Answer to gig economy question depends on baseline

The US Bureau of Labor Statistics reported on Thursday that contingent and alternative work was occurring at about the same rate in May 2017 that it did in February 2001. Even more surprisingly, some categories of this type of work were actually occurring less frequently in 2017 than in 2005. Media coverage of the report noted the lack of growth in these categories of work and the surprise among those who expected confirmation of a growing gig economy. One take that has been missing, however, is the possibility that other changes to the labor market mean we should expect much less contingent and nonstandard work. Two indicators that nonstandard work has been falling come from regularly monthly data showing the long-term decline in self-employment and multiple job holding. Relative to this alternative baseline, the BLS report shows a labor market that has been changing in dangerous ways.

Definitions

“Contingent” work is comprised of jobs that are not expected to last, even if there is no change to the economy. In a related concept, “alternative” work arrangements include independent contracting, on-call work, and working for a temp agency or contract company while assigned to one client’s place of business. The structure of these jobs varies but they all have a nonstandard employment relationship; independent contractors may provide their labor with less intermediation than the standard worker while contract company and temp agency employees operate through more labor market intermediaries. While only some alternative work arrangements are contingent, all of these categories might be associated with a less-smooth income stream.

Importantly, despite the popularity of treating “contingent”, “alternative”, and “gig” work as synonymous, BLS defines gig work very narrowly as only those jobs that include work found and paid for through a website or mobile app. BLS has not yet released any estimate of gig work, but plans to do so by September 30, in a Monthly Labor Review article. The gig work questions ask people whether they have done ANY gig work, whereas the contingent and alternative work questions focus on the main job.

Interpreting the survey results relative to an alternative baseline

With lots of discussion about Uber drivers, shortened job tenure, and the fissured workplace, analysts were expecting to see a growing share of US workers employed in contingent and alternative work arrangements. However, this expectation seems to forget major demographic changes since 2005. The overall employment rate has fallen since 2001, particularly among young people and students, who are traditionally disproportionately “contingent” workers. Simultaneously, people are now much more likely to have a college degree. Since contingent workers tend to be younger, and those with college degrees tend to have more stable employment, a baseline assumption about the labor force, based on demographics alone, would suggest that nonstandard work arrangements become less common, not more common.

Two indicators that show this changing baseline are the incidence of unincorporated self-employed workers and the rate of multiple jobholding. Both measures are proxies of contingent and alternative work and have fallen by a third or more since the first contingent worker supplement was conducted in February 1995.

Fewer unincorporated self-employed

It’s important to remember that the analysis released by BLS on Thursday came from a one-time supplement to a survey that is conducted every month. If, during the week before one of the monthly interviews, someone drove for Uber, provided labor through Task Rabbit, or worked one of many non-gig independent contractor jobs, as either their first or second job, they will show up as “self-employed, unincorporated.” As expected by demographics, the share of the population that is classified this way has fallen since 1995 for both those age 25 to 54 and for those age 55 or older (figure 1).

seu

In 1995, 7.3 percent of the age 25 to 54 population was unincorporated self employed, compared to 6.3 percent in 2001, 6.2 percent in 2005, and 4.8 percent in the year ending May 2018. The share of those over age 55 who are self employed and unincorporated has also fallen, from 5.1 percent in 2005 to 4.3 percent over the past year.

Since my calculation above includes those who are self employed as their second job, it helps to rule out the possibility that the contingent and alternative workforce numbers are misleadingly low because they do not capture the second job. Additionally, using a moving annual average of the data helps to smooth out the volatility that is possible when looking at any single month of CPS data.

Falling multiple jobholder rate

A second indicator supporting a lower baseline for comparison of recent BLS estimates comes from the fall in the share of workers with more than one job (figure 2).

mjh

The share of employed with more than one job has fallen from 6.6 percent in 1995 to 5.3 percent in 2005 and to 4.7 percent in the year ending May 2018. Among women, the trend is less pronounced, with 6.4 percent of female jobholders reporting more than one job in 1995, compared to 5.7 percent in 2005, 5.6 percent in 2017, and 5.4 percent over the past year. Since some explanation of the lower-than-expected BLS estimates has focused on concerns over measurement, and specifically that second jobs are not captured, it is worth pointing out how much less common second jobs now are in the CPS, compared to the period when the earlier survey supplements were conducted.

Policy should focus on people not work

When adjusting for increased education, an aging workforce, and changes to overall employment rates, incidence of contingent and alternative work actually seems to have grown relative to its 2001 rate. In age and education, those who are currently working are more demographically similar to those who traditionally hold standard jobs, with the potential exception of some independent contractor jobs. When compared to a baseline where, all else equal, contingent and alternative work should actually have fallen over the past 12 years, it is concerning to see this type of work retain its share of the labor market.

One explanation for the relative persistence of nonstandard work is that many of the jobs that used to be temporary or unstable have been either outsourced, automated, or lost to the housing bubble, while jobs that used to be stable and standard are being destabilized. The influence that companies in places like Silicon Valley have over labor policy has grown, and these companies seem to be using their influence to try to make otherwise unprofitable ventures profitable, at the expense of workers. Partially as a result of corporate influence, and in the context of other changes in the US labor market, the main difference between the 2001 and 2017 contingent worker supplement results may be that the less stable jobs of today take a lot more student debt to get.

