Poor sales MINUS quality of labour

Update 3: I wish I had titled this post: "Labour quality: the dog that stopped barking".

I see something a bit different in that graph that all the macro-bloggers have been blogging about. Sure, I see the "poor sales" that everyone else sees. But what about "quality of labour"? What's important is that "quality of labour" is not seen as the "single most important problem" by more than a very small number of small US businesses right now. It's usually, in normal times, seen as a problem by a lot more businesses. That tells us something important.

I see booms as times when it's easier to sell stuff and harder to buy stuff. And recessions are times when it is harder to sell stuff and easier to buy stuff. By "stuff" I mean both goods and labour. Forget the distinction between the goods market and the labour market. The important distinction is between money and everything else. It's money that is hard to "buy" in a recession; everything else is easy to buy.

So firms are telling us that it's hard to sell goods and it's easy to buy good quality labour. Yep. That sounds like an excess demand for money and an excess supply of everything else. Standard Keyneso-monetarist recession. Not structural unemployment, where firms can't find the right sort of labour.

If I were technically competent I would figure out some way of playing with that graph and subtract "quality of labour" from "poor sales" to show you what I mean. But I'm not. Which is why I have never been able to make it as a good empirical economist.

Update: Daniel Kuehn picks up on this post and makes explicit a point I really should have emphasised more, and repeated. This data goes right against the recalculation story (or what I call "structural unemployment"). If recalculation was the reason for increased US unemployment, more firms should be saying that finding good quality labour is their most important problem. But we see the exact opposite. In other words, at the margin, recalculation looks less of a problem now than it usually is. (Though, if we solved the deficiency of demand problem, it might of course become a problem; which is why I said "at the margin".)

Update 2: Niklas Blanchard has found a much better chart to show what I wanted to show, and coloured it in to make it clearer. It's all much clearer now. (How come other people can do such things??)

 

Comments

[Adam P]

Nick, while I of course agree with you diagnosis I'm not sure this graph actually speaks to it.

The question asked about the "single most important problem" and so just because sales demand, as a problem, has been elevated to number one status need not imply it's gotten any easier to buy good quality labour.

[Nick Rowe]

Adam; yes. Damn. It doesn't *necessarily* mean it's gotten easier to buy good quality labour.

Hang on though, that argument also works the other way around. Maybe it's gotten so much easier to hire good labour, so that poor sales have been elevated to number one problem? Not that sales have gotten worse, in any absolute sense. Or, something else has gotten better.

Another problem, I was thinking about after I wrote this, is that it does not make sense to simply take poor sales minus good quality. Because all the things don't then add up to 100%, of course. But I do want to "add" those two categories together, because both are indices of demand deficiency. It's just that one is a positive index and the other is a negative index.

My interpretation *fits* the data, but is not, strictly, *implied* by the data. But nor is anyone else's, strictly, I think. We can't really say that "sales have gotten worse". But that's as well as we can usually do in economics, anyway.

[Adam P]

"Hang on though, that argument also works the other way around."

true, but that was the point. Your data doesn't allow you to sort that out.

On the other hand, can't we imply from sales volume data that sales have gotten worse?

[Nick Rowe]

We know from volume data that sales have fallen. But sales identically= purchases. Maybe purchases have gotten harder? (Obviously not, just from reading the news, but a determined economist could always argue it's a supply-side shock that caused sales to fall.)

[Adam P]

some do argue that :)

Surely, though purchases gettng harder shows up as (is defined by?) inflation.

[Phil Koop]

"Because all the things don't then add up to 100%, of course."

I'm not clear on the statistic you are trying to create. Is the goal to quantify the importance of demand deficiency relative to all other complaints? Or to quantify the degree to which either demand or supply is deficient, ignoring taxes, regulations etc?

[Nick Rowe]

Yes, though perhaps not *by definition*. This is how I would say it: if sales=purchases drop because of a fall in demand, then *either* there's excess supply, so it gets easier to buy and harder to sell, *or* prices fall (relative to trend, or expectation?), *or* (more usually) a bit of both.

What bugs me about macroeconomics data is that we don't really get hard data on how hard or easy it is to buy and sell stuff (goods and labour). All we get is data on prices and quantities. Not that I have any solution to this problem. So I listen to news stories and other anecdotes.

[Nick Rowe]

Phil: "I'm not clear on the statistic you are trying to create."

That's because I'm not clear on it either! I keep flipping back and forth in my mind between your two options.

[Nick Rowe]

Phil: "hard to find good quality labour" is a measure of boom. "Poor sales" is a measure of recession. So they are negatively correlated. Somehow I want to invert one index, the labour one, so it becomes a measure of recession, then add it to the other one, poor sales, to get a combined index of recession. But I can't figure out how to "invert" it. I perhaps don't literally mean 1/index. Maybe I should mean 1-index. Or some constant C-index. Maybe C should be the maximum value of the index in the data set?
Dunno.

