Saturday, October 4, 2014

Business Sector: Labor Share


This is the FRED graph:


This is what seems to bother most people:


This is what bothers me:


Friday, October 3, 2014

Velocity: Percent Change


Last April, under the title Velocity and the Great Moderation, I showed this graph of percent change in the velocity of M1 money:

Graph #1: Percent Change from Previous Quarter, M1 Velocity
M1 velocity was tame till 1980, and rambunctious thereafter.

But I was looking at that graph just now, feeling sorry for myself because the velocity of M1 only goes back to like 1960. Suddenly I had a brain fart:

Graph #2: By using two series I get data all the way back to 1869!!!!!
Here's the "Percent Change" version of Graph #2:

Graph #3: Percent Change in Velocity, Annual Data
Two things.

On Graph #3, shortly after 1875, there is a period of a decade or more where the percent change in velocity almost continuously negative. Below zero. It was a time of rapidly falling velocity, as you can see on Graph #2.

It was also a time of deflation following the Panic of 1873, which Marcus Nunes describes as a ‘nonevent’. Marcus writes: "in the 1870s and 1880s the economy performed robustly in spite of a drawn out fall in prices".

Deflation, associated with a consistently negative rate of change in velocity.

The other thing happened just about a hundred years later, around 1975, when for a decade or more the percent change in velocity was almost continuously positive. Above zero. It was a time of rapidly rising velocity, as  Graph #2 shows.

It was also a time of inflation. It was the time of the so-called Great Inflation.

That's interesting, I think: Two divergences from the norm, in opposite directions. There must be some kind of rule that goes along with this.

Thursday, October 2, 2014

The Other Side of That TFP Coin (and something unrelated)


Yesterday I took a quote from a conversation at ResearchGate. I won't give you the link again, because every time I go there ResearchGate pops up a window asking me to "join for free". I close the window and it doesn't bother me again, and it doesn't seem to damage my computer, far as I can see. But still.

However, I do want to look at some of the stuff from that conversation. John Ryding, whom I quoted yesterday, links to the download page for a paper by Charles R. Hulten, who writes:

Economists have long recognized that total factor productivity is an important factor in the process of economic growth. However, just how important it is has been a matter of ongoing controversy. Part of this controversy is about methods and assumptions. Total factor productivity growth is estimated as a residual, using index number techniques. It is thus a measure of our ignorance ...

Ignorance, yeah. What I said yesterday:

Total Factor Productivity is the contribution to economic growth that's left over, after we subtract the contributions to growth that we understand ... It is the contribution to growth that we do not understand.

At ResearchGate, Hak Choi agreed with John Ryding, saying "For me, TFP is merely residue."

Residue. Something left over. Yeah. That's how it is calculated, according to several sources (including the Bank of England). Total Factor Productivity is a statistical discrepancy with a fancy name.

And then, at ResearchGate, Fabrizio Pompei offered a "contradictory" view:

Actually, TFP as mere residual is only one way to observe total factor productivity. Generally, TFP is a productivity measure involving all factors of production.

Yeah, I don't need to read a sweeping generality to know that the part of economic growth which is not explained by capital, and not explained by labor, must be explained by everything else.

I think economic growth is important, and restoring economic growth is important, and I think we need to know more than that "everything" is responsible for growth.

Fabrizio's contradictory view included the mention of the book An Introduction to Efficiency and Productivity Analysis, a 331-page PDF by Timothy J. Coelli and three other writers.

I knew I shouldn't look at it, but I did, and sure enough, it shut down my tirade against Total Factor Productivity.


Dominating, isn't it?

Now... It's over 300 pages long, so I don't expect to read it. But I was looking it over this morning before work, and it did look interesting. Of course, my options were to keep reading, or to go to work.

I want to say one last thing about TFP. I only read a few pages out of the book, so I don't know. But it looks like the book is an investigation of four different ways to measure efficiency and productivity, with "total factor productivity" among them. What it does not appear to be is an investigation of the relation between TFP and efficiency. That relation seems to be assumed. But like I said, I read very little.


