that a stock ends up worthless doesn’t mean that it was a bad investment: in aggregate it may have distributed much more money to shareholders than it ever took from them.

        • Carlos, I 100% agree with you. The explicit point that seems to be missing is that no one should invest in a small number of stocks with constant dividend reinvestments. People should take their excess income and invest it in diversified stocks and bonds and reinvest dividends in a continuously rediversified portfolio until they need to take income out to spend it.

          Also people should invest their wealth in their children and grandchildren as they grow rather than trying to leave everything to them when they’re adult.

  4. Before people swarm in about whose incomes are rising and by how much, and how we measure price changes, the one thing in Wolfers’ statements that bothers me the most is the unabashed optimism about future growth. I know that is a prevailing view – at least among many economists – but it seems to make huge implicit assumptions about climate change, global security, future health, and civil society. While evidence can be provided for optimism in all the dimensions, there are plenty of counter arguments as well. And some of these dimensions are well beyond Wolfers’ expertise, so we have a respected “expert” on some matters expressing his opinion about a wide range of things without clearly specifying which are opinions and which he has the background for his opinion to be “better informed.” Where have we seen such behavior before?

    • That is what I was thinking. Climate change is already driving migrations, which is driving xenophobia and right-wing populist movements here and in Europe. This is only going to get worse. The Guardian recently published a poll of climate scientists, asking what global warming they expected: 80% said 2.5C or more: https://www.theguardian.com/environment/article/2024/may/08/world-scientists-climate-failure-survey-global-temperature. That will take more than 39 years, but still.

      On the other hand, climate change will cause more severe weather which means more property destruction which means spending money to rebuild which means a higher GDP; I’m nor sure that is good.

        • B-b-b-ut . . . the Freakonomics people told us that “The PDO cool mode has replaced the warm mode in the Pacific Ocean, virtually assuring us of about 30 years of global cooling.”

          Who to believe, huh?

        • “FWIW, there is evidence we have already hit 1.7 degrees Celsius of warming:”

          That’s so strange! I still seem to be living!?? Today I checked the market: people were trading stocks as if the planet isn’t even burning! No smoke outside either…huh. It’s floodin’ down in Texas, but SRV was singing about that in the 80s….I guess not much has changed.

          Well, anyway, we know economic growth ended in 1973, we died of starvation in 1980, the world imploded from the Y2K bug in 2000, and peak oil has come twice in the last 50 years. I’m not sure how I’m using this computer when we’ve been out of energy since Carter. And no doubt, the biodiversity crisis that started 35K years ago with the death of the Australian megafauna is going to drop the hatchet on us tomorrow. It only **LOOKS** like humans have been thriving during that time! Really that’s just a cognigive illusion to which only academic scientists are immune! Doom is waiting for us somwhere! And now adding insult to injury, we have FB and TWTR responsible for the mental health crisis of our children.

          And we’re all horribly obese from eating high cholestorol eggs!! My sister is very obese but thankfully her nutrition advisor has helped her lose zero pounds over the last X^Y years. She kindly provided me the many pages of explanation of nutrition given to her by her advisor, but I foolishly took the dangerous and risky step of eating less without consulting a nutrition advisor! Oh my dear, I’m still deeply worried that I’ll be harmed by my inappropriate balance of transfats even though I lost 50lbs. It’s dangerous to eat without your own consulting scientist!!

          Oh Dear! Only statisticians and climate scientists and nutrition advisors can save us!!

      • If an asteroid big enough to knock modern civilization back to the 18th century appeared out of deep space and buzzed the Earth-Moon system, the near-miss would be instant worldwide headline news.

        Two years ago, Earth experienced a close shave just as perilous, but most newspapers didn’t mention it. The “impactor” was an extreme solar storm, the most powerful in as much as 150+ years.

        […]

        By extrapolating the frequency of ordinary storms to the extreme, he calculated the odds that a Carrington-class storm would hit Earth in the next ten years.

        The answer: 12%.

        “Initially, I was quite surprised that the odds were so high, but the statistics appear to be correct,” says Riley. “It is a sobering figure.”

        https://science.nasa.gov/science-news/science-at-nasa/2014/23jul_superstorm/

        Those Mayan calendar doomsday predictions were actually only off by 9 days. And even one winter of such outages would probably trigger a mass migration back south.

        For some reason the news focuses on tiny gradual changes when theres evidence all over that we need to prepare for sudden catastrophic events. This blind spot seems to have developed in the 1800s to deny anything like Noah’s flood could have happened.

        • Anon:

          The Noah’s flood point is interesting. I remember reading a book many years ago about the history of catastrophism and gradualism in geology, and some of that came up.

        • The Noah’s flood point is interesting. I remember reading a book many years ago about the history of catastrophism and gradualism in geology, and some of that came up.

          To me, its like “I get it, they really needed to dismiss the authority figures of the day”. But now its time to move on and interpret the evidence with less bias.

    • The EMH says there’s evidence for optimism in lots of speculative prices. If you expect the world to go to hell in a handbasket, no sense in financial investments long into the future, cash out now. If you expect the polar icecaps to melt and raise up the sea level, then coastal land (admittedly much less fungible than securities on an exchange) should go down in value while inland territory could become beachfront. The posted graph contains the actual Civil War and Great Depression, so it seems robust to some quite large shocks.

  5. Even 50th percentile is a tricky concept when you’re dealing with a group whose composition changes over time. It’s not like they take a randomly selected group of Americans and track their incomes over time and report back the 50th percentile of that group.

  6. When Mr Wolfer writes about the accounting identity of GDP, he does not mention the extremely large measurement errors in GDP. Using different methods (three in all, one for each of the accounting identities) we get results that can vary by up to a third. You do not need to be a statistician to know that this is extremely imprecise. The uncertainty over the 39 years is huge.
    To make matters worse, US GDP forecasters usually compare their predictions with random walk models. If US GDP follows any kind of random walk (i.e. no regression to the trend), the system is unstable: The first doubling could be in thirty years and the second doubling fifty years later.
    Nevertheless, I think he is right in predicting greater prosperity for most people in the Western world in the future.

    • “results that can vary by up to a third”
      Can you provide a source for that? My textbook knowledge (many years ago) was that the different approaches gave fairly close estimates, at least in the US. I did find this paper which alludes to a difference of 1% or less, although with some indication that the errors are larger at some times and perhaps some indication they were rising (https://www.elibrary.imf.org/view/journals/001/1997/123/article-A001-en.xml). Your comment sounds like you are referring to the errors from different measurement approaches (for example, basing GDP on total expenditures or total income). There are other types of measurement error, such as those resulting from revision of estimates. That has also been studies and I believe show much smaller estimates of error, at least for the US. So, your statement of “upt to a third” sounds very high to me. Can you clarify what exactly you are referring to and possible sources for that?

