Updated: “How Stable Are Democracies? ‘Warning Signs Are Flashing Red.’ ” Or Maybe Not.

[Prof. Ronald Inglehart was kind enough to read and comment on this article.  He has correctly noted that data from cohort analysis must be handled differently from the procedures economists mostly use.  I have removed the entire section discussing data.  He also notes that Roberto Foa is now an assistant professor at the University of Melbourne.  I have corrected the biographical sketch to reflect this new information.  For the sake of the historical record, click here for a pdf version of my original article.  You can click here to read Prof. Inglehart’s response to the new Mounk and Foa paper (forthcoming in the Journal of Democracy).]

The quoted text in the title is the headline in a New York Times article November 29. The article reports on a paper that will be published in the January, 2017 issue of the Journal of Democracy.

The authors, Yascha Mounk and Roberto Stefan Foa, believe they have found a set of leading indicators whose values may be able to predict whether a democratic form of government will go into decline in the near future. Here’s the money pull-quote:

Support for autocratic alternatives is rising, too. Drawing on data from the European and World Values Surveys [sic], the researchers found that the share of Americans who say that army rule would be a “good” or “very good” thing had risen to 1 in 6 in 2014, compared with 1 in 16 in 1995.

That trend is particularly strong among young people. For instance, in a previously published paper, the researchers calculated that 43 percent of older Americans believed it was illegitimate for the military to take over if the government were incompetent or failing to do its job, but only 19 percent of millennials agreed. The same generational divide showed up in Europe, where 53 percent of older people thought a military takeover would be illegitimate, while only 36 percent of millennials agreed.

There are many problems with this research. I’ll discuss three of them here:

  1. The research uses, in part, the World Values Survey and European Values Study. It happens that I have quite a bit of hands-on experience with this dataset. And it has issues.
  2. The Journal of Democracy does not adhere to standard academic publishing practices. It is not peer-reviewed.
  3. The authors have no apparent qualifications to conduct empirical research.

I’ll look at these in the order listed above. Hyperlinks should let you jump directly to a section.

The World Values Survey and European Values Study

Ronald Inglehart

Ronald Inglehart

University of Michigan Professor Ronald Inglehart is one of the founders of these surveys. He has posted a response to the authors’ earlier paper here. Or you can just click the screen capture below to see a readable image.

Inglehart Response

(click for larger image)

In the distant past I spent a few years on some research that eventually hit a dead-end. In large part this was caused by the nature of the data in these two data sets.

The first problem is an inconsistency. The wave definitions are different for the World Values Survey (WVS) and the European Values Study (EVS).

Wave Date Ranges

Wave Date Ranges (click for larger image)

The second problem is that not every country is surveyed in every wave. The Times article cites data for six countries. The table below shows the year in which each country was surveyed in each wave. New Zealand is an especially egregious case, with responses in only two of the six waves. (Click here for a pdf file containing the complete list of countries, waves, and years.)

Country Survey Dates

Country Survey Dates (click for larger image)

In cohort analysis, one has to sort out life cycle effects from generational change and period effects.  The latter take into account the fact that the current situation changes from one time point to the next.

Strictly speaking, data used from different years should not be compared directly. Here’s what Prof. Inglehart pointed out in his email to me. ⇒

For example, consider the first wave (1981-1984). In the U.S. this was a period of sharp disinflation with the unemployment rate briefly exceeding ten percent. And, of course, 1995-2000 was the dot com boom. People were likely to be pretty optimistic during this period. Here’s what the Times story says about the periods compared:

Support for autocratic alternatives is rising, too. Drawing on data from the European and World Values Surveys [sic], the researchers found that the share of Americans who say that army rule would be a “good” or “very good” thing had risen to 1 in 6 in 2014, compared with 1 in 16 in 1995.

The global economy was doing pretty well in 1995. By contrast, 2014 was a year of economic stress. People are more likely to turn to authoritarian solutions during periods of economic malaise. That is a more plausible hypothesis for the authors’ results than a declining faith in democracy.  And this is part of what Prof. Inglehart calls “period effects.”  Simply looking at the numbers is not enough.

The Nature of Data

[Section removed per introduction at the top of this piece.]

The Journal of Democracy

Their full paper won’t be published until next month. It will appear in the January, 2017 issue of the Journal of Democracy. It is hardly ever a good sign when researchers publicize their findings in advance of publication.

And there’s even a problem with that. The Journal of Democracy is not peer-reviewed. Here’s their statement on submitting unsolicited manuscripts:

Although the majority of its articles are commissioned by the Editors, the Journal of Democracy also welcomes unsolicited submissions. Every unsolicited manuscript is read and evaluated by at least two members of the editorial staff. Many unsolicited manuscripts are rejected at this stage, often for reasons of focus, timing, style, or space that do not necessarily reflect on their scholarly merits. The Journal is not formally peer-reviewed, but in some cases manuscripts are sent to outside scholars or specialists for comments and evaluation.