In the introduction of the Efficiency and Productivity Analysis book I got severely distracted. They present a graph to show the difference between efficiency and productivity. The graph shows a hypothetical firm's inputs on the X axis, and its outputs on the Y axis:
To illustrate the distinction between technical efficiency and productivity we utilise Figure 1.2. In this figure, we use a ray through the origin to measure productivity at a particular data point. The slope of this ray is y/x and hence provides a measure of productivity. If the firm operating at point A were to move to the technically efficient point B, the slope of the ray would be greater, implying higher productivity at point B. However, by moving to the point C, the ray from the origin is at a tangent to the production frontier and hence defines the point of maximum possible productivity. This latter movement is an example of exploiting scale economies. The point C is the point of (technically) optimal scale. Operation at any other point on the production frontier results in lower productivity.

From this discussion, we conclude that a firm may be technically efficient but may still be able to improve its productivity by exploiting scale economies.

If they move from Point A to Point B they produce more. But if they move from Point B to Point C they get maximum output per unit of input.

As I see it, the optimum size of that hypothetical firm is at Point C.

I talk about the "Laffer limit" sometimes: Doing more of something gives better and better results for a while. But after the Laffer limit is reached, it gives worsening results.

This first occurred to me back in maybe the early 1980s when I used to get the Kiplinger letter -- a B-size sheet of paper, 11"x17", folded in half to make a four-page A-size paper, 8½"x11", with four pages of Kiplinger info, for far too much money, even back then.

Anyway I was reading the Kiplinger letter one day, they were talking about agribusiness as I remember, and they said that agribusiness was growing "beyond its economies of scale". That floored me. It was so big that it was losing money. Its growth was driving its rate of profit down.

Why would anybody do that? My answer is: the tax code. The business income tax gives you a tax break for growing your company. So you grow, to get the tax break. And even if you grow beyond your economies of scale, the government gives you a tax break that more than makes up the difference.

So businesses grow, and big ones buy up little ones, and we lose competition that way, but the growing businesses get a tax break, so to them it's worth it. And at some point they find their profit rate is not as high as it was when they were smaller, but hey, they get those luscious tax breaks. So the government gets less revenue, and businesses merge and consolidate, and the economy is distorted by growth beyond economies of scale, and you can't keep this up or your economy will turn to shit.

And then policymakers strengthen the incentives, thinking that merger and consolidation is the same as growth.


There was a little more in Chapter 2:

In summary, the production function depicted in Figure 2.1 violates the concavity property in the region OD and violates the monotonicity property in the region GR. However, it is consistent with all properties along the curved segment between points D and G - we refer to this as the economically-feasible region of production. Within this region, the point E is the point at which the average product is maximised. We refer to this point as the point of optimal scale (of operations).

The figure 2.1 is similar to 1.1, but annotated differently. And the authors are explicit, defining a "point of optimal scale", meaning scale of operations specifically. And their graph shows plenty of room for a company to grow beyond its economies of scale.

So, we're not talking Total Factor Productivity anymore. I'll come back to that eventually. But I couldn't let this go.

Wednesday, October 1, 2014

The Calculation of TFP


One of the problems with economic models is that they seem to be based on imaginary conditions. So, when I found a PDF titled "Creating a Macroeconomic Model Using Real Economic Data" I judged that book by its cover (so to speak). I liked it right away.

That PDF is SolowProjectNotes.pdf, from the archives of Professor Rodney Smith at the University of Minnesota.

The PDF shows a calculation for Total Factor Productivity. For me that's useful, because I've been wondering for a while now how they calculate TFP. Hey, don't be intimidated by this:


See the equal sign? All that big mess of stuff on the right side of the equal sign is equal to TFP at some time t. I know this because I see TFPt on the left side of the equal sign.

I know. I know. It's not the stuff on the left side of the equal sign that's intimidating. It's all the stuff on the right. But parts of that are familiar. Take the first part:


"Y" is output. "t" is time. Yt is output at some particular time t.