Perhaps more importantly, though, it is off-target to focus on the share of the economy currently claimed by contingent and alternative work. Financial statements already show that many of the new labor market intermediaries are not profitable. The focus should instead be on policy that tries to protect and improve job quality and to prevent gaps in unemployment insurance from needlessly putting people in poverty during the next economic downturn. The ACA, as one relevant policy example, may have been blessed by the 2017 survey data: health insurance among contingent and alternative workers is now much more common.

 

 

Advertisements

Sawing off the top of a step ladder

Policymakers may have accidentally demonstrated that they have the skills needed to curb the pay of the top earners in a profession. Unfortunately, these problem solving skills were not applied to CEOs, who regularly get paid a thousand times what their middle-paid worker makes. Instead, some state lawmakers have tackled the non-existent problem of teachers making too much money.

In 2016, public school teachers in the states with recent strikes or large scale protests (West Virginia, Oklahoma, Arizona, Kentucky, North Carolina, and Colorado) are paid about three dollars per hour less than other teachers nationwide. This is not new. After adjusting for prices, the gap was about the same in 2006. What has changed is teacher pay near the “top”. In 2006 a public school teacher making $40.93 per hour (in 2016 dollars) makes more than 90 percent of public school teachers nationwide, while it takes an hourly wage of $37.95 to make more than 90 percent of teachers in states with strikes and protests. The nationwide 90th percentile wage ticked up to $43.27 by 2016, but fell to $34.98 for the states that have subsequently organized strikes and protests.

teacherpay

The top half of the public school teacher wage distribution is already low compared to other occupations. But after state efforts to curb teacher pay, 90 percent of the public school teachers in states with strikes and protests make less than $35 an hour, and more than a quarter of the group is paid less than $16 an hour. These hourly wage figures are based on annual earnings in the March CPS, so they already account for differences in weeks worked per year and hours worked per week, and are therefore more comparable to the hourly wages of other occupations. Using these wage values, teachers, relative to other occupations, have the additional burden of being paid, on average, for fewer hours per year, making it harder for them to make ends meet.

Policymakers should reverse the misguided efforts to cut teacher pay. If anything, teachers should be better rewarded for their contribution to society. It shouldn’t take strikes and protests to make this point.

Technical note: The box plot above shows the average (green diamond), median (white line), 90th, 75, 25th, and 10th percentile hourly wage. Public school teachers are identified as persons in occupations 2200 to 2340 who are employed by the government. The hourly wage variable is generated as the ERN-VAL / (WKSWORK * HRSWK). 

Can a tight labor market pull young people back to full-time work?

From 2001 to 2013, college enrollment in the US increased rapidly. Young people delayed the full-time work portion of their lives, ostensibly to seek skills for an evolving set of jobs. Trends since 2014, however, suggest that the increase in education is not an entirely structural change. Over the past few years, a tighter labor market has been leading more young people to full-time work, suggesting that a weak labor market, rather than concern over a changing set of jobs, was pushing some young people to enroll from 2001 to 2013.

The Current Population Survey shows that from 1994 to 2001 young people were increasingly taking full-time jobs. By March 2001, an annual average of 40.5 percent of people age 19 to 21, and 61.5 percent of people age 22 to 24, were employed full-time. Following the recession of 2001, the trend reversed completely. By March 2013, an annual average of only 23.3 percent of those age 19 to 21 had full time jobs, and only 46.0 percent for those age 22 to 24.

fte_by_age

While data show a long-term decrease in full-time employment among younger people, the same data also show two signs that a weak labor market plays a role. First, the full-time employed share of both age groups has increased steadily since 2014. As of the year ending in March 2018, 28.3 percent of those age 19 to 21, and 52.2 percent of those age 22 to 24, are employed full-time.

Second, if education were driving the change in youth employment then summer youth employment would increase relative to school-year employment. Since most US schools offer students a summer break, the share of young people employed full-time during June, July, and August is much higher than at any other point during the year. If young people are opting for school because of abundant attractive new jobs, then summer break might provide an increasing chance for full-time employment for those in school during the rest of the year. However, summer full-time employment has actually fallen relative to school-year employment, suggesting the opposite, that young people are having a harder time obtaining summer jobs compared to the period from the mid 1990s through 2001.

The weak labor market therefore plays a role in encouraging school enrollment; those who are able to do so can wait out a poor set of job opportunities while gaining new skills. It’s important to note, however, that the tighter labor market since 2014 seems to have slowed, but not reversed, the previously rapidly increasing share of young people who are not in the labor force, and do not want a job, because they are enrolled in school.

nilf_by_age

In 1994, for about half of those age 19 to 21 who are not in the labor force (meaning they are not employed and have not been looking for work) the reason was school enrollment. The rest are often staying home to take care of family, disabled and unable to work, or sometimes believe that there is no job available. Since then, school enrollment has rapidly increased labor market non-participation. By the latest data, covering the year ending in March 2018, more than two thirds of non-participation is due to school enrollment. Among those age 22 to 24, the rate climbs from about one third in 1994 to about one half in the latest year of data. School-related non-participation growth among the slightly older cohort has largely stalled since 2013, while growth has slowed some among the younger group.

Much of the overall drop in full-time employment among young people comes from school enrollment. At least some portion was motivated by the particularly poor quality of the labor market during the years from 2001 to 2013, with students hoping to emerge more competitive and ideally at a time when job applicants have more bargaining power. It remains to be seen how much the Fed will allow the labor market to tighten  and how full-time employment and school enrollment will respond among younger people.