      • Also very interested in this clarification. I could CERTAINLY believe that year over year growth estimates could vary by a third of a percent… But the value of GDP per capita being somewhere between 66k and 100k would be rather shocking

      • Maybe by GDP he means changes in GDP – like it happened with CPI the other day.

        By the way, one does not need to be a statistician either to know that measuring daily changes in weight using a bathroom scale is extremely imprecise but that doesn’t mean that the uncertainty over a whole year is huge..

      • I tried to trace my source. It was from the introductory macroeconomics slides for my first semester of undergraduate studies. I think I must have deleted it. That means no adequate source, sorry. Now I feel like the boy who cried wolf…

        • I haven’t taught introductory macro for 50 years. In fact, I was denied promotion at one time for refusing to teach it. It’s heartening to hear that I left the teaching of that subject in such good hands.

  7. What a huge crock of shit! This is unironically demonstrating my core thesis which is that macro-economists especially serve the purpose of cheerleading policies that make rich people richer to give a “scientistic” veneer to elites doing what elites always have done.

    “It’s like Lenin says, you look for the person who will benefit… and …. uh…. uhh… you know”
    — The Dude

    People have already come in and said a number of good things already. For example:

    1) the average can increase by just increasing the amount of the top few percent even if the bottom 99% decline
    2) Even percentiles don’t match the path of individuals. What people care about are their personal trajectories
    3) Conditions globally in terms of climate and exploitable resources and technological growth in physical technology (engines, agriculture) in 1850 or 1950 were very different than what we expect they will be in 2050

    I’m not so sure why we should care so much about 1820-1950 that’s a super different world than today. 1950 marks an approximate breaking point from the past, because it’s shortly after WWII a conflict that was unprecedented in the modern world, and also is about the onset of modern agriculture with tractors, fertilizer, and engineered crops (through concerted scientific breeding).

    I’ve got lots to say about what Dale mentioned about how we measure purchasing power and soforth. But I’d like to just focus on the part Dale focused on which is the future of things like the environment and energy consumption.

    According to Our World In Data https://ourworldindata.org/explorers/energy?tab=chart&facet=none&Total+or+Breakdown=Total&Energy+or+Electricity=Electricity+only&Metric=Annual+generation&country=USA~CHN~OWID_WRL

    The world consumed 10,000 TWh per year in 1985 and consumes about 29,500TWh/yr in 2023 The graph looks pretty exponential, so let’s just figure this means growing at exp(0.041*t) where t is in years. Let’s extrapolate that out to 100 years (say the end of your kids lifetime if they were born today)… Thats 1780 TWh/yr

    Now, estimate on average that solar energy of 1kW per square meter falls on the equator at noon, the radius of the earth is around 6400km… Consumption in 100 years is what fraction of the total incoming energy?

    julia> 1780 / (pi*6.4e6^2 * 1000*24*365 / 1e12)
    1.5790870629078693e-6

    Or 1.6 parts per million. Considering that we need the sun to fuel plant growth and we need tons of plants per person per year to fuel the humans, and that only 1/4 of the earth is land, and that half the world is in darkness at any one time and … etc etc This is already a lot, let’s extrapolate to 300 years

    julia> exp(.041*300)*29.5/(pi*6.4e6^2 * 1000*24*365 / 1e12)
    0.005749501830227719

    In 300 years at the current growth rate we’ll be using about 0.6% of the entire input of the sun.

    So, yeah, growth stops, and it is guaranteed to stop on the scale of the length of time that the US has been a country so far.

    Now let’s look at NOMINAL GDP per capita since 1950, on a log scale

    https://fred.stlouisfed.org/graph/?g=1nYn0

    There’s a very obvious super-linear growth prior to 1980, and a very obvious sub-linear growth since 1980. And this is **nominal**, we need to subtract by an inflation measure which is also growing. If we divide by CPI all items and plot change from a year ago we get something with a lot of noise, but if you run a spline through it, it’s clearly maybe 2.5 to 3 % per year in 1960 while today it is **almost zero already**, maybe 0.5 to 1.0 at most (I’m too lazy to load it up into Julia and run a loess through it, but feel free if you’ve got the motivation. My feeling is growth today at 0.5 to 1 percent is reasonable to estimate and that at the rate of decline over the last say 15 years you might expect it to hit 0 in another 10 years)

    https://fred.stlouisfed.org/graph/?g=1nYnn

    Now, let’s look at one of the core effects of Income Inequality, which has definitely been increasing strongly in the US:

    https://data.worldbank.org/indicator/SI.POV.GINI?locations=US

    Here is log of homicide rate vs world-bank Gini, with some centroids of notable countries plotted as their 3 letter symbol:

    https://postimg.cc/R6kHVXzS

    There’s a pretty clear mechanism for this, basically poor people surrounded by crushing extractive rich people get pissed off and start committing crime and violence. When I showed this to one economist who posts here he said something like “Holy shit that’s the strongest signal I’ve ever seen in any social science data of any kind”.

    So, perhaps if somehow countries massively lower their GINI coefficient while growing we can expect to have growth without incredible levels of crime and conflict, but I dunno, both BRA and MEX have lowered their GINI through time, but large portions of Brazil are flooded and have lost their fresh water dams, and Mexico city is about to run out of water entirely. The US has been on an upward trend in Gini and unsurprisingly also a plausibly exponential growth in violence since 2014. And this is just homicide at an individual level, not genocide at state levels which seems to be on the uptick the last few years.

    https://www.msn.com/en-us/weather/topstories/brazil-flooding-will-take-weeks-to-subside-experts-warn/ar-BB1mqhY0
    https://www.cnn.com/2024/02/25/climate/mexico-city-water-crisis-climate-intl/index.html
    https://ourworldindata.org/grapher/homicide-rate-unodc?tab=chart&country=OWID_WRL~Europe+%28UN%29~USA

    If I were teaching an undergrad course in welfare economics and asked for a midterm exam essay on the future of human welfare, and a 2nd year undergrad turned in what this Wolfers turned in I would give them a D+ like Peppermint Patty in Peanuts. Like, at least you cited some data, but your thesis is simplistic and supported by only one extrapolation from a period of time well before any “carrying capacity” considerations, and completely fails to consider either physical limitations of growth or the distribution of outcomes or alternative lines of evidence or concerns other than GDP/capita.

    • Countries differ in a lot of ways other than income inequality (my understanding is that a huge driver of differences between countries is connections to US drug markets). In order to determine how reliable the relationship is, one would want to look at it over time. You’ve said income inequality has grown since 1980. Has that resulted in the homicide rate increasing since then?

      “unsurprisingly also a plausibly exponential growth in violence since 2014”
      There was an increase starting in 2014 (after a long drop starting around 1994), but it started going down again around 2016. It shot up again in late May of 2020, but has been going down more recently.

    • All excellent points, but the crock-shittery that bugs me at least as much is that GDP is defined in terms of market value, which is pretty much meaningless in terms of quality of life. Yeah, sure, quality of life is difficult (maybe impossible) to quantify, and who can argue with the reality of the market? Well, I can.