[Daniel Kuehn]

Daniel Kuehn actually - Evan's my brother and occassionally blogs, but I do most of the econ posting.

[Nick Rowe]

Sorry Daniel! I will correct the post.

[Andy Harless]

Almost all the variation at business cycle frequencies comes from these two particular categories (as is clear from the chart that Niklas Blanchard found). It would be difficult to reconcile this observation with a "relativistic" interpretation. If the problem were only in sales (weak during recessions) or only in labor (difficult to obtain during booms), then we would expect the other categories to change as well (as, for example, weak sales might start to displace taxes as a primary complaint during recessions, at the same time that it was displacing labor availability). The fact that the other categories are fairly flat at the relevant frequencies suggests that both sales and labor availability are varying in absolute terms.

[Phil Koop]

Nick, I just meant that I was wondering about your normalization constraint. If you are focusing on the boom/recession indicators, it seems more natural to use the ratio than the arithmetic difference, regardless of how you transform the quality of labour. If you must have a signed quantity, subtract 1 from the ratio or take the log, if it varies enough. What do you have against:

Recession Index = (% responding poor sales / % responding poor labour quality) - 1

Or more naturally:

Expansion Index = (% responding poor labour quality / % responding poor sales) - 1

[Nick Rowe]

Andy: I really think you are onto something there. I had to read your comment twice to try to get my head around it, and while I understand what you are saying, I'm still not 100% sure I've got my head around it. It's just hard for me to think of linear regressions when the sum of the X variables is constrained to equal 100. Do econometricians do this a lot? I expect they must know how to do it.

Stephen?

Yes! I think that's exactly the right way to think about it. I think you are right.

[Nick Rowe]

Yep. Run a regression of some measure of the business cycle on all the variables in that survey, recognising that they must sum to 100%, then find the amount of the business cycle explained jointly by those two variables (poor sales and labour quality).

[Nick Rowe]

Phil: yes, I think your ratio index might make a lot more sense than my additive (subtractive?) index, now I think about it. The normalisation doesn't matter, as long as it goes up and down. I'm just worried about "overweighting" one of the two relative to the other. I can't get my head around that.

Actually, I should probably stop trying. I don't have a comparative advantage in this.

[Jim Rootham]

The other big blip in that graph is insurance. I gotta believe that is health insurance. Are the GOP really going to repeal that?

[Lord]

Insurance initially jumped in '93. I suspect this was when it was first offered as a choice. It grew during the 2000 recession/recovery and then fell late in the last cycle. It really hasn't budged over the last few years. So much for policy/regime uncertain pols like to tout.

[Rahul]

I dont think such graph would have indicated anything even in times of earlier "recalculation" or structural unemployment.

The reason for structural unemployment comes from shifting business realities. It happens in a small number of new businesses not in majority of the sample. those businesses become increasingly competitive pressurizing others to follow, at that time we have this labour problems. At this time, businesses are hunkered down doing tried-and-tested things.

[GA]

"How come other people can do such things??"...

Of course, even when labour quality is less of an issue, it never disappears entirely ;)

[Nick Rowe]

Rahul: OK, there may be some sort of sample selection bias, in that the firms that *would* answer the survey saying they can't find the right sort of labour don't exist, just because they can't find the right sort of labour. That would bias the numbers down.

But that downward bias can't explain why the number of firms answering "labour quality" has fallen so much. There used to be lots of firms giving this answer, now almost none.

GA: my morning chuckle!

[pcle]

No. The graph does not speak to recalculation. Recalculation according to Kling is about more than mismatch, its also about new firm formation.
Ashwin Parmaseran goes into this on his blog:

"It is likely therefore that there is a significant pool of unemployment that cannot be justified by the simple mismatch argument. But this does not mean that the “recalculation” thesis is not valid. The simple mismatch argument ignores the uncertainty involved in the “Post-Minsky Moment economy” – it assumes that firms have known jobs that remain unfilled whereas in reality, firms need to engage in a process of exploration that will determine the nature of jobs consistent with the new economic reality before they search for suitable workers. The problem we face right now is of firms unwilling to take on the risk inherent in such an exploration."

[David Pearson]

I agree with pcle above. The driver of small business job creation is the birth/death of firms. Ten restaurants might close to produce 100 unemployed workers. Surviving restaurants would tell surveyors they have more than enough quality laborers. Meanwhile, would-be entrepreneurs, surveying the landscape, forgo launching start-ups. is the lack of start-ups a function of a temporary lack of AD or something else?

In the factory sector, AD shocks explain a lot: factory endowments and technology sit idle, waiting for falling wages to put them into use again. How does this work for restaurants, real estate agencies and wedding planners? Do the assets and unused skills of failed firms in those industries constitute "idle capacity", or something else? Are entrepreneurs in those sectors waiting for a (3%?) decline in real wages to launch new firms?

A much higher percentage of losses in this recession have occurred in the services sector than ever before. This makes the difference between the "AD" and "recalculation" stories less black and white.