College degree boom and a changing labor market baseline

The March FOMC meeting minutes, released today, noted that increased education is changing the unemployment baseline:

A few participants warned against inferring too much from comparisons of the current low level of the unemployment rate with historical benchmarks, arguing that the much higher levels of education of today’s workforce—and the lower average unemployment rate of more highly educated workers than less educated workers—suggested that the U.S. economy might be able to sustain lower unemployment rates than was the case in the 1950s or 1960s.

It’s worth reiterating the Fed point with another labor market example. From the CPS, the median weekly earnings of women between the age of 25 and 54 increased steadily from 1997 to 2003 and again from 2014 to present.

uwe

But women with a college degree in this age group did not see the more recent increase. Instead, their earnings are nearly $50 per week below the 2003 peak, after adjusting for inflation.

uwe_ba

The same wage series for those with less than a bachelor’s degree and those with advanced degrees also peak around 2003. The increase in education is what explains overall median wage growth. Because college educated workers earn more than workers without a college degree–and because so many more people have obtained college degrees compared to 2006–median wages for the entire group increase.

The increase in the share of this group with a college degree is so astounding that the population of 25-54 year old women without at least a bachelor’s degree has been falling rapidly since 2007.

pop_ltba

Recent overall wage growth is not driven by an unsustainable unemployment rate, but by families and individuals sacrificing time and money for a decreasing reward.

The U.S. economy looks a lot better, starting in 2014

As we head into 2018, I still occasionally see arguments about the U.S. economy based on data that do not cover the years from 2014 to present. There is presumably nothing wrong with pre-2014 data, but it is worth keeping in mind that the economy looks a lot better since 2014. Perhaps most notably, wages and employment prospects improve for large portions of the population, an encouraging change in direction after many consecutive years of loss or stagnation. Recent data show a stronger economy coexisting comfortably with the environmental and labor policies of the past four years.

Specifically, three major pieces of the U.S. economic growth puzzle fall into place around 2014 and boost growth each year since: lower oil and energy prices, tighter labor markets, and available low-cost credit. These are discussed below and captured by the period averages in table 1.

Capture

Savings on energy boosted household finances

First, the price of oil fell sharply in 2014 and has remained low due to slowed demand from China, a large increase in domestic oil and gas production, increased green energy production, and more efficient energy use (for example more hybrid and electric cars). Several years with an average discount of 35 percent on oil, and similar discounts on other sources of energy, has improved household finances and boosted the economy. It translates to higher real income for Americans facing lower prices less at the pump, at the store, and on utilities.

Household spending on gasoline and similar fuels accounted for 2.5 percent of U.S. GDP from 2010 to 2014, but has fallen to an average of less than 1.8 percent of GDP from 2014 onward and is currently down to around 1.5 percent of GDP. Rather than save the money left over from less expensive commutes, which could have the effect of slowing the economy, Americans have turned it into new spending. Meanwhile, U.S. energy production has been expanding despite lower prices for gas and oil, thanks in part to increased export capabilities. Reasonably good fortune for both consumers and producers in the industry does not scream overly burdensome regulation.

Yellen pushed unemployment lower than expected

Next, the unemployment rate is currently much lower than many economists had believed it could go, and the Fed should get some credit for this. In January 2014, the unemployment rate was 6.6 percent and 76.4 percent of Americans age 25-54 had jobs. Instead of responding to missing-yet-theoretically-predicted inflation as the unemployment rate neared 5.5 percent, Chair Yellen’s Fed allowed the rate to fall to the current 4.1 percent. As of November 2017, inflation is low and stable and 79 percent of Americans age 25-54 have jobs, an increase over the January 2014 level equivalent to 3.3 million jobs. The rebound in employment for ‘working-age’ Americans should bring pause to claims that the economy has permanently written off portions of its previous workforce. The economy is not only still adding jobs at a rapid pace, but it is still pulling workers off of the sidelines and into the labor market. These trend are even more pronounced for black and Hispanic Americans.

Critically, when we near full employment, as we are doing now, there are no longer huge numbers of available unemployed workers to compete for new jobs or to replace workers who quit. As unemployed workers become scarce, businesses must increasingly compete with each other as employers. They do so by raising wages and making jobs better, and by better training and equipping existing workers. Benefits of full employment therefore include both increased wages and increased productivity. As pointed out by Dean Baker and others, recent data suggest that we are seeing this benefit. 

Borrowing is back despite [efforts at] regulation

In addition to benefits from more employment, Fed policy has kept mortgage rates and other borrowing costs low, further boosting the overall economy. At the end of 2017, the average interest rate for a 30-year fixed rate mortgage is slightly less than four percent. For context, this rate, which partially determines how much people pay for a home loan, had not fallen below five percent prior to the great recession. Importantly, unlike in the years immediately following the crisis, consumer access to credit has been growing in recent years, making loans both relatively cheap and more readily available. The often discussed credit freeze seems to thaw from 2014 to 2017.

Mortgages for younger Americans also become more common in recent years. First-time buyers made up only 29 percent of home sales in 2013, according to a June 2017 report from Genworth Mortgage Insurance, well below the 2009 or 2010 level. However, by 2016 new buyers make up 37 percent of sales, the highest share since 1999. I suspect that the average age of new home buyers is older now than it was in 1999, but average years of schooling have also increased over this period, so some delay to home ownership is expected. Commonly-accepted notions of young Americans shunning home ownership by choice (think avocado toast memes) are starting to look as suspicious as the financial sector claims about regulation preventing lending.