      First of all, if quality of life is impossible to quantify then how-the-kids-will-be-doing-in-x-years literature must be quantitatively meaningless, so we can just give that whole topic a rest.

      As to the reality of the market, well, my observation is that a large portion of the changes in our expenditures over time are reflections of corporate gamesmanship rather than life-enhancing consumption. Consider short-lived and non-repairable (or specialist-only-repairable) appliances, cars, HVAC systems, and computers. These used to be DIY-approachable (or handyman-approachable if you’re not a DIYer), and (except for computers) they used to last long enough that repair was worthwhile. Or think about the rise of subscription services replacing ownership of software, music, and movies. Or think about housing scarcity, where prices can double with no change in a house’s characteristics. Or…

      Insofar as they result in more consumer spending, these factors all count towards increased GDP, right? One could argue, “well, how much of the GDP do these factors really represent?” I don’t know the answer to that, but the answer is probably something small (surely someone has studied this, but I don’t know that literature). More importantly though, that’s the wrong question. The pertinent question is, “how big are these factors’ annual growth relative to 2% of the GDP?” Probably not so small now. If the growth of nonsense costs exceeds 2% of GDP, then there’s something to the oft-ridiculed perception that material quality of life sometimes feels like its in decline.

      • Preach my brother! :-)

        Ok, so your excellent point has brought up another of my bugaboos…

        The quantity known as GDP/Capita can’t possibly matter at all, and we know this ahead of time. That’s how unscientific Economics is, that it hasn’t even caught up to 1800’s ideas of dimensional analysis. https://en.wikipedia.org/wiki/Buckingham_%CF%80_theorem the Buckingham Pi theorem was proven in 1878 according to that article.

        GDP has dimensions of Money / Time, so by itself it’s meaningless. “Capita” has dimensions of “People”. What could make this all meaningful? One thing would be a quantity in terms of Goods&Services / Money, but now we’ve got 4 dimensions and 3 variables! So we need something like Goods&Services/Time. But we’ve also got a few other dimensions of interest, Mass and Length. For example, you need a certain amount of food energy per day (~2000 kcal/day), and you need a certain volume of shelter (length^3) and a certain mass of protein per day (Mass) and a certain quantity of energy to heat/cool your house (Mass * Length^2/Time^2)…

        So let’s find some dimensionless groups that are meaningful for people’s quality of life… I can’t be expected to do all the work! but here’s one of interest:

        (GDP/Population) / (Avg Food Energy Price (Money/Energy) * Dietary Requirements (Energy/Person/Time) )

        Finally we can *begin* to discuss a scientific question.

        Interestingly of course we can discuss this quantity at different levels of quality. It’s one thing to eat beans and rice and nothing else, but it’s another to eat beans and rice and marinaded chicken with spices and steamed carrots in mole sauce.

        And we can address quality through a number of variables, one is a ratio of two masses, say (Mass of Protein / Mass of Food), and we can also look at Mass of Vegetables / Mass of Food…

        Now that we’ve got a scientific idea of what we’re doing in terms of nutrition, let’s address living space… We can look at for example Area Occupied / Person / Area Occupied by the circle a person’s outstretched arms spun around… That’s gonna be close to Square Footage Of Living Space / (pi * Height^2/4). We can add up the height^2 of all the occupants of a household, and look at the crowding ratio and then look at the distribution of this ratio.

        Ok, next let’s discuss cooling of the living space. Let’s look at GDP/Capita / (Avg Cooling Power in the Summer * Price of Energy * Occupants of a house)

        and the distribution of that statistic…

        Jesus Christ people, get with the 19th century!

        • Note that by dealing with dimensionless ratios it immediately fixes some of the problems you mentioned too. For example a clothes washing machine that needs replacement more often…

          Gdp/capita / (price of a washing machine / lifetime_of_washing_machine / (People/Washing machine in a household))

          Yep, the numerator is money / person / time and the denominator is money / person / time… rearrange…

          GDP * Lifetime_of_Washing_Machine * (People/Washing machine in household) / (Population * Price Of Machine)

          So now, if the lifetime goes down all else equal… the income measure goes down. We are worse off as a society!

          But if we can share washing machines among more people so people/machine goes up… the income measure goes up! And if the price of the machine goes up, the income measure goes down! And if GDP stays constant but population goes up… income measure goes down! Amazing F****ING SCIENCE!

      • About GDP and market values. It is basic economics 101 to point out the numerous ways in which GDP is a flawed measure of economic well-being, and even more, overall well-being. Unless you live in Bhutan (where national happiness is measured), we don’t have any better measures, despite its many flaws. I would point out, however, that while GDP measures the total market values of final goods and services produced, it is also a measure of the total income earned in the economy. Earning income is also based on market values. Now you can reject that as a meaningful measure, but then I think you are ignoring the reality of people’s lives. Many, if not most, people seem to think their income is an important measure of something. The fact that their income is linked to the market values of goods and services that are produced, means that if those market values are meaningless then so are their incomes.

        I am personally willing to accept that all of our monetary values are flawed – perhaps so deeply flawed as to be meaningless. But what do you propose in its place? I’ve always been disturbed by economic measures of “value” – it is one reason I stopped teaching economics. But it is hard to take the extreme position that these measures are totally meaningless. There have been proposals for a labor theory of value and a thermodynamic theory of value, but they have their own issues. Doing away with markets poses the question: what would you replace them with?

        • Dale, GDP isn’t entirely meaningless it’s just entirely meaningless by itself. Until it’s combined into a dimensionless ratio it just doesn’t have meaning yet.

          I give an example of a dimensionless ratio of GDP/capita to various stats about washing machines… That dimensionless ratio is inherently meaningful. It measures average income to the income needed to carry out laundry… We need a at least about 5-6 such ratios to understand households (likely one about food, one about housing, one about climate control and utilities, one about transportstion, one about personal healthcare, one about childcare and education, one about entertainment and communications, one about construction and maintenance costs of dwellings…) We need at least 5-6 such ratios to understand manufacturing, we need at least 5-6 such ratios to understand healthcare provisioning….

          By the time we are done 20-30 such ratios should give us a usable aggregate description of what’s going on in the economy.

          Until Economists understand this it will not be a scientific discipline.

          Nothing which has dimensions matters (fundamentally this is because the numerical value of the thing changes depending on what you choose as your measurement units, which is not true for dimensionless ratios)

  8. Ok, I figured let’s also take a look at the part about how we measure the purchasing power of goods… We’ve had some convos here in the last week about different kinds of indexes. One I advocated for was equal-weighted index on important categories. Equal weighted because I think for you to thrive you need some baseline fixed quantity of all of these, you can’t be thriving and just consume a lot of cheap food and education while living under a bridge. And I want to expand those important categories from the baseline ones I was mentioning to include a few more because I think we’re talking about not just scraping by here, we’re talking about whether our children will be thriving.

