No, because people do different types of jobs over time and the country’s demographic have changed.
We have more old people. More people in college. More people who have gone to college. Etc.
Now that doesn’t mean there can’t be some worrying trend going on, but digging it out requires a bit more subtlety.
Looks like the civilian noninstitutional population aged 16+ increased 25.57% Aug 97 to 17, but aged 25-54 increased only 7.55%.
That’s LNU000000-00 and -60
If you look at these two BLS graphs (one from 1990 to 2014, and the other from 2007 to 2017), it would appear that the tendency in the 21st century so far is for labor force participation to drop during recessions when the unemployment rate goes up (as you’d expect), but then to stay flat during times of expansion when the unemployment rate drops:
https://www.bls.gov/bls/cps_fact_sheets/lfp_mock.htm
https://data.bls.gov/timeseries/LNS11300000
If this pattern continues, it will be a kind of ratchet effect that may disguise the elimination of jobs. Unemployment goes up during the recession, comes down during the expansion. That part is normal as can be. But if the labor force trend continues, we could go from nonworkers being outnumbered more than two to one by workers at the beginning of the century, to their actually outnumbering the workers within forty or fifty years. My hunch though is that it won’t be linear and will be faster than that. But the net job losses still might well occur entirely during recessions, with periods of flat participation in between.
This pattern is in large part due to the demographic changes I just mentioned a couple posts back. Participation in the 25-54 group has actually risen over the past two years.
But if you look at smaller populations, e.g. prime-age males with only high school education, my prediction is you’ll find a trend that better fits the narrative of this thread.
Sure enough:
https://data.bls.gov/timeseries/LNS11300061
Obviously in the three minutes it took to find these I haven’t read them through yet.
I only looked at the BLS graph, but not only is it noisy, it doesn’t really go against the pattern I hypothesized. Overall since the recession ended, the numbers are flat at best, and maybe even negative.
IOW I was helping you make your point. The plot I linked is just men. The articles discuss men with less education. Overall we’ve seen an increase in core working-age workforce participationover the past two years by one percentage point – a steeper rise than before the recession. But that is not true for every subset. Men have been mostly flat over the past two years. Women have increased by about two points. And the articles point out that participation continues to decline for those with less education.
And of course since those are just rates, we’ve also been creating millions of jobs along the way. Just not, apparently, for everyone.
I’m not sure if we’re talking past each other, but when I said “overall since the recession ended, the numbers are flat at best, and maybe even negative”, I *was *talking about the BLS plot you shared. Sorry if the “overall” made that unclear.
But you keep talking about the “past two years”, whereas I’m talking about in between recessions, which for the most recent time period means “from the end of the last recession until now”. The sources I’ve seen date the technical end of the recession at June 2009 (when it hit its trough and began growing again), and the economy was back to its pre-recession peak in real GDP at some point in 2010, which is another way you could define a true “end”.
So even in terms of your latest graph, showing 25-54 year olds, the participation rate is at best the same now as it was in 2010, or maybe slightly lower.
Yet increasing faster than it ever has since the 80s.
Even that is arguable. If you zoom in close, it jumps up and down over the short term. But we sort of automatically smooth the curve in our minds, understanding that the short-term jaggedness is noise. But you see a longer term trend (“increasing faster”) whereas I’m not sure if that’s not just part of the bouncing around. Time will tell: if it rises to a point where it’s higher than it was in 2010, that’s somewhat significant–although less so if those gains get wiped out by the next recession. Most significant the other way would be if we at some point see a level that’s unquestionably lower than any we’ve yet seen in the 21st century (even if it recovers a bit after that). So right now it’s a bit ambiguous and I’d say the jury’s still out.
That’s not smoothing in anyone’s mind. It’s called a linear regression.
??? Why is that not smoothing?
It’s regression also, but looking at a graph on a weekly basis instead of daily, or monthly instead of daily, definitely smooths the graph. What’s wrong with calling it smoothing?