Peer reviewing is the gold standard for academic publications. Publishing in a journal that does not use double-blind peer reviewing makes the entire body of research suspect.

The Authors Are Not Qualified

The authors, Yascha Mounk and Roberto Stefan Foa, have Ph.D.s in Government from Harvard. Dr. Mounk is class of 2015, Dr. Foa is 2016. Their pre-doctoral studies focus on philosophy and history. I could find no evidence that either has significant training or experience working with data. But Dr. Mounk sure gave a great pull-quote to the Times:

Yascha Mounk is used to being the most pessimistic person in the room. Mr. Mounk, a lecturer in government at Harvard, has spent the past few years challenging one of the bedrock assumptions of Western politics: that once a country becomes a liberal democracy, it will stay that way.

Curiously, Dr. Mounk is a lecturer in Government at Harvard. But he has his union card. (In academia, a Ph.D. is usually required for admittance to the faculty in most departments.) Why isn’t he on a tenure-track appointment somewhere?

Roberto Stefan Foa is an assistant professor in Political Science at the University of Melbourne in Australia.  This information is from Dr. Inglehart; the University of Melbourne website is sorely in need of updating.


I look forward to reading the full article when it is published. Stay tuned.

Visa Overstays Entry Exit Report FY 2015

Entry Exit Overstay Report FY 2015

Visa Overstays Entry Exit Report FY 2015This is a fascinating little document published every yeqr by the Department of Homeland Security.  The Entry Exit Overstay Report FY 2015 gives details of the number of people who have overstayed their visas broken down by country and in several other ways.  Two fundamental categories are those who overstayed their visa but have left the country and those who have overstayed but have (presumably) not left the country.  The report is available as a pdf file here.

Regular readers know my fondness for Excel.  I decided to do a quick-and-dirty export from pdf to Excel.  Surprisingly the tables seem to have come through in very good shape.  I haven’t bothered to separate them from the text.  Anyone who produces the tables with one table on a worksheet, please let me know.  Give me a copy and I’ll post it, complete with credit to you.  Click here for this invaluable addition to the planet’s stock of information.

Marketplace Disses Trump

Kai Ryssdal

Kai Ryssdal

Kai Ryssdal’s editorial comment at the end of “Marketplace” July 26, 2015. Copied and pasted without editing by your faithful correspondent. Over-the-top phrasing, however, is solely the responsibility of Mr. Ryssdal.

A word about the unemployment rate and this election.

This weekend on CNN’s Sunday show State of the Union, an interview between Jake Tapper and Donald Trump Jr. turned to the unemployment rate, about which Trump said: “These are artificial numbers, these are massaged to make the existing economy look good, to make this administration look good, when in fact, it’s a total disaster.”

To use as straightforward a word as possible, that’s a lie.

Reasonable people can and do disagree about the health of the economy, and how to measure the labor market, but the idea that the Bureau of Labor Statistics [BLS] manipulates the monthly unemployment report is without any basis in fact.

It’s at best a fabrication and at worst, and most damaging, a malicious conspiracy theory.

Same thing goes, by the way, for the Republican nominee’s claim that unemployment in this country is at 42 percent.

This isn’t, to quote Jake Tapper, an anti-Trump position or a pro-Clinton position.

It is a pro-truth position.

Mr. Ryssdal has made a fundamental error. He assumes the BLS gathers the data for the unemployment rate. That is not true. BLS outsources data collection to the Bureau of the Census. And, shortly after taking office in his first term, President Obama moved Census from it’s long-time home in the Department of Commerce to the Executive Office of the President. Which means the administration directly runs the critical data collection job for much of the government. (If you’d like to get some idea of the extent to which the federal government relies on Census data, visit http://census.gov and try to find any single piece of data you’re looking for. The sheer quantity of numbers available is overwhelming. I say this as someone who has spent many, many hours trying to track down data on that site.)

In any case, I’ve written about this before. And I have to remind everyone that John Crudele wrote a series of articles detailing how the data was being faked.  I compiled the articles into a single pdf file which you can read by clicking here.

Having written all this, I have to add one disclaimer.  I do not support Mr. Trump’s candidacy.  I am, in fact, #nevertrump.

My Refusal to Use U.S. Government Data Has Been Extended Indefinitely

RIP Government Data

(click for larger image)

In early 2014 I wrote that I would not use government data again for three more years.  I have now extended that deadline indefinitely.  Neither of the candidates for president inspires more confidence in the data than President Obama.  I have updated the image at the top right of the home page to reflect this fact.  Here’s the updated version →

I will shortly post a companion article about Kai Ryssdal’s failure to understand the issues with government data.