Yt-1 is output at the time just before time t. For example: If "t" is the year 1995, then "t-1" is 1994. And if you take output (or Y) for the year 1995 and subtract the output for 1994, then you get the change in output from 1994 to 1995. That's what this is:


Then if you divide that "change" number by Yt-1, the output for 1994, you get the percent change number. I guess you'd have to multiply by 100 and stick a percent sign after the number to make it a percent, but that's just formatting.

And then if you look at the equation again, you've got that "Y minus Y divided by Y" thing there, percent change in Y. And you've got two other things just like it, except one of them has K instead of Y, and the other has L  instead of Y:


So you've got "percent change" calculations three times -- once for Y, which is output... once for K, which is capital... and once for L, which is labor. I didn't make those up, K is for capital and all that. It's in the PDF somewhere. But it's pretty easy to remember.

I guess C is for "consumption" so you can't use C for capital. That'd be my guess.

Going back to the complete equation again...


we've got TFP (for some year "t") equals the percent change in output, minus something times the percent change in capital, minus one-minus-something times the percent change in labor.

It's kind of nice that everything we've figured out so far is "percent change" numbers. Because subtracting "percent change" from "percent change" seems pretty reasonable. It's not like trying to subtract number-of-employees from gross-domestic-product. (You can't do that!)

The only other things in that equation are the "something" and the "one-minus-something". That's how "labor share" and "capital share" are distributed. I'm not sure why it's reasonable to do that the way they do. But it seems like it might be. Anyway, the "something" is capital share, and the "1" is 100% of what is shared, and 100% minus capital share is labor share. So there ya go.

In broad stroke then, TFP is what's left when you take output and subtract the contribution of capital and labor from it. As the formula is written, it is figuring the "percent change" in TFP, because it is subtracting percent change values from percent change values and what's left must also be percent change values.

But GDP is GDP, whether you look at it in billions of dollars or as percent change values. And the same is true of TFP.


In broad stroke, TFP is what's left when you subtract from output the contribution of capital and labor. That's not me talking. It's the SolowProjectNotes.

Anjaree says it too: "TFP growth rate = growth rate of output - growth rate of factors of production."

At Researchgate, John Ryding of RDQ Economics says it too: "total factor productivity ... is the growth of output that cannot be explained by the growth of the inputs (labor, capital etc.)"

Even the Bank of England says it, if not in so many words. In the PDF titled Measuring total factor productivity for the United Kingdom, in the appendix, they show "The formula used to calculate TFP growth":


Now, this doesn't look much like the equation from the SolowProjectNotes. But there are key similarities. On the right side of the equal sign we start with Y -- output. We subtract some messy thing that ends with K -- capital. And we subtract some other messy thing that ends with L -- labor. Total Factor Productivity is calculated by taking output and subtracting, very precisely and carefully, some number arising from capital and some number arising from labor. This is exactly the same as in the SolowProjectNotes.

Total Factor Productivity is the contribution to economic growth that's left over, after we subtract the contributions to growth that we understand -- the contributions from labor and capital. Total Factor Productivity is the discrepancy. It is the contribution to growth that we do not understand.

Don't you want to open that black box and see what's in there? I do.

Tuesday, September 30, 2014

A Statistical Discrepancy


If you go to FRED for the default graph of GDP, the title reads Gross Domestic Product and the left border reads (Billions of Dollars).

If you then change the "Units" for the graph to "Percent Change from Year Ago" the left border changes to read (Percent Change from Year Ago). But the title still says Gross Domestic Product.

It's a pretty minor point. But it makes sense: GDP is GDP, whether you look at it in billions of dollars or as percent change values. That's probably obvious. I just want to make sure it's obvious.


At Vox last May, Yglesias pondered differences between Gross Domestic Product and Gross Domestic Income:
This morning we learned that GDP shrank at a 1 percent annual rate and GDI shrank at a 2.3 percent annual rate. Disturbing. And you may be even more disturbed when you learn that GDP and GDI are the exact same thing.

Check out any economics textbook and they will tell you that GDI = GDP. And it does. By definition. And yet when the government sets about to measure GDP and GDI they are never equal.

In theory, these two series should add up to the same thing. In practice, data sources are always imperfect and they don't add up to the same thing. The difference is known as the statistical discrepancy.