Gains from inherited policy can be saved or squandered

Many stories that make sense when describing the lack of economic recovery from 2008 to 2013 fail to fit the recent data. Peak oil, job-stealing robots, aging populations, crippling regulation, and people playing video games, we have been told, consign us to a different growth path. But, as I have discussed, contrary to the prediction of these stories, data covering the years since 2014 show rebounding employment rates, reasonable energy prices and interest rates, and the potential for strong wage and productivity growth. Beyond serving as face value good news, the sustained economic turnaround started in 2014, suggesting that the worker and environmental protection policies in place over the last four years are not too burdensome. As long as people like them, these policies should be maintained; the economy is growing just fine with them in place. If people particularly like some, like health care exchanges, there’s no reason not to improve or expand them.

Despite rapid improvement since 2014, the economy can still add millions of jobs before reaching full employment. This offers room for the current administration to score more policy gains than just the ones it has inherited (little additional growth is expected from its single policy accomplishment, the TCJA). Low-hanging fruit stems from the current interest rate of 2.75 percent on 30-year treasury bonds. Given the low cost of borrowing, the U.S. has little excuse for carrying a long list of unstarted infrastructure projects into 2018. The current administration, with a Republican majority in all three branches of government, is unhindered in its ability to shift more growth levers in the ‘four percent’ direction and push the country closer to full employment, and can do so with federal investments in infrastructure. Doing the opposite, such as by making unpopular cuts to social insurance, would have the effect of slowing the economy. New data and economic theory agree with voters on this one: keep the policies that work and focus on infrastructure.

 

Using APIs to get economic data with Python — list of Jupyter notebooks

I’ve cleaned up jupyter notebook examples of using data providers’ APIs to request data using python. Each notebook works through at least one example and has links to documentation.

Here are links to the individual notebooks:

Using the IMF Data API: Retrieving an IFS series with Python

April 5, 2017 Update: I’ve put together a newer version of the code/guide located here.

The following tutorial shows how to use python’s requests package to get data directly from the International Monetary Fund (IMF). The IMF’s application programming interface (API) provides access to economic and financial data of more than 180 countries over more than 60 years.

More information on the IMF Data API

The IMF offers guidance on using their data services. The API is not stable, so check the IMF data services news if you receive error messages.

More information on Python

This example works with the Anaconda distribution of Python 2.7.

A short example: Loading IMF data into pandas

Below is a short working example of loading the Australian export price index time-series from International Financial Statistics (IFS) into a pandas dataframe directly from the IMF API.

In [1]:

# Example: loading IMF data into pandas

# Import libraries
import requests
import pandas as pd

# URL for the IMF JSON Restful Web Service, 
# IFS database, and Australian export prices series
url = 'http://dataservices.imf.org/REST/SDMX_JSON.svc/CompactData/IFS/Q.AU.PXP_IX.?startPeriod=1957&endPeriod=2016'

# Get data from the above URL using the requests package
data = requests.get(url).json()

# Load data into a pandas dataframe
auxp = pd.DataFrame(data['CompactData']['DataSet']['Series']['Obs'])

# Show the last five observiations
auxp.tail()

Out [1]:

@BASE_YEAR @OBS_VALUE @TIME_PERIOD
232 2010 94.6044171093095 2015-Q1
233 2010 90.4668716801789 2015-Q2
234 2010 90.4668716801789 2015-Q3
235 2010 85.5465473860777 2015-Q4
236 2010 81.5208275090858 2016-Q1

Breaking down the URL

The ‘key’ in our request is the URL, which contains instructions about which data we want.

The URL has three parts,

  1. http://dataservices.imf.org/REST/SDMX_JSON.svc/CompactData/ The base for requests using the JSON restful data service;
  2. IFS/ the database ID, IFS, for International Financial Statistics;
  3. Q.AU.PXP_IX.?startPeriod=1957&endPeriod=2016 the data dimension and time period information.

The third part, data dimension and time period information, is broken down on the IMF Web Service knowledge base as:

{item1 from dimension1}+{item2 from dimension1}{item N from dimension1}.{item1 from dimension2} +{item2 from dimension2}+{item M from dimension2}? startPeriod={start date}&endPeriod={end date}

For guidance on finding dimension information and building your request, see my previous example of using the IMF API to retrieve Direction of Trade Statistics (DOTS) data. 

Cleaning up the dataframe

Let’s add more meaningful headers and set the date as our index

In [2]:

# Rename columns
auxp.columns = ['baseyear','auxp','date']

# Set the price index series as a float (rather than string)
auxp.auxp = auxp.auxp.astype(float)

# Read the dates as quarters and set as the dataframe index
rng = pd.date_range(pd.to_datetime(auxp.date[0]), periods=len(auxp.index), freq='QS')
auxp = auxp.set_index(pd.DatetimeIndex(rng))
del auxp['date']

# Show last five rows
auxp.tail()

Out [2]:

baseyear auxp
2015-01-01 2010 94.604417
2015-04-01 2010 90.466872
2015-07-01 2010 90.466872
2015-10-01 2010 85.546547
2016-01-01 2010 81.520828

Plot the data

Now we can use matplotlib to create a line plot showing the history of the Australian export price index.

In [3]:

# import matplotlib and pyplot
import matplotlib as mpl
import matplotlib.pyplot as plt
%matplotlib inline

# Create line plot and add labels and title
auxp.auxp.plot(grid=True, figsize=(9, 5), color="orange", linewidth=2,)
plt.ylabel('Index')
plt.xlabel('Year')
plt.title('Australia: Export Price Index (baseyear=' + str(auxp.baseyear[0]) + ')');

Out [3]:

Output from script

Save as a csv file

To save the data for use in another program, you can easily create a csv file.