Tax Revenue and the Obergefell Decision

In light of the Obergefell decision issued by the Supreme Court Friday, June 26, I wondered what impact this decision might have on the federal government’s tax revenue. After all, as part of the decision, the IRS will be required to treat all married couples equally, regardless of the genders of the two partners. And, despite many attempts to eliminate it, there is still a marriage tax penalty likely to apply to quite a few of these newlyweds.

As always, my methods are transparent.  Click here to download the Excel workbook.

Executive Summary

The U.S. Census Bureau collects data on same-sex partnerships via its regular American Community Survey (ACS). Wikipedia contains an exhaustive history of global changes in same-sex marriage laws, including the dates on which same-sex marriage became legal. Luckily, they include individual U.S. states. Finally, the Tax Policy Center has a handy web page that calculates the penalty (or bonus). Pulling together these disparate elements, my estimate is that the federal government will collect about $375 million in additional tax revenue. But there are many, many assumptions and calculations behind that number. I strongly recommend reading this entire article so you can see what I did and how you might want to change my assumptions.


First, I gathered the data on same-sex couples from the American Community Survey. The data is broken down by age, income, and the usual demographic elements. But what interested me was the breakdown by state. After all, some states (27 to be precise) had legalized same-sex unions before Obergefell made it the law of the land. I combed the history provided by Wikipedia. The ACS created “Tab 3” in the Excel workbook, but I added some information, specifically which states fall into which groups (legal and not legal pre-Obergefell) and the month and year on which each of the 27 states approved these marriages.

If only it was that easy. Until this year the IRS did not allow same-sex couples to file joint tax returns. Therefore, all those who married before Obergefell will also be subject to the marriage tax penalty.

The marriage tax penalty was calculated using a handy web page at the Tax Policy Center. I have copied and pasted both the inputs and outputs into the Excel workbook (the MarriageTaxPenalty tab). However, users should beware: this is Excel, not Turbo-Tax. If you want to change my assumptions about income, deductions, or other items you’ll need to use the website. The only number on that tab that feeds into the rest of the calculations is cell E30, the marriage tax penalty.

Calculating the number of same-sex couples who were married pre-Obergefell is straightforward. The ACS reports the number of same-sex households (by state). They also report the “Percent of all same-sex households who are same-sex spouses.” Multiply the number of households by the percentage married gives a total. Note that this applies to all 50 states. Many states in which same-sex marriage was not recognized pre-Obergefell reported a significant percentage of same-sex married couples. California, for example, reported that 37.35% of same-sex households were married couples. This is actually above the national average of 34.00% in states where same-sex marriage was legal![1] That creates certain problems in the data analysis which, fortunately, are easily overcome.

The real issue is the number of unmarried same-sex households that will decide to marry in light of the Obergefell decision. To calculate that figure, I first calculated the expected increase in the percentage of same-sex couples that would get married. I used the difference between the national average in legal states (34.00%) and the actual percent married in each state. For states like California in which the actual percentage was greater than 34.00%, I entered zero. I then multiplied the expected increase in percent married by the percentage unmarried, giving the expected percentage of unmarried couples likely to marry. Finally, I multiplied that percentage by the total number of same-sex households. After all that work, it looks like about 8,092 couples in states where same-sex marriage was not legal pre-Obergefell will decide to get married. The impact on tax revenue is a paltry $11.7 million.

But there are 251,695 same-sex households that are already married. Until now these folks have not been subjected to the marriage tax penalty. In total, government tax revenue will increase by about $375 million as same-sex couples get to experience the same treatment that different-sex couples have endured practically ever since the graduated income tax was introduced. Welcome!

The Marriage Tax Penalty

The marriage tax penalty is inevitable as long as tax rates are progressive. In a progressive tax system your marginal tax rate rises as your income rises. That means two people, each earning $50,000 per year, will have a joint income of $100,000 per year and (most likely) pay a higher tax rate.

Conclusion and Warnings

This is really a quick-and-dirty look at a complicated issue. Here are a few of the many issues I did not take into account.

