Power Laws: Raise Those Eyebrows

My long-ago memory is that basic classroom experiments in physics and chemistry would produce a graph of data that was pretty close to a straight line, or a smooth curve. In social science, it's more common to see a rougher pattern, with the points scattered around. As a result, the social science research on "power laws" is full of graphs that raise my eyebrows, because there are lots of graphs where the points fall very close to a straight line. The fit looks too good! Xavier Gabaix provides a readable overview in "Power Laws in Economics: An Introduction," appearing in the Winter 2016 issue of the Journal of Economic Perspectives, which pushed my  (Full disclosure: I've worked as Managing Editor of JEP for 30 years now. All JEP article back to the first issue are freely available on-line compliments of the publisher, the American Economic Association.)

Here are some of Gabaix's examples from the paper, but he also refers to power law results in a wide array of other papers. (For some readers, it may be useful to add a few words on what a "power law" is. On a typical linear graph, each equal distance on the graph represents a change of the same absolute amount--say, 1, 2, 3, 4, ...--although the units may be expressed in millions or years or percentage points or dollars or whatever is useful. In a "power law," each equal distance on the graph represents a rise in an exponential power--say, 10¹ , 10², 10³, 10⁴ ... As a result, what appears to be an equal visual distance on the graph now represents not an absolute change, but a proportional change: for example, each equal visual distance in the powers-of-10 example represents a 10-fold increase.)

 Consider a graph based on the population of US cities, which in this data is all cities with population above more 250,000. The horizontal axis is population, expressed as powers-of-10. The vertical axis is the rank of the city size--that is cities are ranked by population from #1 New York to #2 Los Angeles and so on. Again, the vertical axis is expressed as powers-of-10. The result is very close to a straight line with a slope of -1.

As Gabaix writes: "A slope of approximately 1 has been found repeatedly using data spanning many cities and countries (at least after the Middle Ages, when progress in agriculture and transport could make large densities viable, see Dittmar 2011). There is no obvious reason to expect a power law relationship here, and even less for the slope to be 1."

Now here's an example looking at the distribution of the size of US firms, measured by the number of employees on the horizontal axis, and the number of firms of this size, measured on the vertical axis. Again, both axes are measured in powers-of-10. Again, the slope is very close to -1. But why should the ranks of cities as measured by population look similar to the frequency of firms as measured by number of employees? (As I said, these are the sorts of graphs that make your eyebrows go up,)

Or here's an example about the distribution of daily stock market market returns. You can read the details of the calculations in the Gabaix article, but again, the axes are expressed as powers-of-10, and a linear relationship seems to emerge.

Or here's an example of a power law in the relationship between the pay of CEOs and the size of firms. Here, the labels on the graph are expressed as logarithms. When the size of the firm rises, the so does CEO pay--but in an exponential power-law kind of way. Gabaix writes: "In a given year, the compensation of a CEO is proportional to the size of the firm to the power of 1/3, S(n)^1/3, an empirical relationship sometimes called Roberts’ (1956) law." He argues that this pattern of larger firms paying more to their CEOs can explain much the rise in CEO pay over time, as well as cross-differences in what CEOs are paid.

Why do these kinds of power law relationships show up so often? Say you start of with a random distribution of something, and the different points in your data all tend to experience proportional growth (positive or negative). However, this particular data (like city size) can't turn negative. In addition, the total size of the system can't grow in an unbounded way. Gabaix explains how with a few additional assumptions this process will result in the kind of straight-line power laws shown here. Of course, this kind of explanation then needs to be adapted and applied to each specific situation.

Want a power law outside of economics? Here's a graph where the typical mass of an animal is shown on the horizontal axis with a power-of-10 scale, and the metabolic rate, or energy requirement of that animal each day, is shown on the vertical axis. I'm sure that clever biologists can give reasons for why this should be so. But given what certainly seem to be substantial differences in animal behavior, it's not obvious to me that, before the data was available, they would have expected straight-line power-law to emerge here, either.