    Let’s take GDP/capita and measure it relative to an equal-weighted CPI average including rent + food at home + transportation + education&childcare + healthcare

    I think we can all agree that to enjoy a good standard of living we need to assume our kids can consume at least some baseline level of all of these, and that they are all nontrivial components. Like we’re not consuming only 1% of income on education or on transportation or anything like that each of these is at least 10% of income averaged over the first say 30 years of life. Rent maybe vs mortgage payments or inherited houses is a bit of a complication, but not for a growing society that will need to make new houses (and hence not inherit them).

    So let’s plot year over year percentage growth of GDP/capita vs this measure:

    https://fred.stlouisfed.org/graph/?g=1nYuR

    And what we see is this is already a random walk around 0% and has been the entire time of the data set since 1980. Basically there has been ZERO growth in “ability of a person making the GDP/capita to sustain a better standard of middle class living” for my entire lifetime.

    Here’s the value of the statistic plotted on a log scale (not the change rate, just the gdp/capita/inflation_index on a log scale)

    https://fred.stlouisfed.org/graph/?g=1nYvi

    If you were a second year welfare economics student and you handed in Wolfer’s thesis on your midterm and I showed you that graph and asked you why you hadn’t even looked at something like that, I think you should feel humiliated.

      • Exactly my point. All of these should have nontrivial weights (ie. 10-40% ish). In some sense, I’d like to use something like a Dirichlet(5,5,5,5,5) prior for the weights and then look at GDP/capita/weighted_sum and see what an ensemble of those curves looks like… but I don’t have the time to put that together right now.

        • It’s exactly your point, that’s what the L stands for. Keeping up with your « middle class living » standard requires higher spending in childcare/education than food and housing combined.

        • You’re saying that today Education&Childcare makes up 35% of my equal weighted graph as of today? Maybe, I didn’t look at the breakdown this time, and unfortunately it actually depends on what fred uses as the base date, which I normally select as 2019-01-01 (Pre Covid) so it should be closer to 20% but in this case I think I didn’t remember to select that. I find Fred’s tedious clicking annoying and don’t always remember to reset the base date.

          Dale, yes we’re looking at a family of “Wolfers like people and their next generation” so this index is not just about base living it’s about living while affording to invest well in two children basically. When we average across the entire population, not everyone has kids at home so the weights are very different for the population avg.

          I’ll re-do the index with 2019-01-01 (pre covid) as the base index.

          https://fred.stlouisfed.org/graph/?g=1nYKP

          This has a moderate growth rate about 1%/yr easier to see in the cumulative curve:

          https://fred.stlouisfed.org/graph/?g=1nYMS

          Let’s examine the weighting within those 5 categories for a family in Los Angeles area who is middle to upper income like the Wolfers presumably are

          Let’s say a family of two parents, two kids is our target for this index (we’re talking about how “great” the kids will be doing 40 years from now, so presumably a 4 person family today headed by parents with young kids vs a 4 person family 40 years from now headed by the kids with young grand-kids).

          Preschool for 2 kids in Los Angeles area is easily $4000/mo, rent is $5000/mo for a 3 bedroom house, food at home is say $1500/mo, Transportation is probably $300/mo in car payments for 2 cars, plus 6 tanks of gas a month, plus $2000/yr in insurance so let’s call it $1200/mo, health insurance for that family is likely $600/mo plus actual health care is likely another $200/mo on avg so $800

          Just adding those up let’s see what that means: $150k/yr after taxes.

          GDP/capita is 84103 according to Fred, if the two parents are working at that income they’re making $168k/yr before taxes, let’s say they’re paying 35% in taxes (federal, FICA, state, and state disability) after taxes they’re taking home $110 so they’re spending $40k/yr of savings which is pretty reasonable for a family of 2 with 2 kids who need childcare, that’s like about what my family lost per year for the first 6 years of my kids life (kids are really expensive).

          So reasonable weights for that family restricted to just those quantities is 32% to childcare and education 40% to rent, 12% to food, 10% to transport and 6.4% for healthcare though if there were an unfortunate healthcare related issue the healthcare could easily double.

          So let’s use those weights:

          https://fred.stlouisfed.org/graph/?g=1nYMW

          The story doesn’t change ~ 1% per year over 30 years, you can play with the fred “units” at the bottom of the edit-graph dialog as well.

          So, yeah, whatever there’s a bunch of uncertainty in what the proper weighting should be because things vary from one family to another which is one of the key points. As I say, the way I’d actually argue to construct that whole index if I were doing it in Julia instead of clicking Fred’s website is to normalize all the CPI values to 2019 levels (because COVID and after was pretty weird) then construct a Dirichlet([5,5,5,5,5]) distribution as a prior (basically making each weight a random variable mainly concentrated in the range of 0.05 to 0.5 with 0.2 as the most likely value, and then plotting the ensemble of curves using the ensemble of normalizers… the point being to emphasize how the importance varies from family to family and that each curve is plausibly some outcome for a family, many of those curves may stay flat or decline though of course some would probably go up.

        • Dang, I really wish I had read this far down in the thread before posting my comment about nonsense expenses in response to Daniel’s previous post! Oh, well.

          Daniel’s back-of-envelope calculations from the FRED data show that rising cost of essentials takes a huge bite out of the portion of GDP growth that could potentially be chalked up to life improvement. A careful accounting of rising nonsense costs would take another bite!

        • I should be careful what I wish for: My earlier comment appears not to have posted after all!

          My main point was that corporate gamesmanship has led to lots of consumer expenses that are counted as GDP growth but have nothing to do with material life improvement. For example, cars, appliances, and HVAC equipment have shorter lives now than they used to, and they are generally no longer repairable by a DIYer or handyman, which leaves consumers with more frequent replacements and more expensive repairs from certified specialists. Another example is the rise in subscription services for things that we used to own (software, music, movies, books). Also nonsense rent-seeking services like turbotax. These don’t add up to as big a factor as changes to education or rent/mortgage costs, which Daniel’s analysis looked at, but they’re probably a few thousand dollars a year.

        • Josh
          My impression has been that automobiles last longer than they used to. A very quick search yielded https://www.bts.gov/content/average-age-automobiles-and-trucks-operation-united-states. There is also plenty of anecdotal evidence that they last longer and also have fewer problems. Your point about them being difficult for DIY fixes is absolutely true, however. Also, I am unsure about appliances. The old Maytag commercial of the repairman with nothing to do certainly doesn’t match what my experience has been lately. But, on the other hand, most appliances are considerably cheaper than they used to be. Televisions are a prime example – large screen TVs are generally much cheaper than the old cathode ray tube TVs used to be. I’m not sure about their longevity.

          I have to say I am skeptical of both your examples and Daniel’s assertions. Many cities have subsidized childcare (preschool). Someone with the average income in the US probably does not qualify for the subsidies, but a considerable number do (e.g. in Alaska, anybody with household income below 85% of the statewide median is eligible). College tuition is another ambiguous area – list tuition prices have been rising much more rapidly than inflation, but not net tuition. There are undoubtedly stories of college education expenses eating up large portions of income and requiring large debts. But there are also stories of these debts being paid back quickly (greatly depending on what you study).

