Money Illusion Delusion

This is going to be a long post with lots of words and graphs.  Grab an apple or your favorite quaff, put your feet up and your thinking cap on.

Over at The Money Illusion, Scott Sumner, an Economist with a PhD in economics from the U. of Chicago recently posted an entry with the title: America’s amazing success since 1980: Why Krugman is wrong.

His thesis is that neoliberal economic policies promote economic growth, and further, that a country's growth rate relative to other countries is a direct result of having or not having employed neoliberal policies.

I am about to deconstruct Sumner's methodology and eviscerate his conclusions.  First, though, a disclaimer.  I am not an economist. I do not speak the jargon, and am unfamiliar with many of the concepts.  However, I have a technical background, two masters degrees (Chemistry and an MBA) some proficiency in math, critical thinking skills, a finely-honed sense of skepticism, and the ability to graph data sets.  Most importantly, I believe in data-driven conclusions, not conclusion-driven data mining.

Next, let's take note of the kerfuffle Mike Kimmel and Spencer over at Angry Bear got into with Sumner.  This illustrates the hazard of getting into a point-counter point with someone who is ideology driven, especially after he has framed the debate in a way that is favorable to his conclusions.   The thing to do, I'm convinced, is go after their basic assumptions, and refute them with actual facts.  That will be my goal in this post.




Let's also take note of Sumner's sloppy methodology, and apparent lack of irony.  He states:

Krugman makes the basic mistake of just looking at time series evidence, and only two data points:  US growth before and after 1980.

It's fine with me if somebody disagrees with Krugman.  Hell, it's fine with me if somebody disagrees with me.  I do insist, though, that the disputant think clearly and make some sort of sense, if he's out to prove anyone wrong.  Here, Sumner calls two time series, "two data points," when each series is -  how else can you say it - a series.  Then he goes on to draw global conclusions from three data points each for the several countries he deals with.   These are three data points chosen from a series that contains 28.  The best you can say about this is that it's lazy.  It smacks of  cherry picking.  When I saw it, my bull shit alarm started flashing bright red.

The data series Sumner harvests is GDP/Capita PPP (purchasing power parity) from the World Bank  There are several GDP series available from the WB, and other sources.  Though I'm using a WB GDP/Cap PPP data series, I'm not certain the series I'm using is the same one that Sumner used, but there are certainly similarities.  More on this later.

Sumner takes each country's GDP/Cap as a percentage of  U.S. GDP/cap for 1980, 1994, and 2008.  These are the beginning, middle, and end of the WB data series values.  He reasons that if a country's GDP/Cap is increasing or decreasing relative to the U.S. -  as indicated by how the ratio changes over time - that indicates how much better or worse their economic policies are compared to ours and other countries under consideration.   His final answer is (with his empahasis):

So there you are, all these countries support my hypothesis that neoliberal reforms lead to faster growth in real income, relative to the unreformed alternative.

To which he adds in an update (emphasis added):

But I was primarily interested in international comparisons, for which all I had was GDP data in PPP terms.  The point was that countries that did more reforms did better than those that did fewer reforms.  I am not denying that growth in US living standards slowed after 1973, rather I am arguing that it would have slowed more had we not reformed our economy.

 Neoliberal. in this context means the low tax, low regulation, free-market, pro-business, anti-union policies of Reagan and Thatcher.  Philosophically, it also implies what G.W. Bush called "the ownership society" as he attempted to dismantle Social Security, and the replacement of public good or community ideals with those of individual responsibility.

Sumner presents a table including GDP/cap data for several countries as a fraction of  U.S GDP/cap for his three selected years, then gives some text as to why these results confirm his idea that neoliberalism is superior to whatever it is that Krugman propounds.

This neatly finesses a number of issues:

GDP is a one-dimensional look at a country's economic performance.  A time series gives you growth and growth rate, not much else.

GDP/cap relates to the mean income of a nation, and tells you nothing about how the income is distributed.

In the U.S wealth and income have been redistributed from the have-nots to the haves in dramatic fashion since 1980, and the disparity is now the greatest it has been since before the Great Depression. I suspect the same might true in Sumner's favored countries.