In [4]:

auxp.to_csv('auxp.csv')

Brexit and the feedback of uncertainty through asset prices

The medium- and long-run consequences of Brexit are unknown, but increased uncertainty in response to the event will likely have real global effects of its own. Individuals are now less confident in their guesses about the future, and this uncertainty changes behavior, increasing risk aversion and decreasing business investment. Global market reactions to Brexit reflect not only the new average belief about expected future earnings, but also the indirect feedback effects of the higher level of uncertainty.

Medium- and long-term effects unclear

The U.K.’s June 23 vote to leave the European Union, termed ‘Brexit’, surprised many analysts. As the referendum vote neared, markets (representing the average belief of individuals) increasingly expected the ‘remain’ party to win. Some well-educated analysts explained how Brexit would never happen (possibly what Nassim Nicholas Taleb would consider proof of a black swan). The ‘leave’ campaign won with 51.9 percent of the vote, prompting David Cameron’s announced resignation.

The medium-term consequences for the U.K. and E.U. (or anywhere else) are entirely unclear. For example, should unemployment increase and production falter in the U.K. during the next few months, the Bank of England may need to ease monetary policy further and adopt very unconventional tools. However, if the market reaction to Brexit can be fully absorbed by exchange rate adjustment, the Bank of England may need to raise interest rates in response.

On June 24, the pound sterling fell more than eight percent against the dollar (figure 1). In response to the currency depreciation, many here in the U.S. joked about buying goods online from the U.K. or planning vacations. Many others expressed uncertainty about the future of the U.K. and the E.U., and the economic and political implications for the U.S.

gbp_jun262016
Figure 1. The pound sterling fell against most currencies following the Brexit vote, including by more than eight percent in one day against the U.S. dollar.

Markets reminded of their anxiety issues

Within hours of the ‘leave’ campaign victory, stock prices of some financial services companies, such as Lloyds, fell by 30 percent. Although real changes from the vote may take two years to implement, the average belief about the appropriate cost of equity for these firms changed dramatically overnight. The fundamentals for people and businesses in the U.K. do not immediately change, but their behavior and asset prices do, and this has an indirect feedback effect on the fundamentals

Likewise, the U.S. is clearly not part of the E.U. or U.K., however, it is intimately linked to both markets through trade and investment. The Dow Jones Industrial Average fell 610 points (3.39%) on Friday in response to the news. The most widely-used measure of market volatility jumped nearly 50 percent (figure 2). The sales revenue of the largest U.S. companies is practically unchanged, however, individuals’ collective behavior has changed in response to uncertainty.

Individuals are repricing the assets that they previously overvalued. However, if the Bank of England is forced to wait to decide which direction its key policy will turn, how well can individuals immediately reprice future earnings? The immediate reaction to Brexit reflects not a new certainty about future earnings but increased uncertainty. Individuals never know what the future will hold, but these large unexpected events make them less confident in all of their guesses, and it changes their behavior.

vix_jun252016
Figure 2. VIX, the CBOE measure of expected near-term equity market volatility, saw an almost 50 percent daily increase following the Brexit result

Searching for safety and the uncertainty feedback loop

When people are less certain about the future, they take fewer risks. For example, business investment falls when equity market volatility is high. Individual and institutional investors also show preference for safe assets during times of increased uncertainty. In an immediate reaction to Brexit, U.S. treasuries, gold, and Japanese Yen saw inflows, while equity markets saw net outflows. Neither business fundamentals nor any rules had changed; people were showing risk aversion.

Investors’ increased aversion to equity investment raises firms’ equity cost of capital. This cost of capital is a major factor in firms’ investment decisions, especially for smaller and less-cash-flush firms. Therefore, the uncertainty induced asset price shock has a feedback mechanism through which it affects the future value of companies. This dangerous feedback mechanism can be procyclical.

What it means for the U.S.

The medium- and long-term direct effects of Brexit on the U.S. are very unclear. Brexit-induced equity market volatility, higher levels of uncertainty, and the negative effects these entail can however be analyzed with attention to the current U.S. macroeconomic environment. Three potential short-run consequences emerge:

1) Continued volatility in equity markets and strong demand for treasuries;

2) More downward pressure on business investment, which was already negative in the first quarter of 2016; and

3) Delay of the next Federal Funds rate hike in response to the above.

While the downward pressure on already weak business investment is worrisome, none of the above are enough to induce major concern. Households, many of which have recently started seeing long-awaited wage and income increases, will play a key role in determining whether recent asset price volatility will spill into consumer sentiment and awake a much worse set of feedback mechanisms.

Check out the dashboard and please leave feedback

Nearly 100 freshly updated charts:

U.S. Macro and Markets Dashboard (Updated June 26, 2016)

References and additional reading

Networks, Crowds, and Markets: Reasoning About a Highly Connected World by David Easley and Jon Kleinberg

Nick Bloom comment on Brexit 

BBC: The UK’s EU referendum: All you need to know

 

Global trade network analysis with Python: central players in bluefin tuna and large aircraft

Network analysis provides useful insights into complex bilateral trade data. Two methods are presented for calculating with Python each country’s influence in the global trade network for individual goods. Related concepts in graph and international trade theories are discussed.

Modern goods have complex trade networks

The things we buy increasingly travel long distances and from scattered origins before they reach us. Take one man’s repeatedly failed attempt to build a toaster (from scratch). Over several decades companies have changed their production techniques, relying on global value chains, for example, to keep costs low. These changes have gradually contributed to long-term growth in global trade.