          More to my point in today’s blog post, the “state of the economy” is not easily established. Yes, many economists state that it is very good. Lots of the comments over the past few days have stated it is terrible. There is evidence on both sides of the argument. So, when we see the public reaction, I am not sure if the measurements are out of whack with reality or if the perceptions are very misguided.

        • Dale… see my post on Economists finally partying like it’s 1899 and beginning to use dimensionless ratios, which is the point at which Economics will *begin* to be a science.

          https://statmodeling.stat.columbia.edu/2024/05/15/the-income-of-the-average-american-will-double-approximately-every-39-years-who-says-that-sort-of-thing-show-some-respect-for-uncertainty-dude/#comment-2372423

          Your points are all real valid ones, and it turns out they’re all related directly to the fact that people are pre-scientifically trying to understand the world without forming dimensionless ratios… I give an example of washing machines, all the considerations you mention are present, Price, longevity, number of people who share the machine etc.

          We can also include operational cost in the same kind of ratio, instead of just purchase price, include purchase price + cost of operation over the lifetime. If purchase price goes up but cost of operation goes down more than compensating, then income measure vs durable goods and cleaning goes UP.

          As for childcare transfers, those just wind up creating a poverty trap. Earn a little more money, become ineligible for childcare, take home less! My own mother had to reject multiple raises when I was a kid because she would have lost money on them. It’s another example of terrible systemic problems in the economy that absolutely affect people and they *feel them* even though the economists don’t see them in the dollars here and the gdp this and that and the unemployment whatchacallit numbers.

        • Daniel
          Are you really rejecting the notion of childcare subsidies, or are you objecting to the way the policy is implemented? If the latter, then I think most economists would agree with you. Very few economists support policies with eligibility cliffs. But I think it is wrong to label childcare subsidies as a “poverty trap.” If we ever get a UBI, that too may seem like a poverty trap. In fact, many public policies are ill-designed. But most of the fault, I believe, lies in our political process not at the hands of economists. If I were to fault economists, it is for recommending too few policies, not too many. It is too easy to point out the ways that policies can go wrong, thereby ignoring the existence of the problems they would address.

        • Dale. I reject any policy which creates a negative slope or cliff in the consumption capacity vs earned income curve.

          This is PRECISELY why I advocate for flat tax +UBI. Flat tax leads to a constant slope of consumption vs earned income and Ubi raises the line so it doesn’t reach zero at zero earned income.

          This is one of the most important reasons for flat tax + UBI together.

          When it comes to helping children I advocate that each person on their 1st birthday receives an account at the Fed with M2/capita in it and that the Fed has no other way to increase the money supply than to deposit an equal quantity of base money into each living persons Fed account simultaneously.

          So yes I’m against the way we do childcare subsidy and all that sort of stuff with the caveat that it’s because I specifically advocate specific alternatives.

      • Carlos, Daniel
        Can you tell me what data you are referring to? When I go to the CES data (https://www.bls.gov/cex/tables/calendar-year/aggregate-group-share/cu-income-deciles-before-taxes-2022.pdf) I see aggregate spending on education of about $117 Billion and total income after taxes of around $11 Trillion (multiplying the average income after taxes by the number of consumer units) , for around 1.6% of the total. The table also shows the breakdown of total education spending by income decile: 6.1% is the share of the lowest decile, gradually rising through the deciles, then rising rapidly to 41.9% of the total being spent by the top decile. So, where are you getting your ideas from? Or are you referring to Wolfers’ children (since he is in the top decile)?

        • I’m referring to the “measure” in Daniel’s charts. If middle class living meant spending 20% of your income in education four decades ago when it was cheap you really need to spend much more now to maintain that standard. Education is now 35% of that measure and healthcare 25%.

        • We really have apples and oranges here because the numbers are not close. Daniel’s numbers have too many steps for me – the CES data reports spending in dollars and it isn’t close to matching Daniel’s derived numbers. I don’t believe that middle class people spend an average of 35% of their income on education.

        • Dale, I have a post that details it out and utilizes alternative weightings, it’s being held for approval.

          We’re talking about families with say young kids today compared to those young kids families 40 years from now… We’re also talking about families where say each parent works and earns GDP/capita which is 84k each before taxes.

          The CES survey averages across all people in the US. If you can find families of 2 parents and 2 kids with incomes in the range say 2*GDP/capita +- 20% you’ll see very different weights than the whole-country avg.

        • For example, https://mybrightwheel.com/search/l/guides/cost-guide-los-angeles-ca-daycares-preschools#how-much-does-infant-daycare-cost-in-los-angeles suggests that if you have 2 preschool kids (0-5 years) and you need 5 days a week full time care, you’ll be spending on average $1500/mo for EACH of them in the LA area, or $3000 a month total, I put $4k in the index example calculation that will hopefully be approved soon, but that’s well within the range of plausible.

          https://www.apartments.com/rent-market-trends/los-angeles-ca/ says you’ll be paying $4000/mo for a 3 bedroom apartment, I clicked and looked at the map and found a 3 bed 2 bath 1300 sqft home in Highland park for $4450 it was nothing special (5249 Hub st, los angeles CA)

          By the time you work it out, a 2 parent 2 child family in the LA area is going to be spending after-taxes easily ~30% on childcare, 40% on rent, 10% on each of transport and food and healthcare is not unreasonable.

        • I have to say I’m not fond of this idea of creating a stylized family (which no doubt represents some real families), then analyzing those circumstances, and then making conclusions that appear more general than they are. It is a form of cherry-picking, though I realize it does describe real circumstances that face real people. The averages, of course, represent no real people themselves. But the averages do represent something – so when your stylized families look very different than the averages, it says to me that there are other families that also look very different (but in opposite ways).

        • The original quote suggests to me:

          “The income of the average American will double approximately every 39 years. And so when my kids are my age, average income will be roughly double what it is today. Far from being fearful for my kids, I’m envious of the extraordinary riches their generation will enjoy.”

          1) So we’re talking “average” levels of income ie. GDP/capita is a resonable income for the parents
          2) There are some kids involved, a typical number for a family is 2
          3) We care about what happened to that group of people vs what the world would look like for a similar group of people 40 years later (ie. “my kids”)

          Averaging across “everyone” is useless. There are households with 70 year old retirees, there are households with 19 year old single mothers there are households with 48 year old women with no partner who have no desire to have kids… all these people are going to make very different choices about what they spend their money on… you’ll be averaging across massively different expense distributions, the expense distribution won’t look anything like **any** of them. This is basic high-dimensional-space type of stuff often discussed in MCMC sampling… the overall average is actually far from any typical draw from the distribution.