The Fed Semiannual Update: An Interest Rate Gap and the Balance Sheet

Twice each year, the Federal Reserve is required to submit to Congress a report about "the conduct of monetary policy and economic developments and prospects for the future." The most recent Monetary Policy Report went to Congress on February 10, 2016. A number of the themes will be familiar to regular readers of this blog. For example, there's a discussion of how US unemployment rates have more-or-less bottomed out, and whether there are some signs of a pickup in wage growth (see my post on "Unemployment is Bottoming Out, So What's Next?" January 25, 2016). There's a discussion of how the inflation rate is affected by fluctuations in energy prices, food prices (see "Breaking Down US Inflation Rates by Category," February 9, 2016). There's a discussion of how Federal Reserve policy of raising interest rates may not have a huge effect on emerging markets, because any negative effects they experience from capital outflows will be largely offset by how a lower exchange rate spurs their exports ("Bernanke on the Fed, the US Dollar, and the Global Economy," January 8, 2016).

Here, I'll just note a couple of other figures that caught my eye.

A divergence has emerged in the interest rates of advanced market economies, between the US and UK on one hand and the euro-zone and Japan on the other. Here are a couple of examples.
The first shows nominal yields on 10-year government debt.

The other is what's called the "overnight index swap rate." In general, an interest rate "swaps" contract occurs when one party that's getting a variable interest rates over time swaps with another party that is getting a fixed interest rate over time. Obviously, the fixed interest rate in the swap will reveal what the average of the variable rate is expected to be. The figure shows the two-year overnight index swap rate for several economies. For the US, this is the fixed interest rate which shows what the average value of the US federal fund interest rate (the interest rate targeted by the Federal Reserve) is expected to be in the next two years. The other lines show what average interest rate is expected for the target interest rate of other central banks--and notice that it is turning negative for the Bank of Japan and the European Central Bank.

Both sets of interest rates show that the rates for the US and the UK are substantially above those for Japan and the euro-zone. The reasons for such a divergence are a jumble of expectations about growth rates, inflation rates, financial stability, and central bank policies. But whatever the reasons, this difference helps to explain why demand for US dollar assets is up and the US dollar exchange rate has been rising.

The big shift in the Federal Reserve balance sheet has leveled off. Just about anyone who is teaching a serious course about Federal Reserve Policy in recent years uses some version of this figure. It shows the assets and liabilities of the Federal Reserve system. For example, two of the main assets of the Fed are the US Treasury securities that is owns, along with the mortgage-backed securities and housing-related debt that it owns. The "other assets" are mainly certain premiums or discounts on these two other categories that for one accounting reason or another aren't built into the usual price of the asset. The two main liabilities of the Fed are currency--that is, Federal Reserve notes in circulation--along with the deposits it holds from member banks. The remaining "Capital and other liabilities" includes, for example, the US Treasury General Account, which is more-or-less the checking account through which the federal government pays its bills, and also includes some financial deals like reverse repurchase agreements.

The figure shows how the Fed balance sheet has evolves over time. Back in mid-2007, Fed assets were basically all Treasury securities, and its liabilities were basically all currency. But when the financial crisis hit in late 2007, several shifts happened.

The Fed set up an alphabet soup of agencies to make temporary loans to various financial institutions during the recession: for example, the Primary Dealer Credit Facility, the Asset-Backed Commercial Paper Money Market Mutual Fund Liquidity Facility, the Commercial Paper Funding Facility, and the Term Asset-Backed Securities Loan Facility. On the asset side, you can see how these loans blossomed for a time, then were repaid after the crisis, and these temporary agencies have now been shut down. Almost no one talks about these temporary agencies any more, but they were a rapid and innovative success in helping to prevent financial contagion from spreading.

As the temporary lending facilities phased out, you can see the quantitative easing policies come into play. The Fed started investing directly in mortgage-backed securities. It greatly increased its investments in US Treasury debt. On the liability side, banks began to hold much larger reserves at the Fed than the bare legal minimum they used to hold. The banks now receive interest on these reserve, too. These changes leveled out in the later part of 2014. But clearly, the Fed as a financial institution has been fundamentally changed from those not-so-long-ago days when it was pretty much all about US Treasury securities and currency.