The more deeply a country becomes involved in global trade, and in global supply chains in particular, the more subjected its economy becomes to changes abroad. This can be good; historically many powerful cities began as ports. However, the potential for higher returns from servicing foreign demand carries with it increased risk of economic contagion.

It is not hard to imagine how global supply chain connections transmit effects from other countries. If a strike in France delays the delivery of a crucial intermediate good, it may cause an assembly line stoppage in Taiwan. On an aggregate level, the results do not necessarily average out, and can result in vast shifts of wealth.

It may therefore prove useful to examine the complex networks of global trade using the tools provided largely by graph theory. As an example, let’s start with a graph of the global trade of tires in 2012.

401110
Trade networks are complex and large. Data source: UN Comtrade

Each country that exported or imported automobile tires in 2012 is represented above by one node labeled with its three letter country code (for example Germany is DEU). The precise location of a node on the graph is not critical (it is often arbitrary), but generally countries more central to the trade of tires are closer to the center of the network. Likewise, countries are generally graphed near their largest trading partners.

Each trading relationship is shown on the graph as an edge (a line connecting two nodes). If France exports tires to Aruba, the graph will include an edge connecting the two nodes labeled FRA and ABW. Trade network edges are considered directed, as the flow of goods has a direction (either imports or exports).

Rat’s nest or Rorschach?

You may look at the above ‘visualization’ and simply see a rat’s nest. This is a correct interpretation. The graph shows overall complexity in the trade network, not individual bilateral relationships (there are more than 4400 edges in this network). Indeed the automobile tire trade network is particularly large and dense. Many countries currently produce internationally competitive tires and all countries use them and import at least some. In fact, the average country imports tires from many other countries. A graph of the resultant trade network is reminiscent of a gray blob and practically as useful.

More useful, however, are individual metrics of network structure. For example, which countries tend to trade only with a select subgroup of other countries? Which goods are traded in networks where one country dominates trade? These questions relate theoretically to the respective graph theory concepts of clustering and centrality.

Let’s take a look at how the Python programming language can be used to measure centrality in trade networks, and discuss two specific measures of centrality.

Python for trade network analysis

What follows is a more technical segment with sample code for trade network analysis of using Python 2.7.

Let’s start by importing the packages

In [1]:

import networkx as nx
import csv
import numpy as np
import matplotlib
import matplotlib.pyplot as plt
from matplotlib import cm
from mpl_toolkits.axes_grid1 import make_axes_locatable
%matplotlib inline

We will rely heavily on NetworkX and give it the short name nx. Numpy is used to do certain calculations, and matplotlib helps with the visualizations.

Load the data and build the network

The example uses cleaned bilateral UN Comtrade trade data for scrap aluminum exports in 2012. The data follow the HS2002 classification system at the six-digit level of aggregation, and are sourced from WITS (subscription required for bulk download). Data are read from a csv file with the equivalent of three ‘columns’: the exporting country code, the importing country code, and the inflation-adjusted US Dollar value of exports in the one year period.

Data from the csv file are read line by line to build the network quickly. NetworkX is used to build the network, which is called G according to convention,  as a series of edges.

Read the data and build a network called G

In [2]:

G = nx.DiGraph() # create a directed graph called G
 
# Loop reads a csv file with scrap aluminum bilateral trade data
with open('760200_2012.csv', 'r') as csvfile:

     csv_f = csv.reader(csvfile)
     csv_f.next()

# Now we build the network by adding each row of data 
# as an edge between two nodes (columns 1 and 2).
     for row in csv_f:
          G.add_edge(row[0],row[1],weight=row[2])

Let’s look at a specific bilateral trade relationship to verify that the new network, G, is correct. Exports of scrap aluminum from the U.S. to China should be quite large in 2012.

Check individual trade flow (edge)

In [3]:

usachnexp = G.edge['USA']['CHN']['weight']
print 'USA 2012 scrap aluminum exports to China, in USD: ' + str(usachnexp)
USA 2012 scrap aluminum exports to China, in USD: 1199682944
Central players can affect the market

Now that the network has been built, we can use indicators from graph theory to identify potential weaknesses and risks in the network’s structure. In this example, we will look for the presence of dominant countries in the trade network. Dominant importers or exporters have the ability to influence supply and demand and therefore price. These dominant countries are highly influential players in the trade network, a characteristic measured in graph theory as centrality.

There are several measures of centrality and two are discussed briefly in this post. The first is eigenvector centrality, which iteratively computes the weighted and directed centrality of a node based on the centrality scores of its connections. The example below scores each importer country as a function of the import-value-weighted scores of its trading partners. That is, an importer is considered influential to a trade network (receives a high eigenvector centrality score) if it imports a lot from countries that are also influential. Mathematically, eigenvector centrality computes the left or right (left is import centrality, right is export centrality) principle eigenvector for the network matrix.

See the NetworkX documentation for questions on the code or this for more details on the the math.

Calculate eigenvector centrality of imports

In [4]:

# Calculate eigenvector centrality of matrix G 
# with the exports value as weights
ec = nx.eigenvector_centrality_numpy(G, weight='weight')

# Set this as a node attribute for each node
nx.set_node_attributes(G, 'cent', ec)

# Use this measure to determine the node color in viz
node_color = [float(G.node[v]['cent']) for v in G]
Calculate total exports

Next we calculate each country’s total exports of scrap aluminum in 2012 as the sum total of its individual exports (edges) to other nodes. In the script, total export data is assigned as a node attribute and set aside to be used as the node size in the visualization.