ETA: In sound-processing, it’s like using a “click-removal” tool, to delete anomalous spikes and dips.
The objection being to “in our minds”, as if the trend line over the past two years is a result of our minds desparately seeking patterns (which does happen) rather than a simple matter of babby’s first statistics. The LSRL over the past 24 months gives us an increase of 0.040 points per month, R[sup]2[/sup] = 0.83. Before this recent surge, the last time we saw that steep an increase was in the two years ending with April 1990. That isn’t “arguable”.
Now that doesn’t there aren’t any troubling trends re: workforce participation. It just means that “flat or decreasing” doesn’t apply to the core working age population over all recent multiyear periods. I wouldn’t even argue against “flat or decreasing or only increasing slightly (and not enough to offset the decreases).” We’re definitely down from the peak in Jan 1999. And certain sub-populations are down even more. And for a recession that supposedly ended years ago, it took an awful long time for the decrease to level off. And we’ll see how long this current surge can continue.
The last two sentences are basically my point Two year trends mean nothing if the next recession leaves us lower than we are now.
You wrote that a simple and true mathematical statement was “arguable”. And now you think that an unusually steep increase doesn’t matter. Please share more of your scholarship with us.
Something to keep in mind is the shift to e-commerce which moves retail store worker jobs to the fulfillment center.
DC labor to support distribution to retail stores is much less labor intensive than DC labor to support e-commerce fulfillment. This is due to the difference between an aggregated order for the entire store, vs an order for one or a few individual pieces to a consumer.
There are ways to build automation to chip away at the dc labor, so over time it will trend down, but there are a few tricky problems that make it a slow process to automate completely.
Sigh. :smack: To recap: I hypothesized that based on the 21st century data so far, labor participation may be dropping during recessions and then becoming flat during recoveries. I didn’t think I had to note that I meant the *net *was flat, since there are always fluctuations. You have been insisting that an increase (“unusually steep”) over the past two years falsifies this hypothesis, but I keep trying to point out that if you measure from the end of the last recession, there has been no increase at all, just a decline followed by an “unusually steep” increase, cancelling each other out in a “U” shape. Big whoop, get back to me when there’s a net recovery to numbers more like those in the '90s that is largely sustained after the next recession.
NPR reports on the bipartisan “AI Caucus” in Congress. I think it’s a good idea to have such a caucus, but they come across as a little bit Pollyanna-ish. Elon Musk is afraid of change? Really?
I just finished reading all of this thread over a few days after wandering onto this board from an unrelated Google search. Let me start by thanking you guys for keeping this going for 5 years and providing all the links! It’s a wealth of knowledge on the subject and interesting perspectives I haven’t found elsewhere, and I’ve been following this subject very closely for a decade.
I have to agree with most of the more pessimistic outlooks expressed; if something like a basic universal income is not implemented, it will spell disaster in the short to medium term as fewer and fewer workers are able to afford to buy products (or pay rent, for that matter, something I didn’t see addressed in much depth in this thread).
I don’t see it as a problem that will particularly affect the owners of the robot factories immediately. If the workers that were displaced in their home countries stop buying products, they simply find a market in another country to sell their nearly-zero production cost items to, and shift their sales there. Once that country automates, they shift to another market, and so on, until basically everywhere is fully automated and no longer has a consumer base to sell to.
But long before reaching that point, hundreds of millions have lost the basic ability to sustain themselves. In some regards, industrialized nations will actually suffer the worst of the effects because the vast majority of our population does not grow food, and does not own a home (or if they do, will lose it by getting foreclosed on when they can no longer pay a mortgage.) They may not be out on the streets the same day the paychecks stop coming, but they will be once their savings, if any, are exhausted and they can no longer pay their rent or mortgage.