  1. State income taxes are deductible from income at the federal level for those who itemize deductions. California, notably, has a high state income tax as well as many same-sex households (107,991 is the ACS number). That will probably reduce my estimates somewhat. Mitigating this is the fact that the Tax Policy Center’s calculator asks about state income taxes.
  2. Looking only at the states in which same-sex marriage was legal pre-Obergefell, there is an interesting (and significant) downward trend. States that legalized same-sex marriage in 2009 had a 50% marriage rate. This falls to 30% for states that legalized in 2014. A trend like this is to be expected. After all, unmarried same-sex couples in states that legalized in 2009 have had five years to make plans and decisions. A cartoon making the rounds shows a same-sex couple. One is saying to the other, “You said you’d do it once it was legal.” Exactly. Negotiating a marriage proposal and working out the details takes time. The Excel workbook contains an SPSS linear regression that may be helpful. On average, for each additional year same-sex marriage is legal, the percentage married increases by about two percentage points. Note, however, that the fit is far from exact (to put it mildly).(Warning: if you try to duplicate my results you need to use the SPSS coding for Julian dates. For your convenience, these are shown in the Excel workbook on the tab LegalStatesOnly.)Trend in Marriage Rate
  3. I’m certain that there are many issues related to income distribution and incentives that I have overlooked. I hope to hear from some readers and will happily publish alternative analyses.

I started this as what I thought would be a simple exercise. It has taken me the better part of two days to finish it. My lovely wife tells me this sort of thing is good for my brain. I hope she’s correct.

[1] Another interesting point: the average percentage married in legal states was 34.00% while the corresponding percentage in illegal states was 36.58%.

Global Interest Rates Update

Late last year I wrote about global interest rates. I attempted to rationalize the eye-popping disparity between U.S. and foreign government bond interest rates. Today, doing some work for a class I’m teaching, I once again noticed something pretty amazing. U.S. ten year notes have a yield of 2.017%. Germany and Japan are in the vicinity of 0.3%. That will get anyone’s attention.[1] This is your global interest rate update. As always, my methods are transparent. Click here to download the Excel workbook (includes both monthly and annual data).

Before proceeding with the update, I must confess a minor error in my previous article. My equation for the default risk premium was much more complicated than necessary. I’ve corrected my mistake below.


So first let’s look at interest rate trends. For some reason, Japan’s data is only reported through October, 2014 in the OECD dataset. I’ll exclude that country just to keep the data comparable by using a single source.

Here are a few representative countries. As always, the yield to maturity shown is on the ten-year government note. I assume (hope?) that OECD has corrected for coupon payments in putting these numbers together. This time I have included only one eurozone country, Germany.


Interest Rates

Monthly Interest Rates, 2014 (click the image to enlarge)

Notice something very interesting. The data falls into three rough groups. (Russia, the top line, is a special case.) The first group with interest rates around 4% includes Chile, New Zealand, and Australia. It’s fairly easy to understand why New Zealand and Australia might move together. A simple gravity model of trade will tell you these economies should be closely linked. But Chile? Is there a strange southern hemisphere effect? The data is there — I leave it to more intrepid researchers to figure this out.

The second group consists of Israel, the U.K., the U.S., and Canada. These four countries are linked by a shared language and historical ties. And, of course, the U.S. and Canada have both the gravity model and NAFTA to encourage economic linkages.

The lowest two countries are Germany and Switzerland. In fact, in January, 2015, the yield on Swiss government bonds was -0.07%. This is the nominal rate. Savers are paying the Swiss government for the privilege of lending it money.

Expected Future Exchange Rates

We can explain some of these interest rate differentials by using the uncovered interest parity (UIP) model: [2]

Uncovered Interest Parity

This equation simply says that the yield to maturity on U.S. bonds should equal the yield on euro bonds plus the expected future depreciation of the dollar vis-à-vis the euro.[3] When the dollar depreciates, one euro will buy more dollars. Thus a dollar depreciation increases the required return on U.S. securities, causing their interest rate to rise.

Here’s the expected depreciation (+) or appreciation (-) of various currencies vis-à-vis the U.S. dollar. I’ve excluded countries that did not have data for January, 2015. Some of the depreciations are truly remarkable. Not surprisingly, Russia is the leader in this group with expected annual depreciation of 23.85%. Iceland is also notable with 14.54% annual depreciation.

Expected Exchange Rate Changes

Expected Exchange Rate Changes

Currencies expected to appreciate sharply include the Swiss franc and the Swedish kronor. The eurozone countries are interesting: Belgium (−11.3%), Finland (−12.6%), Germany (−14.90%), the Netherlands (−13.60%) and Spain (−3.36%) all use the same currency. But the expected rates of depreciation are markedly different.

Default Risk

The only way to make sense of this is to introduce another variable: default risk. We can calculate the default risk premium as the difference between yields on government bonds issued by different governments but denominated in the same currency.

The usual explanations for interest rate differentials among various countries are (1) differential inflation rates and (2) expected future exchange rate changes. Naturally the two are related. However, expected future exchange rate changes also incorporate a variety of other risks, including exchange rate risk and default risk. The appropriate short-run model is a slight modification of UIP.