In the U.S. at least, since WW II poverty has increased whenever there was a Republican president, most dramatically during the Reagan and Bush II eras, and decreased whenever there was a Democrat in the White House.

GDP/cap isn't just a fraction, like cutting a pie into 8 equal slices. It is a ratio of two values that are not independent of each other.  The nature of their dependence is complex and can well vary over time and between countries.  Further, a times series of ratio values is influenced by changes in both the numerator and denominator.  Consider two countries with identical GDP growth.  Country A has high rates of live birth and immigration, and low rates of mortality and emigration.  Country B has low rates of live birth and immigration, and high rates of mortality and emigration.  Clearly, Country A is a better place to live, but over time, country B will outperform it in GDP/Cap because of relative population decline.  Ratios are traps for the unwary. 

The actual numbers that Sumner looks at are GDP/Cap for each country as a fraction of GDP/Cap for the U.S.  This is a ratio of two ratios.  We aren't quite at the level of angels dancing on pinheads yet, but this is a mathematical game at a high level of abstraction, with a relationship to reality that is, at best, tenuous.


And not least, both Reagan and Bush II inflated GDP by running enormous deficits, so their values are artificially large(in spite of which, Bush II has given us the lowest GDP growth since the Great Depression.)  Clinton was pretty far right - he gave us NAFTA, welfare reform, and financial deregulation - but no neoliberal.  I'm not a big Clinton fan, but he reduced poverty, raised taxes and balanced the budget.  He will be a touchstone before this post is finished.

Reagan, in particular had at least some actual  policies that are conceptually inconsistent with neoliberalism.  As an example, he was the only president since WW II "who increased both the size of the national debt as a percentage of Gross Domestic Product (GDP) and the percentage of Americans employed by the federal government."  (See Pg 2 of the intro, included with Chapter 1 at this link.) 

Sumner's weak conceptual framework is further undermined his assumption that GDP growth is solely or primarily a function of the native degree of neoliberal policy implementation, ignoring business cycles, economic shocks, political unrest, imports, exports, inventory adjustments, and all other endogenous and exogenous factors that can come into play.

Let's agree on something before we get into the meat of the discussion. Since 1980, there have been two periods of nominally neoliberal economic policies in the U.S: 1) under Reagan (such as it was) from 1981 to 1989, and 2) under Bush II, from 2001 to 2008. The 12 year span encompassing Bush I and Clinton was a non-neoliberal interregnum.

Here is Sumner's table of each country's GDP/Cap, as a fraction of U.S GDP/Cap, along side my values for those same items. Both of us got our data from the World Bank. I am baffled by the differences, highlighted in red. I've ignored those with small differences in the third decimal place.
Our values for European countries differ. Some of these difference are pretty minor, some are pretty substantial.  Make of it what you will.  Our values for all other countries are either right on, or very close.

Sumner identifies countries that have done well and those who have done poorly over that 28 year span. 
On the good side: the city-states of Hong Kong and Singapore, Great Britain (with Thatcher's neoliberalism), Australia, Canada, and Sweden, all of which introduced neoliberal reforms on or about 1994, Sumner tells us, and Chile, South America's primary neoliberal experiment.

The also-rans are Japan, France, Germany, Italy, Switzerland, and Argentina.  See Sumner's post for details.

Looking at that table of numbers doesn't do much for me (after all, for many of these countries, decades of presumably good or bad policy have lead to changes in the second decimal place.) so I graphed them.  First, Sumner's good guys, vs the U.S., pegged at 1.0.  In this set of graphs, for European countries, where the data differ, I am using Sumner's numbers.

First easy observation: great gains by Singapore and HK, nice positive for Chile, so-so, at best, for the others.

Second easy observation: an inflection point at 1994 is a common feature, though not all bends go the same way.   And where there is one bend, there could be two, or five, or seven, if we look at the entire data set.

Next, the bad guys.

First easy observation:  Switzerland looks like the big loser here, Argentina holds its own after 1994, other changes are not particularly dramatic.
 
Do you see a lot there to hang your hat on?  Frankly, for most of the countries considered, there's not much there, other than an inflection point.  Again, for the majority of the countries considered, there is a difference in the second decimal place, after 28 (or 14) years of presumably definitive policy action.