          The real problem is *looking at broad averages* and expecting them to tell you something about the average behavior of a relevant specific group of people. That’s the whole point I’m trying to make, which is looked at for families with average income per capita of the parents and two kids… growth of income was meager 1% per year for the last 40 years kind of stuff.

        • So please analyze what happens to the 70 year old retiree with grown kids and grandkids who receive full scholarships to a good university and lives in a house that they owe nothing on. And, what about the minimum wage restaurant worker whose employer has instituted a surcharge to pay for employee benefits, and who works in a city that increased their minimum wage, and where there are unfilled openings, who received a rent subsidy courtesy of their local government, and who lives in a state where childcare subsidies pay almost all of the tuition for preschool.

          I didn’t think we wanted to analyze the reasonableness of Wolfers’ particular statement for him personally. If that is the issue, then we need to know more about his particular family – perhaps his kids will, in fact, receive full-pay scholarships.

          I certainly don’t want to defend Wolfers’ statement. I find it offensive for a number of reasons. But I also find statements that 35% of income is spend on education to be of no value.

        • Dale, sure it should generalize beyond Wolfers in particular, but I don’t see the relevance of analyzing anyone at all who doesn’t have kids… since we’re asking how well their kids are doing relative to them… and I don’t see the relevance of asking questions about anyone who is far from “gdp/capita” in income since we’re talking some kind of middle class-ish family. And of course some people will win the lottery and some people will get full ride scholarships and all that, but if we pick numbers near typical for the costs of things, 35% on education/childcare for someone with young kids today is pretty much spot on… I quote websites above: https://statmodeling.stat.columbia.edu/2024/05/15/the-income-of-the-average-american-will-double-approximately-every-39-years-who-says-that-sort-of-thing-show-some-respect-for-uncertainty-dude/#comment-2372303

          and I also re-analyze the same index with 20% weight to each item as of 2019-01-01 and re-analyze the same index with the estimated weights from a basic back of the envelope budget based on real world expenses for a family of 4 in that above linked post… those are both basically the same conclusion… 1% per year growth for families

          The 70 year old who had kids 40 years ago is just the same person moved backwards in time 40 years…

        • Let’s put it another way that maybe makes the mistake in analysis that I think economists tend to make more clear.

          Let’s take a time machine 40 years in the future and take a big sample of 40-year-olds at that time and ask them how they’re doing relative to how their parents were doing when they were young.

          When they think back to their parents’ experience, 100% of those people we ask in the future will be thinking of people who are having Young children right now. The experience of people right now who don’t have children is irrelevant. none of it matters for answering the question we ask the 40-year-olds of the future, because none of the people who have no children are the parents of people 40 years in the future.

          So when we construct a CPI that averages over everything everyone today spends it just doesn’t address the question of how well people are doing 40 years in the future compared to their parents 40 years previously because it averages over a huge number of people who shouldn’t be in the average.

          I’ve heard people comment on this in terms of so-called “nonergodicity economics” which is kind of a buzz word I guess among some heterodox economists. But whether I buy into those heterodox economists entire thesis I don’t know. all I am definitely saying is that when analyzing people in the future and comparing them to their parents, people who never have children are completely irrelevant.

    • And I’d hate to be the leftover fifth of a kid in that family with 2.2 children! I guess they have to get together with the other one-fifth kids and try to make a go of it. :( We need a program to help these poor children!

  9. For the way a group of Berkeley economists look at it, go to https://realtimeinequality.org/, where you can find interactive graphs that show the distribution of various measures of income, as well as wealth. Looking at their “average real disposable income,” the income of the bottom half of the distribution has gone from $17k to $25.8k since3 1976. For the top 10%, the numbers are $106.5k to $297.2k. It only gets worse as you go up the distribution. So, I guess it depends on which economist you ask.

  10. Andrew: It’s not the same people in any given percentile bracket at any given time. The “average American worker” moves through several brackets during their working life, probably typically starting in the 5fth (lowest) or 4th and winding up in the 3rd or 2nd or even 1st (highest) at retirement – when, presumably, most people will backslide one or two brackets.

    See the graphic here:
    https://manageyourmonies.com/index.php/2018/11/10/average-income-by-age/

    Also, modern retirement accounts (e.g., IRA/401K) did not emerge until the mid 1980s and have presumably changed the income landscape subsantially since:

    1) people reduce their income by tens of thousands of dollars using 401Ks and IRAs
    2) this income is then available later in life. Prior to the 1980s, it’s likely that:
    3) *most* americans had little or no retirement income other than SS, so
    4) the effect of IRA/401K is to distribute income over a longer time and increase total income
    5) for obvious reasons many people use IRAs and 401ks to depress their taxable income tot he maximum extent possible.

    Beyond all that, since the 1960s the US has had a unprecedented inflow of poorly educated people from 3rd world countries. Immigration as a percentage of the population is way up, and Europeans as a percentage of immigrants is way down, with most being less educated and from 3rd world countries:

    1960-9: 85% the 320K immigrants/year were from Europe or Canada
    2010-2019: 87% of 1063K immigrants / year were from Latin America (50%) and Asia (28%).

    You all may not like that poorly educated immigrants earn low-end wages, but *they* love it becasue they’re earning much more than they would have in their home countries. That’s why they come! In effect, most immigrants get a big wage increase just by showing up. Where does that show up in the historical wages in the US??

    So, Andrew et al, if you want to do some **real** stats instead of using misleading aggregate data without much thought, why not back adjust “income” in a way that adjusts American incomes downward in the past for what would have been the historical earnings of immigrants in their home countries?

    • Also since the availability of investments to the public has increased massively over the last 50yrs, it follows that more educated people would be able to leverage that availability more readily, thus increasing their income, while less educated people – immigrants in particular – probably can’t leverage investing nearly as well. Thus, even though many immigrants save, their savings likely earn very little, which acts to supress their income relative to the rest of society.

    • Chipmunk:

      Agreed about people moving up and down the income distribution during their lives. I was thinking about this when writing the above post but didn’t bring it up because it seemed to me that Wolfers was making a comparison between generations, rather than making a comparison within individuals over time.

      • Andrew: that’s good to hear! Just the same, most people in any given generation also move up and down the income brackets. Seems like the relevant consideration is lifetime “income” – e.g., all forms of “gains” including capital gains, retirement income, pension income, beneficial government services (e.g., Medicare / medicaid) etc – not income bracket or average income at any point in a person’s life.

    • What you conveniently forgot to mention is that 401Ks and such replaced security with insecurity/casino investments, susceptible to all sorts of manipulations, only a few insiders have access to. The no. of jobs with available guaranteed pensions (in addition to SS) has probably gone down.

      Regarding the immigration to US since ’60s, that doesn’t add up either. This country is nothing but a handful of generations of immigrants. Some like to pretend that there was some gap/pause in immigration, but that’s an illusion really. There were slow periods, but this country has always needed fresh muscle (not brains) to build the wealth for a handful of those who happen to have immigrated a little bit earlier.