We can already see the beginnings of this with the number of college students who are staying with their parents until later and later in life. They can’t find the well-paying jobs that they were seeking by getting a degree, that would allow them to move out and become self-sufficient. Automation requires fewer and fewer skilled workers in all sectors, and fewer hours required to achieve the same result, so the vast majority of jobs are part-time, no benefit jobs working for at or near minimum-wage. Meanwhile, rent and cost of living continue to skyrocket in most cities.
As to the capacity of the elite to ignore the millions who will end up on the streets; I am equally pessimistic. We already have 600,000 people homeless in the U.S. (not counting all those living with family, sleeping in cars, etc). When was the last time you heard a politician or news person even mention homeless people, much less suggesting solutions that would minimize the problem? Our capacity for ignoring the issue is far from being exhausted. I predict it will continue in that vein until long after it is a serious problem affecting the millions of people living paycheck-to-paycheck who simply can’t find another paycheck to pivot to when their current job is automated. Food insecurity is an issue we’ve addressed with food stamps, soup kitchens, food banks, etc. But rent is one where we haven’t even scratched the surface.
Finally, I’d point out that I am one of those millions who will be on the streets in the coming years/decades, when “worker” becomes a word mostly synonymous with “robot.” I currently work in an office in a position that can be best described as “A.I. Proofreader,” in that it is my job to comb through errors that crop up in the process of an algorithm automatically approving or denying millions of applications for health insurance.
When it finds something its deep learning algorithm has recognized is an issue, or human management has flagged to be addressed, that application is routed to me with the relevant error highlighted, so that I can pull up the computer-stored image of the application, review it, and decide on an action to take to correct the situation. This could be anything from correcting a data entry error the algorithm’s OCR made, to calling a customer to obtain missing information, to making a judgement call as to whether it should be approved based on the information available.
It should be fairly obvious that this situation as it stands won’t last forever. Every time I work with an error, I’m training the algorithm in the correct way to handle it. All it will require to fully automate all, or most, of the people doing what I do is slightly better and cheaper automation with the ability to read and “understand” scanned documents, plus “good-enough” robocalls to request information. All of which are marching forward from everything I’ve read.
I happen to be one of the lucky workers who still exist, albeit at a substantially lower wage versus inflation ($14/hr) than someone doing a similar job would have received a decade ago. For every application I review, there are a thousand more that were assessed by the software and automatically completed. It’s a prime example of the theme I saw repeated throughout this thread of automation not entirely replacing humans, just drastically decreasing the number of them needed and dramatically raising the amount of productivity.
By the time I’m made fully redundant, I expect that I’ll be completely unable to find another office job, and it will be a struggle to find any manual-labor job for minimum wage. While I have 10 years of experience in the health insurance field, it’s simply not economically viable to pay me a reasonable wage and benefits for that experience when we now have deep-learning algorithms that can accumulate thousands of years of far more comprehensive work experience, and immediately share it across the network. Paying me $14/hr makes no sense to a corporate board with a fiduciary responsibility to their shareholders to maximize profits. The moment they figure out how the tasks I’m doing now can be shifted to the algorithm that costs them cents per hour, they will.
Seeking a job in another field that doesn’t employ such automation will be relatively impossible for me given that I have a felony on my record and a poor credit rating. Employers already perform stringent checks for both for any kind of white-collar work, and some manual labor jobs as well. Job application requirements will no doubt become more stringent as fewer workers are needed, and demand for every open position swells.
I’m just one of many millions of people in the same position; about 1 in 3 American adults have some kind of criminal record, which is not surprising given we have 25% of the prisoners on the planet but only 5% of the world’s population. As time goes on, we can expect fewer and fewer people in that category to be able to find any job at all. Anyone without a degree, or with a poor credit score, will find their job prospects almost as abysmal, and only getting worse as automation continues to increase competition for jobs. One of these 3 impediments can be addressed, albeit mostly by taking on crushing student loan debt. But nothing can be done to remove a criminal record, and the paradox of a bad credit score is that it requires you to get a well-paying job to improve your credit; but good jobs require that you have a good credit score in order to get hired.