As scant as this all appears, my approach is to look at the entire data series, not three isolated years, and see what picture that paints for us.

Let's look first at Sumner's good guys.

The red/blue line at 1.00 is U.S performance, of which the other country values are percentages.  Yellow = Singapore, pale blue = Hon Kong, pink = Australia, Red =Sweden, dark blue = U.K.  Same color code as above.   I've place a few horizontal blue lines on the graph to indicate the duration of the Clinton administration.  The Reagan years show two gains, two losses, and not a lot of change for  the U.K.  Trends during Clinton's time were flat or down.  Everyone gained against Bush II

To get another - and frankly, more dramatic - look, I took a 5 Yr slope  (rate of change) by simply subtracting from each year's value, that from 5 yrs earlier: ROC =Val (85) - Val (80), etc.  Here is a graph of that data, same color code as above.

   

For Singapore and HK we see positive and increasing slopes during Reagan and Bush II, sharp declines during Clinton.   Australia and Sweden have fairly constant, slightly negative slopes vs Reagan.  Australia gains during Bush I, but starts to falter against Clinton.  Sweden goes from negative to positive slope during Clinton's time.
Great Britain meanders along, mostly a little bit above the zero line.  Well, they do have socialized medicine there.  They all gain vs. Bush II.  

Now, for Sumner's bad guys.  Color coding same as above.

Not a lot of positive here - except during the late 80's - the latter part of the Reagan regime, when even laggard Italy showed some positive action.  Note that the biggest declines are relative to the Clinton administration.   These unreformed economies all gain against Bush Sr. and mostly hold their own against Bush Jr.

Let's look at the 5 Yr ROC.

Every country in the group had its best performance from about 1985 to 1991 , and it's worst performance during the Clinton era.  After 2000, it's a bit of a tangle, with all slopes, except Italy's, reaching or approaching 0 by 2005.   

We haven't yet taken a close look at Chile and Argentina.  Their values are numerically similar, and far from those of other countries, so I've graphed them together.

Top two lines are the actual data points for Chile and Argentina, and the bottom two are ROC.  We see Argentina in a steep decline during the 80's - a time of local political turmoil and economic uncertainty. Meanwhile, Chile is pretty flat relative to Reagan, and both countries take off against Bush I.  They flatten and lose ground relative to Clinton.  Both gain again vs Bush II, Argentina far more dramatically.  

I just realized I left Canada out of the graphs.  For the sake of completeness, here it is.

Pretty lame for a good guy.  The values for 1994 and 1999 are identical at .816.  The only gain is against Bush II, and it's not much. 

Conclusions:

Remember, this whole line of reasoning is very highly suspect.  Given that caveat, what Sumner would have us believe - that neoliberal reforms lead to superior performance vis-a-vis other countries, is the exact opposite of what this data tells us.

First off, most of the countries looked at ended up 28 years later at pretty close to where they started.  How can you draw broad conclusions about policy when the results are trivial?

Second, the big gains came against both Bushs.  Sr. had a major recession, while Jr. had a recession and more.  He also gave us the lowest sustained level of GDP growth in the entire post WWII era.  Endogenous U.S. factors during those times inflated the numbers of other countries.

Third, other countries' performances relative to Reagan are a mixed bag, with both gains and losses - mostly small - taking place during the span of his 8 years.  Between 1985 and 1989 the poorest performing countries were either flat or showed small gains.

Fourth, nobody made a big gain vs Clinton.  Those countries who lost vis-a-vis the U.S. lost the most during his presidency.  Those countries who gained the most  vis-a-vis the U.S. either flattened or declined after 1995.  In fact, none of these countries gained at all against the U.S from 1995 to 1999.  Coincidentally - or not - that is when neoliberal reforms were introduced in Australia, Canada, and Sweden.

So there you are,  none of these countries support Sumner's hypothesis that neoliberal reforms lead to faster growth in real income, relative to the unreformed alternative.

The point is that countries that did more reforms are only vaguely distinguishable from than those that did fewer reforms, and their biggest gains came against U.S. recessions and Bush II's generally declining growth in GDP.  I am emphasizing that growth in US living standards slowed after 1973, and I am arguing that a great deal of this slow down occurred because of Republican economic policies.