That's it: The difference is known as the statistical discrepancy.


But today's topic is TFP, not GDP.

Back in June I wrote:

I don't know how they figure Total Factor Productivity. Maybe it's an error term. Maybe there's a discrepancy in the calculation and they use TFP to reduce or eliminate the discrepancy.

Back in June I got the idea that TFP is "an error term". I quoted Simplicitus:

Total factor productivity is the otherwise unexplained productivity left over after all the individual factors (labor, capital, etc.) are taken out.

and I said

See? That's why I say TFP seems like an error term, a correction value, an adjustment to make the answer come out right.

I quoted Diego Comin:

Total Factor Productivity (TFP) is the portion of output not explained by the amount of inputs used in production.

and I said

This is starting to get familiar... the growth of labor and capital do not fully account for the growth of output. The difference ... is called TFP, Total Factor Productivity.

So that's pretty much how I came to think of Total Factor Productivity as an error term. It is the measure of a discrepancy.


There is a statistical discrepancy between GDP and Gross Domestic Income. They call it a statistical discrepancy.

There is a statistical discrepancy between GDP growth and aggregate production functions like the Solow Growth Model. They call it "Total Factor Productivity".

With GDP and GDI they don't take the statistical discrepancy and make up a name for it like Total Income Productivity and make it not seem like a discrepancy. They just call it a discrepancy.

But with GDP growth it's different. They take labor and capital and fit them to an equation, and compare the result to GDP growth. And when it comes out on the low side, they say there must be something else involved, and they give it the name "Total Factor Productivity".

They explain it by saying Oh, well, that's the increase in labor efficiency. And yes, it makes good sense. But it's just a story. It's not empirical. It's not based on anything. And it doesn't account for the effects of things like the growing accumulation of private debt.

TFP cannot be measured directly. Instead it is a residual ... a mismatch in the calculation ... a statistical discrepancy ... a big one.


This is what fascinates me about TFP: It's a black box.

I want to open up the box and see what's in there.

Monday, September 29, 2014

Hopscotch Graphs


The other day I looked at a recent Noahpinion post that used an old David Beckworth graph showing Total Factor Productivity data from John Fernald.

Fernald's TFP (left) looks significantly unlike FRED's TFP (right):

Graph #1: Fernald's TFP (Beckworth)

Graph #2: FRED's TFP

To improve the comparison, I overlaid the one graph on top of the other:

bottom image
top image

Graph #3: Beckworth's TFP (blue) Overlaid on FRED TFP (red)

Then I kept looking, and found another TFP graph of Fernald's data from Noah. So now I want to repeat the above layout of graphs, but using Noah's graph in place of Beckworth's:


Graph #4: Fernald's TFP (Noahpinion)

Graph #5: FRED's TFP (again)

Again, the overlay:

bottom image
top image

Graph #6: Noah's TFP (blue, green, dingy red) Overlaid on FRED TFP (bright red)
On Graph #3 I adjusted the "overlay" graph up-and-down to get a good fit in the early years. On Graph #6 I adjusted the overlay to get a good fit in the later years. (Noah's graph is taller, and already sits a lot higher than the FRED graph. So I didn't want to move it up more to make the early years align.)

But you can see on Graph #6 that the years from 1980 to maybe 2006 are a good match. The red lines follow the same general trend. Also in the early years, until maybe 1966, the red lines seem to share a trend.

I'm wondering now if maybe the FRED data is calculated in a way similar to Fernald's red line, which shows the TFP for "durables".

One more comparison graph, a straightforward one this time, two FRED series:

Graph #7: Total Factor Productivity (red) and Multifactor Productivity (blue)
Too bad that blue line got such a late start.

Sunday, September 28, 2014

Garbage in, garbage out

Continuing yesterday's search of this site for Total Factor Productivity . . .

After the Van Zandweghe post in March 2014, my next mention of "Total Factor Productivity" came in June with Oh! I've seen this shape before.... That was my reaction when I first looked at FRED's version of TFP: Where have I seen this before?