How Many in the Gig Economy?

The "gig economy" essentially refers to workers who are available when someone wants to hire them, but who don't have any long-term guarantee of how many hours they will work or in some cases even how much they will earn. Those who work for temp agencies are part of the gig economy, and so are those who drive for a company like Uber. But one problem overhangs all discussions of the gig economy. The discussions often end up being about anecdotal cases, either of those who find the gig economy to be a useful and preferred arrangement, or others who feel pressured into the gig economy because they couldn't find a steadier ongoing job.  There's not broad agreement on how to define the gig economy, and partly as a result, there also isn't good systematic evidence on how many workers are in the gig economy or how those workers perceive their jobs.

The Secretary of the US Department of Labor, Thomas Perez, announced a couple of weeks ago that the DoL would be working with the US Census Bureau to add as set of survey questions about "contingent workers" to the May 2017 Current Population Survey. But while we're waiting for that survey to happen, and the results to be tabulated and released, what do we know now?

The US Government Accountability Office (GAO) published a report in April 2015 called "Contingent Workforce: Size, Characteristics, Earnings, and Benefits."  The GAO writes:

"The size of the contingent workforce can range from less than 5 percent to more than a third of the total employed labor force, depending on the definition of contingent work and the data source.  ... However, no clear consensus exists among labor experts as to whether contingent workers should include independent contractors, self-employed workers, and standard part-time workers, since many of these workers may have long-term employment stability. There is more agreement that workers who lack job security and those with work schedules that are variable, unpredictable, or both—such as agency temps, direct-hire temps, on-call workers, and day laborers—should be included. We refer to this group as the “core contingent” workforce. We estimate that this core contingent workforce comprised about 7.9 percent of employed workers in the 2010 GSS [General Social Survey] and also made up similar proportions of employed respondents in the roughly comparable 2005 CWS [Contingent Worker Survey] and 2006 GSS—5.6 percent and 7.1 percent, respectively."

The data surveyed by the GAO suggests that the contingent workers tend to be younger and less educated, with about a 15% chance of leaving the labor force or being unemployed one month later. Even after adjusting for other observable factors that affect wages (like experience and education), contingent workers earn about 10% less per hour. They are less likely to have benefits, and less likely to be satisfied with their jobs overall.

But notice that the definition of a "contingent worker" in this GAO study is not especially new. Job categories like "agency temps, direct-hire temps, on-call workers, and day laborers" have been around for awhile. The evidence they cite is only updated through 2010. Somehow, these categories and teh timing of the data don't quite seem to cover what it means to work for a company like Uber.

A more recent piece of evidence comes from "Changing Patterns in Informal Work Participation
in the United States 2013–2015," by Anat Bracha, Mary A. Burke, and Arman Khachiyan, which the the Federal Reserve Bank of Boston published as a Current Policy Perspectives in October 2015. These authors designed the Survey of Informal Work Participation, which was then included as part of the Federal Reserve Bank of New York’s Survey of Consumer Expectations in December 2013 and January 2015. They define "informal work" like this:

By informal work we refer to any income-generating activity that does not involve a contract between an employer and an employee (except possibly for contracts involving a single task). This definition includes activities that monetize possessions (such as selling used goods or renting out one’s property) as well as activities that monetize free time and skills (such as babysitting). Typical features of informal work are the following: (1) it involves a greater degree of scheduling freedom than a formal job would, (2) the worker is paid on a per-service or per-good basis, and (3) the work does not provide benefits such as health insurance or pension contributions. ... The number and types of paid informal work opportunities have expanded in recent years, in no small part due to the appearance of new technologies facilitating the so-called peer-to-peer economy. Well-known peer-to-peer businesses include Uber, a taxicab-like business that connects drivers with riders via mobile phones; Airbnb, which enables individuals to rent out their home for brief stays; Amazon Mechanical Turk, which offers the opportunity to do basic computing work from home on a fee-for-service basis; and Taskrabbit, which facilitates spot contracting for personal services."