Default Risk Equation 

In this equation, i€P is the yield on government securities for countries in the P group.[4] Similarly, i€G is the yield on German government securities. In the eurozone, German government securities are the benchmark against which other countries are measured.

What, then, is the meaning of the expected change in the exchange rate, the term on the right side of the = sign? The Greek euro cannot depreciate against the German euro. In this case, we are measuring the default risk premium of the various countries.

To measure this I’ll use data from December, 2014 because all the eurozone countries have data reported for that month. Also note that we apparently need to change the acronym. Markets seem to think that Ireland has worked their problems out. But Slovenia is now in the mix. I propose PIGSS, mainly so I don’t have to change a bunch of subscripts in this article.

Default Risk Premiums

What’s really interesting about this list is that only Luxembourg approaches Germany’s level of stability. And France’s default risk premium is higher than you might expect for the second-largest economy in the eurozone.


Wasn’t that fun? I hope you’ve learned a bit about using data to learn something interesting.

[1] OECD (2014), “Finance”, Main Economic Indicators (database). DOI: 10.1787/data-00043-en (Accessed on 27 February 2015)

[2] Material on UIP is based on chapter 4 from Robert C. Feenstra and Alan M. Taylor, International Macroeconomics (2012). Worth Publishers, New York. Disclaimer: I wrote the instructor’s resource guide for this textbook.

[3] It has become conventional to express exchange rates as units of home currency per unit of foreign currency. That means an increase in the exchange rate implies a depreciation of the home currency and vice-versa. This is only confusing the first 247 times you work with it. After that it’s simple.

[4] P stands for PIIGS, the five countries in the eurozone that have very risky government debt.

U.S. Government Debt Revisited

Today’s article is prompted by another idiotic statement from the White House coupled with a few very stupid responses on Twitter. I’m encouraged, however, by the number of intelligent, correct replies. Today I’ll discuss the U.S. government debt and deficit revisited.

You can save some time by reading this first. The article discusses the basic accounting relationships between a deficit, the debt, and the quantity of money in circulation. This stuff is not difficult. The fact that so few people do not understand it is a testimonial to the sad state of economic and financial literacy in the U.S. And, as always, my data is transparent.  Click here to download my Excel workbook.

The Tweets and Responses

I noticed this little controversy in the summary at Twitchy.com. It begins with a tweet from the White House:

White House Tweet

White House Tweet

Twitchy summarized the readers’ comments starting with this one (which is 100 percent correct):

First Tweet Response

First Tweet Response

But if you read to the end you’ll find this (which asks for the impossible):

Second Tweet Wrong

Second Tweet Wrong

A Review for Those Who Did Not Click the Link

Now I know many of you will not click the previous link. (If you did, click here.) So here’s an ultra-short summary. Although I’ve described this using the government as an example, the relationship between debt and a budget applies to individuals, businesses, and every other entity.

The government deficit is the difference between government spending and government revenue. Deficits are financed by issuing bonds. Those bonds make up the government debt. The relationship is simple. When the government runs a deficit it issues enough bonds to cover the difference between spending and revenue. Those newly-issued bonds add to the national debt. When (if?) the government runs a surplus, revenue exceeds spending. The “excess” revenue is used to buy back some of the outstanding debt. Thus we have the fundamental relationship between debt and the budget.

Equation 1

where Dt is the debt at the end of year t, Dt-1 is the debt at the beginning of the year, and Bt is the government budget during the year. If the government runs a surplus, Bt is positive and the debt falls. If the government runs a deficit, Bt is negative and the debt rises.

For example, if the government debt at the beginning of the year is $17 trillion and the government runs a deficit of $1 trillion during the year, the debt at the end of the year will be $18 trillion.

Note that none of this affects the money supply. The government borrows because it needs money to spend. The money that buyers pay for the bonds is almost immediately spent by the government, leaving the quantity of money in circulation unchanged.

In developed economies there is a central bank that determines the money supply. In the U.S. the central bank is the Federal Reserve system (“the Fed”). In many European countries the central bank is the European Central Bank (ECB). The central bank attempts to control the money supply by purchasing or selling government securities. When the central bank buys securities, it pays for them with newly-created money.[1] If the central bank decides to reduce the money supply it sells securities. Bingo, there is less money in the hands of the public.

Where does the newly-created money come from? The central bank creates it. Central banks can decide to create or destroy money based entirely on the decisions of people assigned to make those decisions. They do not worry about gold, silver, yak butter, sharks’ teeth, or the infamous stone money of Yap. They only worry about the current and future state of the economy.

Now You’re Ready for Current Data

It happens that another project led me to update the data from my 2010 article. The data is from the U.S. Department of the Treasury, Office of Debt Management, Office of the Under Secretary for Domestic Finance Table OFS-2 as of March, 2014.

