Sunday, October 12, 2014


I want to look at each of the images used for yesterday's animation sequence. Each of the similarities, I mean. But I'm having a hell of a time getting started with the writing. So I'm gonna just start.

Image #1: No lag... Similarity: January 1919 - November 1930

Graph #1: No Lag
I'm using another FRED graph, monthly data, same two series shown in the animation sequence, to see the dates of similarity. Maybe that'll move the writing along.

I will describe a "lag" in terms of how many months I have to offset the Base Money (black) graph to the right to obtain a highlightable period of similarity. I will describe a "similarity" using the dates of the Inflation (red) graph, the one that doesn't move back and forth thru time.

The similarity highlighted in Graph #1 ends suddenly in late 1930, when it appears the Federal Reserve suddenly decided to do something about the Great Depression. Just after the yellow highlight ends, base money (black) shoots up while the rate of inflation falls for six months more (till June 1931) and then drags bottom at -10% till March 1933.

That separation of the lines shows the lag developing.

Image #2: Lag 29 months... Similarity March 1933 - March 1934

Graph #2: Base Money Lagged Two Years
During the similarity highlighted on Graph #2, base money (black) increased from November 1930 to October 1931 -- a period of eleven months. The rate of inflation (red) increased from March 1933 to March 1934 -- twelve months. So the lag itself was a month longer at the end of the similarity than it was at the start.

At the start of that similarity the lag (November 1930 to March 1933) was 28 months. By the end of that similarity, it was 29 months. I'll call it 29 months.

Image #3: Lag 50 months... Similarity March 1934 - May 1937

Graph #3: Base Money Lagged Four Years
On Graph #3 the similarity starts with the peaks that end Similarity #2. The lag this time is clearly visible as a downtrend that's twice as long for the red as for the black, followed, then, by a more hesitant red uptrend. The similarity ends with rather modest peaks in both the red and black lines.

Note -- This image has been revised. Originally I showed one yellow stripe of highlight, running through those modest peaks. (You can see that version in yesterday's post.)

Incidentally, the peaks that end Similarity #2 no longer appear aligned because on Graph #3, base money (black) is shifted two additional years to the right. (If that doesn't make sense, go fiddle with the interactive experiment graph again.)

The end of this similarity is the highlighted peak of the black line in March 1933. And for the red line, in May 1937. The difference -- the time the black line is lagged -- is 50 months.

Image #4: Lag 63 months... Similarity May 1937 - October 1938

Graph #4: Base Money Lagged Five Years
Number four is a brief similarity. I'm figuring that it runs from the end of Similarity #3 to the start of the highlighted jiggles just before the black line skyrockets.

For the black line it starts with the end of Sim #3 (March 1933) and ends in July 1933.

For the red line it starts with the end of Sim #3 (May 1937) and ends in October 1938.

It's a challenge to get the dates right on these graphs, what with these images lagged (but not providing pop-up info windows) and the FRED graph providing the info windows (with nothing lagged). It's making my head spin. Hey, if you see anything that looks like a mistake, let me know.

Image #5: Lag 92 months... Similarity October 1938 - March 1951

Graph #5: Base Money Lagged Seven Years
For this one the similarity will start with the jiggles where the previous similarity ended. So July 1933 for the black line, and October 1938 for the red, these dates mark the start.

Where to end this similarity? For the black line I'm gonna go with the high point in the third tall spike in the group of three, so July 1943

For the red line I'm gonna stop it at the top of the third red spike of three, March 1951.

The highlighting is a bit off from these dates that I'm choosing now. It's a little bit arbitrary, choosing the dates. If you know a better way to figure it, do let me know.

(It's completely arbitrary.)

Image #6: Lag 74 months... Similarity March 1951 - March 1955

Graph #6: Base Money Lagged Six Years
We reached a maximum lag of 92 months (between 7 and 8 years) with Graph #5. On graph #6 the lag begins to recede. I'm calling the dates for this similarity as follows: For the black line, start at November 1944 and end at September 1949.

For the red line, start at March 1951. And end at, I'll say March 1955

For the red line, I start Similarity #6 where #5 ended. For the black line I'm leaving a little gap. Number six starts at the black peak where the yellow highlight starts. But number five ended at the nearby, higher, slightly earlier peak. So there is a brief gap between similarities at that point.