Bracha, Burke, and Khachiyan are quick to point out that the US economy was in a different place int the two survey dates: for example, the unemployment rate was 6.7% at the time of their Survey 1 in December 2013 but had fallen to 5.7% by the time of their Survey 2 January 2015. Thus, drawing comparisons between the two surveys needs to be done cautiously. Also, although this survey is designed to be nationally representative, it is carried out online and pays the respondents $15, which could introduce some biases in terms of who is likely to answer. That said, here are some findings:

"In Survey 2 [the January 2015 survey], the share of survey-takers who reported participating in informal paid work increased significantly—from 40 percent to 52 percent among men and from 40 percent to 60 percent among women. Among both women and men, participation rates became more equal across education classes in Survey 2. Among women, this equalization reflected in part a large increase in participation among those with high school or less, while among men, the equalization embedded a large increase in participation among those with a graduate degree.
Among men, participation rates became more equal across groups classified based on
employment status. As of Survey 2, men from across the formal income distribution are roughly equally likely to participate in informal work, while among women, the negative association between formal income and informal participation remains in force. ... 

At the same time, however, informal participation increased between the surveys among highly educated and highly paid men, an outcome that likely reflects the fact that recent technological innovations have expanded the set of informal work opportunities and made it easier to engage in such work. Indeed, among both men and women and in both surveys, more than half of those who report engaging in informal work are performing internet-based tasks. In addition, one of the categories with the highest increase in participation between surveys was “online tasks,” which refers to activities such as rating pictures or copy-editing online. Female informal work participants in Survey 2 were more likely than those in Survey 1 to report both that informal earnings were their main source of income and that informal work helped at least somewhat to offset recent negative employment shocks. Taken together, our results suggest that some individuals continue to seek out informal work in order to offset negative economic shocks, while others engage in informal work—despite already being fairly well off—because it offers an easy way to earn extra cash.

They slice up the data in a number of ways, based on whether people have other full-time jobs, other part-time jobs, or no other jobs, as well as by education level. But as an overall summary, it's fair to say that men in the informal economy in this survey were working 10-15 hours per week and earning about $240 per month. Women in the informal economy in this survey were working in the range of 8-18 hours per month at these jobs, and earning in the range of $135-$185 per month. The most common activity for both men and women was "selling online." These results are quite heterogenous: those working in the informal sector include high-educated people who have full-time jobs,  low-educated people without a job, and everyone in between.

There's reason to believe that these kinds of informal jobs are going to increase in number. The Boston Fed researchers also point to some estimates by PricewaterhouseCoopers, who back in 2014 made predictions about sales growth in five "sharing economy" sectors, "peer-to-peer finance, online staffing, peer-to-peer accommodation, car sharing and music and video streaming. PwC estimates that these five sectors had $15 billion in sales in 2013, but are headed for $335 billion in sales by 2025.

One more source that offers some discussion of these issues is Sarah A. Donovan, David H. Bradley, and Jon O. Shimabukuro who wrote a report "What Does the Gig Economy Mean forWorkers?" published by the Congressional Research Service on February 5, 2016. They offer a nice specific definition of the gig economy:

The gig economy is the collection of markets that match providers to consumers on a gig (or job) basis in support of on-demand commerce. In the basic model, gig workers enter into formal agreements with on-demand companies to provide services to the company’s clients. Prospective clients request services through an Internet-based technological platform or smartphone application that allows them to search for providers or to specify jobs. Providers (i.e., gig workers) engaged by the on-demand company provide the requested services and are compensated for the jobs. Business models vary across companies that control tech-platforms and their associated brands. Some companies allow providers to set prices or select the jobs that they take on (or both), whereas others maintain control over price-setting and assignment decisions. Some operate in local markets (e.g., select cities) while others serve a global client base. Although driver services (e.g., Lyft, Uber) and personal and household services (e.g., TaskRabbit, Handy) are perhaps best known, the gig economy operates in many sectors, including business services (e.g., Freelancer, Upwork), delivery services (e.g., Instacart, Postmates), and medical care (e.g., Heal, Pager).