Calculate each country’s total exports

In [5]:

# Blank dictionary to store total exports
totexp = {}

# Calculate total exports of each country in the network
for exp in G.nodes(): 
     tx=sum([float(g) for exp,f,g in G.out_edges_iter(exp, 'weight')])
     totexp[exp] = tx
     avgexp = np.mean(tx)
nx.set_node_attributes(G, 'totexp', totexp)

# Use the results later for the node's size in the graph
node_size = [float(G.node[v]['totexp']) / avgexp for v in G]
Visualization of the scrap aluminum network

NetworkX works well with matplotlib to produce the spring layout visualization. It is another rat’s nest, but you may notice a different color on one of the medium-sized nodes.

Create graph using NetworkX and matplotlib

In [6]:

# Visualization
# Calculate position of each node in G using networkx spring layout
pos = nx.spring_layout(G,k=30,iterations=8) 

# Draw nodes
nodes = nx.draw_networkx_nodes(G,pos, node_size=node_size, \
                               node_color=node_color, alpha=0.5) 
# Draw edges
edges = nx.draw_networkx_edges(G, pos, edge_color='lightgray', \
                               arrows=False, width=0.05,)

# Add labels
nx.draw_networkx_labels(G,pos,font_size=5)
nodes.set_edgecolor('gray')

# Add labels and title
plt.text(0,-0.1, \
         'Node color is eigenvector centrality; \
         Node size is value of global exports', \
         fontsize=7)
plt.title('Scrap Aluminum trade network, 2012', fontsize=12)

# Bar with color scale for eigenvalues
cbar = plt.colorbar(mappable=nodes, cax=None, ax=None, fraction=0.015, pad=0.04)
cbar.set_clim(0, 1)

# Plot options
plt.margins(0,0)
plt.axis('off')

# Save as high quality png
plt.savefig('760200.png', dpi=1000)
760200
China (CHN) is influential to scrap aluminum trade in 2012. Data source: UN Comtrade
Central players on the demand side: scrap aluminum and bluefin tuna

The graph above shows plenty of large exporters (the large nodes) of scrap aluminum in 2012, including the US, Hong Kong (HKG), and Germany (DEU). The demand of one country in the network, however, actually dominates the market. In 2012, the booming Chinese economy was purchasing large quantities of industrial metals, including scrap metals. The surge in demand from China was enough to cause global price increases and lead to increased levels of recycling. Since 2012, however, Chinese imports of scrap aluminum have nearly halved, as has the market price of aluminum. The recent boom-and-bust cycle in scrap aluminum prices has a single country of origin but global ripples; the downturn generates domestic consequences for the large exporters and reduces the financial incentives for recycling.

The central influence of China in the 2012 scrap aluminum trade network is captured by its high eigenvector centrality score (node color in the graph above). We can also easily infer the out sized influence of China from a more simple measure–the high value of its imports relative to other countries. Centrality metrics, of which there are many, often prove useful in nuanced cases.

760200_cent
Chinese demand for scrap aluminum dominates the market in 2012. Data source: UN Comtrade

Another example of a central influence on a trade network can be found in Japanese demand for bluefin tuna. As shown below, Japan has very high eigenvector centrality for imports of this key ingredient in many sushi dishes.

030346

Australia (AUS) dominates bluefin tuna exports, but by eigenvector import centrality Japan (JPN) is the influential player in the market. The first Tokyo tuna auction of 2013 saw one fish fetch a record 1.76 million USD. Indeed in 2012 Japan imported more than 100 times as much bluefin tuna as the second largest importer, Korea.

Like scrap aluminum, the story here follows the familiar boom-and-bust cycle; prices for bluefin tuna have returned to lower levels since 2012. The structure of the trade network, with one central player, introduces a higher level of price volatility. During a downturn in prices, this transmits financial consequences to fishermen throughout the world.

Supply-side influential players: large aircraft production

Trade network analysis can also help to identify influential exporters of goods. Cases that come to mind are rare earth minerals found only in certain countries, or large and complex transportation equipment. Commercial aircraft manufacturers, for example, are limited (unfortunately this may have more to do with subsidies than limited supply of technological prowess). Very large aircraft production is dominated by two firms: Airbus, with production sites primarily in France and Germany, and U.S. competitor, Boeing.

Instead of using eigenvector centrality to measure the influence of each exporting country in the large aircraft global trade network, let’s use a more simple method called outdegree centrality. We compute outdegree centrality for each country, i , as its number of outgoing (exporting) connections, k^{out}_i , divided by the total number of possible importers, (v - 1) :

c^{out}_D (i) = k^{out}_i / (v - 1) .

You can think of this measure as the share of importers that are serviced by each exporter. Nodes with a high outdegree centrality are considered influential exporters in the network.

Calculate outdegree centrality

In [7]:

oc = nx.out_degree_centrality(G) # replaces ec in the above
880240
France, Germany, and the US dominate exports of large aircraft. Data source: UN Comtrade

As expected, France, Germany, and the U.S. receive high outdegree centrality scores. There simply aren’t many alternative countries from which to buy your large aircraft. Beyond lack of choice for buyers, central exporters in a trade network may introduce (or represent) vulnerability and barriers to competition.