Just a week later, in Uptrend to 1966, then stable to 1982, then uptrend again I found the matching shape. The match was SRW's Real GDP relative to the size of the labor force, at Interfluidity.

If productivity is output per worker, then maybe it makes sense that the graphs are similar. But TFP isn't "output per worker" so I'm not clear on this.

In any case, I ended that post "wondering now how they calculate TFP."

Before long I was looking at the Solow growth model.

//

From Roberto Chang of Rutgers:
In the original Solow model, time is continuous and the horizon is infinite. Without loss of generality assume that time is indexed by t in [0,∞).

At each point in time, there is only one final good. The good is produced via the aggregate production function:

Y=F(K,AL)

Here Y, A, K, and L denote output, labor productivity, capital, and labor, and are functions of time (i.e. we should really write Y(t) and so on, but we omit time arguments when not needed). The product AL is called effective labor.

Once you assume the existence of the aggregate production function Y=F(K,AL), it is clear that the growth of output can be due to growth of A, K, or L.

I like that one: If we assume the program we wrote for the Holodeck accurately mirrors reality, then we can assume the results the Holodeck shows us are correct.

It always cracked me up that on Star Trek: Next Gen, when they ran into a problem they couldn't solve, they would write a computer program and let the Holodeck solve the problem. Obviously, if you cannot determine the calculation that will solve the problem, you cannot write a computer program to run that calculation for you.

On the other hand, if you know the answer you want to get, you can write the code so it gives you that answer. Then you call your code a "model" and you call yourself an economist. Roberto Chang lays this out clearly:

Once you assume the existence of the aggregate production function Y=F(K,AL), it is clear that the growth of output can [only] be due to growth of A, K, or L.

//

Right or wrong, the answer we get is determined by the calculation we give the computer. Still, I am fascinated by the Solow model and want to understand it.

True confession: Usually when I'm reading a PDF and I get to a formula I just completely stop. (It even happened with Mason & Jayadev's Fisher Dynamics paper.) For an old math major like me, that's embarrassing. But the 6-page SolowProjectNotes PDF gave me no trouble at all. Things are well explained:

... for each time period t, Yt is gross domestic product (GDP), Kt is the capital stock, Lt is the size of the labor force, and At is labor productivity. The parameter α is the "capital cost share" we discussed in class, δ is the rate at which capital depreciates, and s is the savings rate for a country. We said labor grew at a rate n and labor productivity grew at a rate x ...

We next discuss how to calculate/estimate the parameters in our model...

We begin with the simple calculations first.

Some of the numbers come right out of a table. And the additional explanation you want is at hand:

α is the share of production costs paid to (owners of) capital, and 1 − α is the share of production costs paid to labor. For example, for the U.S., on average 62.9 percent of GDP was paid to labor and the rest went to owners of capital: i.e., α = 0.371 and 1 − α = 0.629. For Zambia, α = 0.533 and 1 − α = 0.467.

For me, even expanding the common concept the share of production costs paid to capital to

the share of production costs paid to (owners of) capital

helps a lot. Highly recommended reading...

But remember: Back in June I wrote:

spent most of Saturday going through the PDF, gathering data and entering calculations in a spreadsheet. It's one of those things where you have to put a lot of hours into it before you can make a graph to see if you're anywhere near right. I have patience for that kind of work, but it was most of Saturday.

Good luck with it.

//

By the way: The A in Roberto Chang's formula and the At in the SolowGrowthModel PDF both stand for "Labor Productivity". I think Labor Productivity is related to, but not the same as, Total Factor Productivity. I can't explain the difference yet. Couple days, maybe.

Saturday, September 27, 2014

When the moment is right


The earliest mention of "Total Factor Productivity" I find on this blog occurs in Suddenly remembered where I caught Noah considering the relation between TFP and unemployment. I wrote:

"TFP" being "Total Factor Productivity" which doesn't matter at the moment.

I think that's funny because it was three years ago, almost to the day. And now, finally, I'm looking at TFP. The moment is right, I guess.