Debt Ownership March 2014

If you’re more comfortable with a chart, here it is:

Debt Ownership


Compare the above with data from my 2010 article:

Debt Ownership July, 2010

Foreign ownership is large and growing.

“But,” some people say, “our GDP is very large.  What about the debt-to-GDP ratio?”

Debt To GDP

(click image for a larger version)


We don’t need to worry about the U.S. government debt because it’s mostly owed to U.S. residents.

When I first studied economics several decades back, textbooks commonly made this statement →

That simply meant that very little debt was owned by foreign entities. As the tables above clearly show, that is no longer true. About one-third of U.S. debt is owned by foreigners.


Honestly, this material is not difficult and the data is readily available. I always have trouble understanding why people don’t simply look at the facts instead of trusting the media. Nearly everyone in the media, including most of those who report on economics, are illiterate about even the most basic economic concepts. Remember, their college degrees are in journalism or communications. Somewhere along the way an editor decided they knew some economics. They don’t.

[1] Please don’t give me a hard time about this simplification. I know the purchase actually creates bank reserves. I even understand why the money supply mechanism in the U.S. is broken today.


The Mystery of Third Quarter GDP

Executive Summary

There has been speculation that a substantial part of the growth in third quarter GDP was caused by subprime auto loans. According to the third estimate, real GDP grew 5.0% at an annual rate. In this article, I use data from the Bureau of Economic Analysis, Experian Automotive, and Equifax to estimate the actual impact. Using Experian’s estimates for leases, loans, and risky lending I conclude that the contribution to third quarter GDP growth from subprime lending was about 0.05 percentage points. I also note that there are subprime loans for recreational vehicles. However, I could not find any data. I assumed the automobile percentages applied to the RV market as well.

A second source is a speech given by Joy Wilder Lybeer, a senior vice president at Equifax. Ms. Lybeer claims that one-third of auto loans are subprime. This is in stark contrast to Experian’s estimate of 10.57%. Using Ms. Lybeer’s estimate, the contribution to third quarter growth from subprime lending was 0.13%. Using BEA standard rounding technique, third quarter growth would have been 4.9% instead of 5.0%. That’s a noticeable difference. But 4.9% is nothing to sneeze at.

There are two areas that pose far greater problems. The first is the persistent growth of inventories. Whether that reflects business optimism about the future or lower than expected spending today is a topic for debate.

The second is the burst of growth in intellectual property investment. This is a fairly new addition to GDP and, as many have noted, a category that is fraught with estimation difficulties. Let me just say that between inventory and IP growth, the 5.0% overall growth rate looks a bit shaky. But it’s not because of subprime vehicle loans.

The Mystery of Third Quarter GDP


The third (“final”) estimate for U.S. real GDP growth in the third quarter was a whopping 5 percent. There have been various reasons cited for this astonishing performance. One explanation is the sluggish economic performance in the first quarter created pent-up demand that exploded over the summer months. Another is the boost from sub-prime auto lending. I’m going to once again violate my promise to not use government data to dive into the GDP numbers and see if there’s anything there. Except as noted, all data is from the U.S. Department of Commerce, Bureau of Economic Analysis. [1]

As always, my methods are transparent. Click here to download the Excel workbook with my calculations.

Gross Gross Domestic Product

At the gross level, let’s first look at the growth rate compared to the second quarter. Overall growth was 4.97%. The components with the largest quarter-to-quarter growth were:

  • Consumption spending on durable goods.
  • All components of gross private domestic investment except business structures, residential construction, and the inventory change.
  • Federal government defense spending.

GDP Growth The Mystery of Third Quarter GDP

(Click image for larger version)


The growth of durable goods spending bears further examination because that’s where automobile sales are buried. I’ll look at the details in the following section.

Gross Gross Domestic Product: Component Growth

The robust growth of business investment is a welcome improvement from the first quarter’s anemic levels. However, the change in business inventories was +$82.2 billion dollars, down slightly from the second quarter’s +$84.8 billion. This is somewhat worrisome because inventory change has been above +$80 billion for four of the past five quarters:

IP Investment The Mystery of Third Quarter GDP

(click image for a larger version)

As Keynes noted, the real issue is the extent to which this inventory growth is intentional (“planned inventory change”). Part of the growth is undoubtedly planned by businesses expecting sales to grow. The rest is unplanned, caused by current sales falling short of production. Planned inventory increases are good because they reflect business confidence. Unplanned inventory increases are bad because businesses are likely to cut production to reduce unwanted inventories. But inventory change fluctuates quite a bit from quarter to quarter, so this is probably not much of a concern.