      Regarding some of your other comments on education, please see figure 3:
      https://nces.ed.gov/pubs2019/2019179/index.asp

      The world is more educated than it was in the ’60s by definition, so by extension, on average people immigrating now should be more educated. On average, at least.

      There are very few countries, including some very unstable and poor ones, that would or should trade their basic education system with the US.
      It doesn’t get much worse than what we have.
      Again I’m talking mode, not median or mean here.

  11. Using the “income of the average American” as a shorthand for “average income of an American” is *tremendously misleading. If incomes follow a bimodal distribution it could be that essentially no American has an income close to the average income of all Americans.

    Using the mean as the “best proxy” for the full distribution is reasonable if the data is known to come from a unimodal distribution of a particularly nice form but is completely misleading when the data is known to come from a heavy tailed assymmetric distribution …

    It’s ridiculous that such data is not accompanied at least by data showing the evolution of the variance (provided such data is available, which it surely is).

  12. So this paper popped up when I was looking for relevant data

    https://www.weforum.org/agenda/2020/09/social-mobility-upwards-decline-usa-us-america-economics/

    https://opportunityinsights.org/paper/the-fading-american-dream/

    The graph on the first link is pretty stark. This research group had access to tax records where they could match children and parents. It seems they used standard CPI all items to adjust incomes, but they show that people born in 1980 whose parents had 50%tile incomes have ~45% chance of making more than their parents CPI adjusted. Compared to people born in 50s to mid income parents who had ~80% chance.

    So that pretty much blows the Wolfers bullshit out of the water. Not only did people born in the 80s not experience 2% per year growth relative to their parents, even at the median outcome, in fact MORE THAN HALF OF THEM DECLINED.

    • “Compared to people born in 50s to mid income parents…”

      I love this comparison.

      You mean this: before the deflation of the post-war wage bubble caused by the WWII destruction of the world industrial capacity, people had a higher chance of making more than their parents. However, after the rest of the world began to compete with the US again, that chance – for groups born in two very carefully picked years – declined. Yes and if you invested in MSFT in 1999 you did poorly but if you invested in MSFT any other time in history you did pretty well. This is not news.

      Also see my comments on tax deferred retirement accounts, the income of immigrants (what is their chance of making more than their parents?:). Also, does your paper benefits? Health care benefits are worth more than ever event hough they don’t count as income. And note that in the period under discussion, Fed spending on medicare and medicaid has increased wildly, as has spending on many other governement services across the spectrum of county, city, state and Fed govt. Your “income” claims might be overlooking that?
      https://chiro.org/BLOG_Graphics/HC_SPENDING.jpg

      Your point doesn’t “blow wolfer’s bullshit out of the water”. It just undermines your credibility. Like most data people you think you can use stats in a knowledge-free vacuum and find some kind of truth. Hey! I found this data on the internet! I declare it means XYZ!! WCGW??? That’s the bullshit. Any given column or groups of columns of data are worthless without thorough knowlege of **all** the processes that generate *all* the *relevant* information – whether that information is codified into rows and columns or not, and whether it is adequately represented in any model or not. That’s why statistical and data anaylsis is appropriately viewed as nothing but sophisticated cherry picking until the work is verified multiple times with multple different forms of analysis and demonstrated to be consistent.

      IMO Wolfers’ statement strikes me as offhand and likely inaccurate claim on a specific level. Just the same, the broader idea that things are getting better for most people in the US is pretty clear if you look at *all* the available information and not just the info that suits your claim, and your and many others’ incessant cherry picking of a few circumstances here and there that you claim indicate otherwise is where the gist of the bullshit lies.

      • Did you even click the link and look at the graph? Pretty sure you didn’t.

        In EVERY decade since 1940 despite the “fact” that real gdp per capita increased exponentially at 2% per year continuously since 1820 according to Wolfers’ graph, when you match people born in that decade to their **actual individual parents** using **actual tax records**, and inflation adjust their income using CPI-all-items… IN EVERY DECADE the probability that a child born to a median parent did better than their parents when they were 30 declined, until by the 1980’s it was only about 45% such that MORE THAN HALF of people born in 1980 did worse than their parents by age 30. Not worse than the expected 1.8x increase you’d get from 2% growth for 30 years… literally WORSE, or in other words NEGATIVE growth.

        Explain that away with some stuff about “data analysts” and “lies” and “cherry picking” and “greatest country on earth” and “miracle of capitalism” or whatever verbal masturbation you like. Individual microdata shows you’re full of shit.

        • > MORE THAN HALF of people born in 1980 did worse than their parents by age 30.

          Depends on the definition of “doing worse”, I guess. If over 5% of those 30-year-olds born in 1980 have zero income – probably because they are maintained by their parents – somehow they are “doing better” than their parents who had to work to maintain them already when the parents were 30-year-olds themselves.

          Even if the those born in 1980 maintain themselves a large share of them will be single at thirty – while their parents were very likely to be a couple. In those cases it seems quite a high bar to compare the income of one person with the joint income of two people.

        • Carlos,

          In fact reading into the second link a bit it appears those graphs are done by comparing male children individual incomes to their fathers individual incomes. So it’s not households with 1 vs 2 people that’s the cause. Of course, this is somewhat problematic because in the 1970’s women started working a lot more than they had in the 50’s etc so by the 2010’s when you’re looking at 1980 children more women are working and the male to male comparison is interfered with with this second order trend. I’d love to have access to IRS detailed data even if it were on an air-gapped computer and I give them my code and they return the report… but I’m not likely to get that any time soon, so I can’t really do anything but point at the paper.

          The paper does mention that they tried multiple ways of accounting for inflation and multiple types of specifications for their models, and the take home message was consistent. Each decade since WWII we’ve made it so less and less of GDP growth goes to families until family income growth went NEGATIVE for all but people whose parents were below ~ 25%tile. How do you go from an **expectation of 1.8x income growth after 30 years** which Wolfer’s implies due somehow to GDP growth at 2%/yr to “more than half of children of the middle class make less than their parents” while those born in every previous decade did better each decade back.

          You have to really screw up your country’s wealth distribution to turn 1.8x expectation into 0.9x actual or whatever.

        • > In fact reading into the second link a bit it appears those graphs are done by comparing male children individual incomes to their fathers individual incomes.

          Wrong.

          The chart is for the main analysis where they “measure absolute mobility by comparing children’s household incomes at age 30 (adjusted for inflation using the Consumer Price Index) with their parents’ household incomes at age 30.”

          “Notes: This figure plots the fraction of children earning more than their parents […] Children’s income is measured at age 30 in the CPS-ASEC as the sum of individual and spousal income, excluding immigrants after 1994. Parental income is measured in the Census as the sum of the spouses’ incomes for families in which the highest earner is between age 25-35.”