Note that other than this brief gap, and the dis-similarity described under Image #1, there are no gaps where one similarity ends and the next one begins. Granted, this is all by eye, and arbitrary. But these "similarities" are nothing but short sections of the overall graph, short enough that we can compare them and describe them.

But if the similarities connect like railroad cars, then what we're seeing is that the pattern of the red graph and the pattern of the black graph are actually quite strikingly similar. The main difference between the two lines is nothing more than a lag -- a long and variable lag. And I'm thinking that the best way to show similarity would be with an "accordion axis" if I knew how to make such a thing.

I'm also thinking that this accordion axis would essentially be a standard axis modified by a calculation that shortens or lengthens the time units in an irregular pattern. I want to create such an axis, because I want to take that calculation and see what it looks like as a normal graph. If it has a familiar look, if it reminds us of something, then perhaps we will have found a major component of the rise and fall of this lag.

Image #7: Lag 60 months... Similarity March 1955 - May 1960

Graph #7: Base Money Lagged Five Years
On Graph #7 the lag recedes again. I begin Similarity #7 where similarity #6 ended: September 1949 for the black line and March 1955 for the red. I end the red one in May 1960. And I end the black one at May 1955. We end with a five-year lag.

I figured these start- and stop-points by looking for similarities in the patterns. Tomorrow I'll graph the lag lengths.


Oilfield Trash said...


Since you are just getting started I offer the following for your consideration.

Kyd­land & Prescott, Busi­ness Cycles: Real Facts and a Mon­e­tary Myth, Fed­eral Reserve Bank of Min­neapo­lis Quar­terly Review, Spring 1990.

Some conclusions from the paper

"There is no evi­dence that either the mon­e­tary base or M1 leads the cycle, although some econ­o­mists still believe this mon­e­tary myth.

Both the mon­e­tary base and M1 series are gen­er­ally pro­cycli­cal and, if any­thing, the mon­e­tary base lags the cycle slightly. (p. 11)"

"The dif­fer­ence in the behav­ior of M1 and M2 sug­gests that the dif­fer­ence of these aggre­gates (M2 minus M1) should be con­sid­ered… The dif­fer­ence of M2 — M1 leads the cycle by even more than M2, with the lead being about three quar­ters.” (p. 12)"

So rather than inflation being cre­ated with a lag after base money increases the data supports the view that inflation is pro­cycli­cal.

Introducing lags and lead to empirical data to explore relationships requires that causation to be clearly understood.

As I have suggested before "p*y=m*v" seems to be a better frame work.

Something to think about.

The Arthurian said...

I reserve the right to explore any empirical relationships I want.

J&P Abstract: "A finding is that the price level is countercyclical in the post-Korean War period."

I'm thinking that "the post-Korean War period" translates to "since the mid 1950s". That could be right. It would not really contradict what I am finding for the 1919-1960 period. A few years of overlap, a few years of transition, I can live with that.

I have already argued that the forces driving inflation changed by 1960:
Before 1960, money was dominant. After 1960, credit was dominant.

K&P: "The dif­fer­ence in the behav­ior of M1 and M2 sug­gests that the dif­fer­ence of these aggre­gates (M2 minus M1) should be con­sid­ered… The dif­fer­ence of M2 — M1 leads the cycle by even more than M2..."

I've spent a lot of time with M2-M1. It is mostly savings.

It's not clear to me how increases in savings would "lead the cycke" or be any kind of driver for increases in economic growth. I know savings equals investment and all that, but there is no nexus which unites decisions to abstain from present consumption with decisions to provide for future consumption.

Oilfield Trash said...


"It's not clear to me how increases in savings would "lead the cycke" or be any kind of driver for increases in economic growth. "

To increase savings at an aggregate level you must increase spending.

The Arthurian said...

Hey Oilfield, good link. I read about a third of it so far. Interesting. I like the history it provides.

I skipped to the end and looked at their charts 5 thru 8 and 9 thru 12.

5 thru 8 show datasets that look like the GNP they use for comparison. These are "procyclical" I guess.