Inventory change The Mystery of Third Quarter GDP

(click image for a larger version)

Ever since intellectual property investment was added to U.S. GDP, some economists have worried about the difficulty of obtaining reliable numbers — and therefore the temptation to use this category to fudge the results. The figures for 2014 do nothing to quell those suspicions. In the third quarter IP growth was 8.83% (annualized). And this continues a trend that began in 2o13:III.

IP Investment The Mystery of Third Quarter GDP

(click image for a larger version)

Ironically, the final major contributor to quarter-to-quarter growth was federal defense spending which grew by 16.01%. Tomahawk missiles at $1 million each probably accounted for quite a bit of this growth, followed by drones and their armaments.

Gross Gross Domestic Product: Contributions to Growth

A second way of disassembling GDP growth is by looking at the contributions to the current quarter’s growth. Luckily, the BEA produces this breakdown in their Table 1.1.2. Once again, consumption spending leads the way with non-residential private investment in second place. At the detail level, consumption of durable goods, investment in equipment, and government defense spending were big contributors.

Contributions to GDP growth The Mystery of Third Quarter GDP

(click image for a larger version)

Fine Gross Domestic Product

What we’ve seen so far gives us guidance about where to explore further. Specifically, what drove that increase in consumer durables spending? The BEA has the answers once again. Tables 1.5.1, 1.5.2, and 1.5.6 have the data we want. Let’s start by looking at the components of consumer durable spending.

Consumer Durable Spending Details

First, let’s look at the quarter-to-quarter growth rate (annualized). Sure enough, motor vehicles and parts grew 11.3% with “recreational goods and vehicles” contributing a whopping 15.7%. An immediate question comes to mind: how much of that growth was caused by subprime lending? (We’re verging on some sort of risk-adjusted GDP here, but I’ll put that exercise off for another day.)

Growth of consumer durables spending The Mystery of Third Quarter GDP

(click image for a larger version)

Naturally we also need to look at the contributions to that 5 percent GDP growth:

Contributions to GDP growth The Mystery of Third Quarter GDP

(click image for a larger version)

We’ll come back to car and RV sales in the next section. But now I want to look at gross private domestic investment.

Gross Private Domestic Investment Details

First, here are the annualized quarter-to-quarter growth rates:

Growth of gross private domestic investment, detailed The Mystery of Third Quarter GDP

(click image for a larger version)

And the contributions to that 5 percent growth:

Contributions to growth, gross private domestic investment The Mystery of Third Quarter GDP

(click image for a larger version)

Industrial equipment, transportation equipment, and IP were the main contributors from this category. Transportation equipment? Most likely railroad cars to carry the oil that won’t fit through existing pipelines. I’ve already expressed some doubt about the veracity of the IP numbers so I won’t repeat that argument. Also note that despite a large quarter-to-quarter growth rate, computers and peripheral equipment was a small contributor to the overall growth rate. That is probably because this category was only 0.44% of third quarter nominal GDP.[2]

Automobiles, Recreational Vehicles, and Subprime Loans

Experian follows automobile financing quite closely. There are two sources of financing for new vehicles: loans for purchasing and leasing. Experian has broken the data into five risk groups: Super prime, prime, nonprime, subprime, and deep subprime.[3] Here’s the breakdown:[4]

Experian loan quality data (new cars only) The Mystery of Third Quarter GDP

(click image for a larger version)

Leasing by risk segment (new cars, Experian data) The Mystery of Third Quarter GDP

Leasing by risk segment (new cars, Experian data)

Loan volume by risk category (new cars only, Experian data) The Mystery of Third Quarter GDP

Loan volume by risk category (new cars only, Experian data)

Experian also notes that 84.8% of new car sales are financed.[5] Of that, 29.14% are leased.[6] It’s easy to calculate the percentage of overall sales that were financed by loans versus leasing:


Sources of financing, new car sales The Mystery of Third Quarter GDP

(click image for a larger version)

Now we can get some work done. Assume that these percentages apply to both auto and recreational vehicle financing. Those two categories contributed 0.59% to the 5.0% third quarter growth. Of that 0.59%, 0.15% was leasing and 0.35% was loan financed. (The remaining 0.09% of sales were not financed, presumably cash purchases.) Multiplying by the percentage of leases and loans made to subprime and deep subprime borrowers and adding leases and loans gives an overall contribution to GDP growth of about 0.05%.

Experian results The Mystery of Third Quarter GDP

(click image for a larger version)

However, there is another estimate from Equifax. In an interview, Ms. Joy Wilder Lybeer said,

“Don’t take the bait on claims that there will be a bubble,” says Joy Wilder Lybeer, a senior vice president at Equifax, a credit-reporting firm. “Subprime is as healthy as it’s ever been.”