        • Apparently you’re right about the graph, but also they say “These findings are unaffected by using alternative price indices to adjust for inflation, accounting for taxes and transfers, measuring income at later ages, and adjusting for changes in household size.”

          then they go on to say: “The decline in absolute mobility is especially steep – from 95% for children born in 1940 to 41% for children born in 1984 – when we compare the sons’ earnings to their fathers’ earnings.” which is what I was referring to when I thought that was how they made the graph. Basically suggests if they did male children vs fathers the graph would look even worse.

          So, in any case, it appears that the general trend is independent of most of the confusing factors such as household size, women vs men in the workplace, changes in the tax code, transfer payments, and measures of inflation. I hardly am going to claim to have read and understood and checked the whole thing but on the surface, it looks as if people who have detailed access to government records of many many **individuals and their parents** for 80 years of comprehensive data say… children have consistently done less well than you’d expect from 2% per year growth and in more recent generations have had **negative** growth.

          So in this discussion we have Wolfers saying “when my kids are my age, average income will be roughly double what it is today. Far from being fearful for my kids, I’m envious of the extraordinary riches their generation will enjoy.”, and then an Economist in the comments saying “To my ears, everything in the original quote is trivial and unobjectionable” and then actual data analysis by Economists with detailed access to data says exactly the opposite has been going on for Wolfers’ entire lifetime.

          This is kind of what triggers me about Economists, and I’m going to admit to being maybe unreasonably irritated by this stuff, but basically if you find an Economist in the news or online who is talking about general issues with human welfare and policy and soforth and it’s not them describing some specific data analysis they personally did, then they are likely to be saying “Cheerleading” type stuff based on really simplistic models of the way the world works using inappropriate averages (Like CPI all items to compare children to their parents, when CPI is averaged across households ~75% of which are childless and using GDP per capita growth and applying it to everyday people without any consideration of income distribution) and based on extrapolations that are ridiculously provably wrong even without data (such as growth “forever”) and it’s entirely likely that some Economist who has looked in detail at the data actually has said at some point something precisely contradicting the cheerleading stuff.

          It’s so common that it’s really striking when you see something else going on, like https://www.politico.com/news/magazine/2021/12/28/inflation-interest-rates-thomas-hoenig-federal-reserve-526177

          A couple of times a week in the last 3 months though you’ll see Economists saying stuff like “the economy is doing really well and individuals rating it not so good are just out of touch with the reality”… Who’s out of touch with the reality now Wolfers?

        • > independent of most of the confusing factors such as household size

          With their “adjustment” for household size (dividing by household size) it’s not true that “MORE THAN HALF OF THEM DECLINED”. The percentage of 30-year-olds with higher (adjusted) income than their 25-35 parents is 60%.

          They could have looked at the half of 30-years old who had children – and compare this subgroup with their parents – but that’s not what they did. I don’t know if that would have raised or lowered that 60%.

  13. The census numbers are garbage by the way. They are CPI adjusted, and CPI is known to overstate (and therefore over-deflate) numbers over time, painting a falsely pessimistic picture. We’re not talking about small difference either since they compound over time. Here’s a simple example to illustrate: take CPI adjusted wage real wage growth, un-adjust it, and then re-adjust it using PCE instead, which is a far better measure than CPI.

    https://fred.stlouisfed.org/graph/?g=19U9Q

    Using CPI instead of PCE since 1993 *halves* supposed real wage growth by 2024.

    I’m not throwing shade at anyone in particular, but when you see someone deflating prices over time using CPI, either they have a pessimistic outcome in mind that they’re trying to find (which I find to be unethical in research terms), or they just don’t have enough domain expertise to know any better. But in either case, the results should be obviously looked at with even more skepticism than usual when someone is making a clear methodological mistake.

    • Just to elaborate (correct me if you see it differently), the over-estimate of price changes in the CPI is largely due to ignoring substitution effects: households tend to substitute relatively cheaper goods and services for ones that rise in price. So, their actual expenditures (represented in the PCE) don’t rise as much as the CPI would indicate. Despite this well-known issue, attempts to replace the CPI in many policy programs (e.g. social security) have not been successful. I would view that as largely driven by politics, although you could make a case for saying that the PCE is also an imperfect measure. I suspect that an attempt to measure the “true” cost of living will remain controversial, since any such attempt must be some sort of average, and individuals and subgroups will experience different price changes. So, I’m not sure I agree that anybody that uses the CPI either has “a pessimistic outcome in mind” or that they “don’t have enough domain expertise to know any better.” I think there is enough room for ambiguity that someone could continue to use the CPI without such motivations, simply because it is the most used measure by many people. I view it as analogous to the use of NHST – people continue to use it despite its many flaws. Increasingly, I move away from it – almost to the point of omitting it when I am teaching introductory statistics – but I wouldn’t go so far as to say anybody using NHST has such ulterior motives.

      I would recommend that any analysis using the cost of living over time should consider alternative measures of price changes and report those results rather than choosing any single measure.

      • “I think there is enough room for ambiguity that someone could continue to use the CPI without such motivations, simply because it is the most used measure by many people.”

        Ugh, you just gave me academia chills. I heard this a lot when I was in grad school –justifying something only because it is “standard” in the literature.

        There’s plenty of reason to use PCE instead of CPI, rather than using them both. There is for example no reason (yet…?) to believe that PCE either overstates or understates inflation. Bullard make the case a decade ago. Even aside from bias concerns alone, it’s also more comprehensive in what it covers (both urban and rural consumers, beyond out-of-pocket spending, more), and how it covers them (more frequent weight adjustments).

        https://www.stlouisfed.org/publications/regional-economist/july-2013/cpi-vs-pce-inflation–choosing-a-standard-measure

        CPI is just bad for long-term price adjustment, there are better alternatives for that task, and there isn’t much justification to continued use in that context beyond “it’s commonly used.” CPI can have some other uses: if you’re doing a macro model and you’re looking at period-to-period changes, then the flaws aren’t going to compound over time.

        • The study you link to finds a 7% difference in the price index (using PCE instead of CPI) over an 18 year period – hardly enough to shake the world. I agree that merely using things because they are “standard” or “commonly used” is a poor practice. But when alternatives fail to gain a foothold, despite amply evidence they are superior, perhaps there are reasons. One reason I mentioned NHST is that similar things can be said about that – we should know better. But plenty of people still conduct NHST without nefarious motivations. I won’t defend using either the CPI or NHST as representing “truth” or the “correct” analysis, but I do think there is room for them to continue to exist.

          If the issue is consumer substitution, then I think both measures have issues. The CPI does change the market basket occasionally, but not enough to really capture the ability to substitute relatively cheaper goods for more expensive ones. The PCE does not account for the possibility that such substitution entails some loss in value or quality. For example, I may substitute a cheaper university for a more expensive one, but the quality of education may not be the same.

          This is not an area I keep up with research in – so please correct me if these examples are off-target.

  14. Perhaps his kids were too noisy in the back seat when he was writing the piece

    Btw he never really apologized with Borjas about his dishonesty. I don’t trust the guy and this interaction confirms it

Leave a Reply

Your email address will not be published. Required fields are marked *