9 thru 12 seem to run opposite to the GNP series. "Countercyclical".

Okay... now I admit I skipped to the end, but I have a problem with their graphs. The GNP they use as a basis for comparison is "Real Gross National Product". It has the inflation stripped out of it. Okay...

Charts 5 thru 8 show Real GNP in comparison to "its components" -- consumption, investment, and government spending. But as these are "components" of Real GNP, I have to assume that these components, like the GNP they use, have had inflation stripped away.

By contrast, charts 9 thru 12 compare Real GNP to other things -- to prices, and to various measures of money. I'm pretty sure that these comparison series have NOT had the inflation stripped away... I'm VERY sure that the "prices" series has not had the inflation stripped away.

If I am right, the countercyclical appearance of series in charts 9 thru 12 is a result of comparing inflating series to an inflation-adjusted series. And the "procyclicality" in charts 5 thru 8 is a result of comparing inflation-adjusted data to inflation-adjusted data.

This appears to be a rather common trick.

Oilfield Trash said...


I will go back and look at the graphs. It is a good catch.

Do not take this the wrong way, but my experience has shown me that Inflation can’t be measured precisely since there is no observable price level.

The price level is an index and an index is an arbitrary figure that could be arrived at through different paths and rationales.

Sumner made a point once in regards to CPI

"Government price indices don’t measure the prices that are of macroeconomic interest. For instance in the 6 years after the housing bubble peaked the US, BLS data shows housing prices rising by about 10%, while Case-Shiller showed a 35% decline. Housing is 39% of the core CPI. That’s a big deal."

So if you adjusted CPI with Case-Shiller data you get a different lens to view CPI.

Meaning with without Shiller Data CPI has been over stating inflation for about six years and prior to the housing crisis CPI was understating inflation.

Is Shiller data important, I think so since much of the spending increase and decreases was due to mortgage debt.

The Arthurian said...

"Do not take this the wrong way, but my experience has shown me that Inflation can’t be measured precisely since there is no observable price level."

Or maybe, can't be measured precisely since there are so many observable price levels.

But forget the explanation; let's suppose you're right: Inflation can't be measured precisely.

What does that say about the reliability of "Real GDP" data? If you take GDP at actual prices and divide inflation out of it, you get "real GDP". But if the inflation numbers are questionable, then the Real GDP numbers must be AT LEAST as bad, probably worse.

Oilfield Trash said...


"What does that say about the reliability of "Real GDP" data?"
Depends on the question about the economy you are asking.

If you are interested in the welfare of the people associated with this consumption rather than the monetary value of production, then it is worthless since it only measures the latter.

GDP is measuring a flow of monetary value of production, as you rightly pointed out, at numerous price levels within numerous markets.

CPI or GDP is therefore nothing more than a measure of price changes.

Since prices change from year to year, and a nation’s productivity level over time is tracked in monetary terms using GDP, of course poses a problem.

How can you tell if a change in a country's level of output is due to a real change in productivity or whether it is due to fluctuations in the level of the prices of that output?

The answer is you can't without constructing a GDP fudge factor called the deflator.

Unlike the CPI, the GDP deflator is not based on a fixed basket of goods and services; the "basket" for the GDP deflator is allowed to change from year to year with people's consumption and investment patterns.

Some will argue that statistically trends in the GDP deflator are similar to trends in the CPI so interchanging them is reasonable. However they are both independent indexes with variable hedonic adjustments, weighting and market baskets used to produce an observable macro price level.

They are only as reliable as their present construct within its associated time scale, which is not saying their present construct is reliable over any historical time scale.

Back to the increase in base money drives inflation I asked what was the transmission mechanism.

I should have been more specific and asked how does an increase in base money significantly change the price level in the private sector?

Debt contracts generate spending (who would enter a debt contract and not spend it), and spending is the strong force on the price level.

With the assumption of a closed economy Growth of debt accelerates faster that the productive output GDP goes up (inflation), Growth of debt accelerates at an equal pace with productive output GPD goes sideways. Growth of debt decelerates faster than productive output GDP goes down (deflation).

When you factor in the CAD it can also influence the price level as well. But this post is to long as it is.