//

I find the phrase Total Factor Productivity a few times after that, but nothing catches my eye (at the moment) until 26 March 2014: The relation between productivity and growth. In that post I quoted Willem Van Zandweghe of the Kansas City Fed:

In recent years, the U.S. economy has undergone a change in the behavior of productivity over the business cycle. Until the mid-1980s, productivity growth rose and fell with output growth. But since then the relationship between these two variables has weakened, and they have even moved in different directions.

I did some graphs comparing RGDP and TFP, then. But what strikes me just now is the contrast of Van Zandweghe's remark to that of Noah in the earlier post:

The recessions of the early 1980s were very short and "V-shaped," despite the extremely low rate of TFP growth at the time... By contrast, the early-2000s recession, though shallow, was much longer and "U-shaped", despite the recovery in TFP growth.

When I came upon that quote this morning I got stuck on the word "despite". Noah's use of that word implies things that shouldn't have coincided with other things, did coincide. Short recessions coincided with low TFP growth. Long recessions coincided with high TFP growth. I don't object to the coincidences. I object to Noah trying to shove "despite" down my throat before I've considered whether it's appropriate to say "despite" or whether it would be better to say "which makes sense because of". To whit:

The recessions of the early 1980s were very short and "V-shaped," which makes sense because of the extremely low rate of TFP growth at the time... By contrast, the early-2000s recession, though shallow, was much longer and "U-shaped", which makes sense because of the recovery in TFP growth.

I don't think it makes sense to say "despite". In a time of high productivity -- lots of output per worker -- you lose a lot of output with each unemployed worker. So I think it makes sense that severe recessions and high TFP go together. And in a time of low productivity, losing a worker means less loss of output. So I think it makes sense that mild recessions and low TFP go together.

I don't think it makes sense to say "despite".

//

Noah is implicitly arguing that there is a particular relation between productivity growth and economic growth. Van Zandweghe of Kansas City is arguing that there *was* a particular relation, which then changed, and now there is a different particular relation.

I couldn't see the change Van Zandweghe described. Nor can I see the relation Noah described.

Friday, September 26, 2014

I have to look into this more


Like pulling teeth, but pushing: to get ideas into my brain.

If I understand it right, "Total Factor Productivity" is a discrepancy. When you try to explain growth, you say: Okay, we've got labor, and we've got capital, and we take and mix them together in a big bowl and look at what we've got. You put another bowl next to it, that has GDP growth in it. And you compare the two.

What you see is that the GDP growth bowl is pretty close to full, but our labor-plus-capital bowl is half empty. So we have to add something to the labor-and-capital mix, until the one bowl is just as full as the other. And then we'll say we understand economic growth.

The thing we add to the bowl is called "Total Factor Productivity". The amount we have to add is the amount that makes the bowls equal. That's why I say TFP is a discrepancy. For every spoonful that our bowl is low, we add a spoonful of TFP.

//

So... the amount of TFP depends on the size of the discrepancy. Or take the bowls away and look at graphs, and say the value of TFP depends on how much we have to push up our calculation-of-growth line to make it match the official estimate of GDP growth. Something like that.

So the TFP number is the number that gives us the answer we were looking for, the answer that matches the GDP number. I'm not saying I'm doing sophisticated analysis here. I'm just saying it's what I understand.

But you see what I'm saying: If GDP is a line on a graph, and our labor-and-capital calculation is a lower line on that graph, then Total Factor Productivity has to be the number that fills the space between the two lines. There's not a lot of wiggle room.

//

So, assuming that's correct. then how can there be two different measurements of Total Factor Productivity??? There can't be. If we know what the GDP number is, and the labor number, and the capital number, then TFP has to be some particular number. If GDP is 10, and labor is 5 and capital is 3, and if our formula tells us that 10=5+3+X, then X has to be 2.

Like X, TFP has to be some particular value to work with the other numbers, which are givens. But remember the overlay graph I showed the other day?

bottom image
top image

Graph #1: Beckworth's TFP (blue) Overlaid on FRED TFP (red)

We've got two different measures of Total Factor Productivity.

What I think is, the numbers Beckworth used are figured a different way. They're not just a measure of the discrepancy. There's more economic thought behind those numbers. I have to look into this more.