880240_cent
Large aircraft suppliers are limited. Data for 2012, source: UN Comtrade
Network structure and (preventing) domestic consequences

Global trade is increasingly complex. Open economies are vulnerable to supply and demand shocks from the other countries in their trade network. The structure of the trade network itself determines in part the level of vulnerability and how and where supply and demand shocks may be transmitted. Certain trade networks, such as those for scrap aluminum or bluefin tuna, face dominant consumers and additional price volatility. Networks can also be subject to supply-side market structure issues, such as the virtual duopoly with very large aircraft.

Hindsight makes bubbles more visible; we easily find the previously missed warning signs once we know where to look. Decision makers aim for early detection of vulnerabilities, but face a geographically growing set of possible sources. Network analysis tools, such as centrality, can be applied to existing sets of complex bilateral trade data to provide new insight in the search for today’s warning signs. Such nontraditional tools may prove increasingly useful in a world where an individual is not capable of building a toaster from scratch, yet they sell down the street for $11.99.

Additional resources and reading
For fun:

The Toaster Project

Some references and further reading on networks and graph theory:

Easley and Kleinberg (2010) Networks, Crowds, and Markets: Reasoning about a Highly Connected World

Jackson (2010) Social and Economic Networks

Reading on trade and trade networks:

De Benedictis (2013) Network Analysis of World Trade using the BACI-CEPII dataset

Feenstra (2005) World Trade Flows

Nemeth (1985) International Trade and World-System Structure: A Multiple Network Analysis

Trade data source:

World Integrated Trade Solution (WITS)

Some python related resources:

Anaconda distribution for python

NetworkX

ComputSimu: Networks 1: Scraping + Data visualization + Graph stats (more useful code here)

Dashboard update: bond yields fall with renewed demand for safe assets and lower interest rate expectations

Monitor more than 80 economic indicators with the macro and markets dashboard:

dash_open
United States Macroeconomic and Markets Dashboard: Updated June 11, 2016

Dashboard weekly update summary:

The latest labor market data show continued overall improvement in wages and low levels of new jobless claims, offering some consolation after the surprisingly weak May jobs report (see last week’s update). Equity market volatility increased, however, as already lackluster global growth forecasts were revised down by the World Bank. Domestic and foreign investors are shifting their portfolio of assets to fixed streams of income. Global bond yields, including on U.S. government and corporate debt, fell considerably during the past week’s rally. Investors are searching for higher returns on safe assets and responding to lower interest rate expectations.

Wages grow faster than productivity

Narrowing a long and persistent gap between productivity and wage growth, recent data suggest wages have been increasing in many industries. For most of the past decade, worker’s productivity (the output for each hour of work) grew more rapidly than their wages. The gap was in part from technology making work more efficient, but it also came from a weak labor market. An economy in which there are many qualified workers for each opening makes workers less likely to quit and more likely to accept no or small increases in wages. Companies simply do not need to rely on pay increases as a motivation when fear of unemployment is very strong.

While wage growth crawled along for a decade, productivity growth remained strong. Recent data suggest, however, that the long upward trend in productivity may be facing at least a hiccup. Meanwhile, overall measures show wages have been growing at a reasonable pace since mid-2014. Revised first quarter 2016 index data on wages and productivity shows the former nearly catching the latter.

wageout_jun112016

When oil prices fall, for example, there is a transfer of wealth from the stakeholders of oil producers (who face a fall in revenue) to households (who spend less on fuel and energy). Likewise, a fall or stagnation in corporate profits can result in an increase in worker’s relative share of output. Wages, unlike commodity or stock prices, tend not to be cut. Where the past year has seen labor markets and workers’ bargaining power improve, it has seen productivity and corporate profits stagnate.

lshare_jun112016

Additionally, data for the week of June 4 on new claims of joblessness was strong, with only 264,000 such claims. Overall, reasonable wage increases, an increasing labor share of output, and low headline unemployment paint a better picture for households and aggregate demand than the last jobs report’s payroll growth and participation rate data suggest. That said, labor market improvements are traditionally slow and gradual, while deterioration is rapid and steep. The June jobs report should therefore have major implications for the Fed’s rate hike decision.

Equity market volatility climbs

U.S. equity markets gains over the first four trading days of the week were erased on Friday by a large sell off. The CBOE volatility index, VIX, increased to 17 from 13.5 a week earlier. The bond rally described below suggests that there has been a flight to safety. Investors have been adjusting in part to new forecasts of generally lower global growth. Likewise, there are several large events on the horizon (brexit, elections, central bank policy divergence, etc.) suggesting fluctuations and jumps in equity markets (as well as debt and forex markets).

vix_jun112016

Bond markets rally as investors seek safe assets and interest rate expectations fall

Over the past week, U.S. treasury bonds, t-bills, and corporate bond yields fell. Elsewhere, ten-year German bund yields are nearly negative and Japanese ten-year government bond yields have fallen to -0.13%. People’s tolerance for extremely low returns is limited, and U.S. government debt offers a relatively higher return. During the past week, ten-year U.S. treasury bonds reached a four year low, partially as a result of strong foreign demand.

yield_jun112016

The spread between 10-year and 2-year treasuries sits currently at a nine-year low. Potential causes for the flat yield curve include the following: 1) investor search for return driving driving down long- and medium-duration bond yields, 2) investor fear of a business cycle downturn and a near future need for monetary easing, and 3) lower interest rate expectations as a result of recent data taking June (and potentially July) rate hikes off the table.

spread_jun112016

Check out the full dashboard for more than 80 indicators of U.S. economic activity:

U.S. Macroeconomic and Markets Dashboard, June 11, 2016

I also updated the dashboard for Japan:

Japan Macroeconomic and Markets Dashboard, June 11, 2016