“One in every three cars sold goes to a subprime borrower, and those loans are performing well,” she says, citing historically low delinquencies, defaults and repossessions as a percentage of outstanding loans.[7]

Well, that’s quite a difference. One third compared to 10.57% is not insignificant. Fortunately, it’s easy to calculate the new numbers. I‘ll assume the lease percentages are those estimated by Experian.

Equinox results The Mystery of Third Quarter GDP

(click image for a larger version)

So without subprime lending, third quarter growth would have been 4.9% instead of 5.0% (using BEA rounding conventions). That’s a bit of a difference, but 4.9% growth is nothing to sneeze at.


While subprime vehicle loans contributed to third quarter growth, the total contribution was minor. Far more concerning is the continuing growth of inventories and the apparent burst of growth in IP investment.


[1] All BEA data is from the NIPA Interactive Data site at http://bea.gov/iTable/iTable.cfm?ReqID=9&step=1#reqid=9&step=1&isuri=1. Accessed late December, 2014 and early January, 2015.

[2] I used nominal GDP because no estimate is available for chained real GDP. See Excel worksheet RealGDPdetailChained especially footnote 4: “4. The quantity index for computers can be used to accurately measure the real growth of this component. However, because computers exhibit rapid changes in prices relative to other prices in the economy, the chained-dollar estimates should not be used to measure the component’s relative importance or its contribution to the growth rate of more aggregate series; accurate estimates of these contributions are shown in table 1.5.2 and real growth rates are shown in table 1.5.1.”

[3] Melinda Zabritski, Sr. Director Experian Automotive, ” State of the Automotive Finance Market, Third Quarter 2014.” http://www.experian.com/assets/automotive/white-papers/experian-auto-2014-q3-presentation.pdf?WT.srch=Auto_Q22014FinanceTrends_PDF Accessed December 31, 2014.

[4] Ibid., slides 17 and 28. Note that deep subprime leases are calculated by me to make the sum 100%.

[5] Ibid., Slide 14.

[6] Ibid., Slide 16.

[7] http://wardsauto.com/finance-insurance/subprime-just-fine-dire-predictions-aside. Accessed January 3, 2015.

The BLS Has a Problem

The latest jobs report (December 5, 2014) includes outright contradictions. According to the New York Post’s John Crudele, a comparison of the seasonally adjusted (SA) and non-seasonally adjusted (NSA) data is, um, instructive. Let’s look at the data first:

BLS data November 2014 The BLS has a problem

The ratio of the seasonally adjusted to non-seasonally adjusted data should be pretty close to the same in two consecutive years. But it’s not. Even worse, the NSA figures decreased between 2013 and 2014. But the SA numbers increased. There’s something very, very wrong here. (Click here to download my Excel workbook.)

While trying to track down the NSA figures for 2014, I ran across this table: ESTABLISHMENT DATA Table B-1. Employees on nonfarm payrolls by industry sector and selected industry detail [In thousands]. Clearly subtracting the October figure from the November number should equal the 1-month change. And it does, confirming Mr. Crudele’s data.

Long-time readers will recall my refusal to use government statistics until 2017. This is yet another example of malfeasance at BLS. Future economists will be forced to exclude the 2013 – 2016 time period from their research because the numbers are not believable.

Thanks to John Crudele and Twitter satirist @hale_razor whose Tweet brought this to my attention.

The BEA News Release Archive

As noted in the corrected version of my article on healthcare spending and first quarter 2014 GDP, I accidentally deleted the Excel workbook that had the data and calculations. Every cloud has a silver lining. In this case I got to learn about the BEA News Release Archive.

As far as I can tell, every BEA press release since at least 2002 is archived here. Once you’ve selected the kind of data you want (Figure 1) and the specifics (Figure 2), the press release appears. Look carefully at the right sidebar and you’ll see a link to Tables Only (Figure 3). That link downloads the entire dataset as a single Excel workbook. You will have to work a bit if you want to reconstruct, say, Table 1.5.6. And the data is in a rather annoying format with some row titles occupying two rows (Figure 4). Those blank rows can be a real pain. Excel thinks <blank> − <blank> = 0.0. I left those blank rows in the workbook just so the data would be consistent across all four worksheets.

BLShistoricalDataMenu1 The BEA News Release Archive

Figure 1


Figure 2 The BEA News Release Archive

Figure 2


Figure 3 The BEA News Release Archive

Figure 3


Figure 4 The BEA News Release Archive

Figure 4


I’ve complained a lot about economic data being provided by the U.S. government. And I’m still quite skeptical of many of the numbers emanating from Washington, D.C. But kudos to the BEA for data transparency.