The authors also emphasize various ways that gig workers are different from freelance workers. On-demand firms that contract with gig workers control the brand name, place various requirements on how the job is done, and take a percentage of what is earned. As they write:

However, gig jobs may differ from traditional freelance jobs in a few ways. The established store-front and brand built by the tech-platform company reduces entry costs for providers and may bring in groups of workers with different demographic, skill, and career characteristics. Because gig workers do not need to invest in establishing a company and marketing to a consumer base, operating costs may be lower and allow workers’ participation to be more transitory in the gig market (i.e., they have greater flexibility around the number of hours worked and scheduling).

But while this definition of the gig economy is nice and specific, the report then runs into the problem that there is no systematic survey evidence on the "gig economy" defined in this way. As one example, the number of what the Census Bureau calls "Nonemployer Establishments," which are firms that don't employ anyone but earn at least $1,000 in a year and pay income tax, seems to be on the rise. But it is not at all clear what share of this increase is the "gig economy" as narrowly defined, or those in the "informal economy" or in the category of "contingent workers."

There's considerable discussion over whether the rise of the gig economy, or the informal economy more generally, represents a broad shift in the conditions of the labor force that should push us to broader labor market reforms about issues involving minimum wages, overtime pay, unemployment insurance, or other benefits for these kinds of workers. I discussed one such proposal in "New Rules for the Gig Economy?" (December 9, 2015). The problem at this stage is that it isn't yet clear, even roughly, how many workers are facing what kinds of problems. Full-time workers picking up some extra income in the gig economy are one thing. Part-time workers or those without other jobs who are earning most of their income in the gig economy pose other issues. With a broad array of new labor force relationships, it's  very hard to sort out the costs and benefits of different set of rules--especially given the ability of on-demand firms and workers to alter their labor force relationships in response to any new rules that are enacted.

Do Business Cycles Die of Old Age?

Whenever the US economy looks shaky, one of the most common questions I hear is whether this recovery has "run its course" or "gotten old." The downturn of the US economy during the Great Recession ended back in June 2009, so it's now been about 80 months of an economy on an (often frustratingly slow) upswing. There's a basic statistical answer to this question, but there's also a broader issue that tackles of how to think about a "business cycle."

When looking at the path of economies over time, you see recessions and recoveries. But there also a well-known is a common cognitive pattern of "paraedolia," which refers to looking at randomness and perceiving patterns that aren't really there.  Even using the conventional term "business cycle" for patterns of recession and recovery hints at a belief that the economy is be based on underlying patterns and dynamics that will cause it to rotate in a preordained way from recovery to recession and back again. When people ask whether the recovery is "getting old" or has has gone on "long enough," they are presuming this kind of "cycle."

The statistical answer to whether economic upswings die of old age can be answered statistically, and Glenn D. Rudebusch summarizes the conventional wisdom very nicely in "Will the Economic Recovery Die of Old Age?"  written as the Federal Reserve Bank of San Francisco "Economic Letter" for February 8, 2016. Rudebucsh uses a kind of graph called "survival analysis," which can be applied to people's chance of dying, to part of a machine wearing out or breaking, to whether economies fall into recession, and many other applications.

As an example, here's a survival curve for the probability of an American male dying in the next year, The graph shows that the chance of dying in the next year doesn't rise very much at all for men up to the age of about 50 or 60, but then it starts to rise steadily with age.


A survival curve for the economy asks a question like: "What's the chance of an economic recovery ending in the next month?" Based on data for US business cycles going back to 1858, the patterns look quite different for before and after World War II. Here's Rudebusch's figure. Before World War II, there was a substantial rise in the chance of recession as an expansion aged: that is, after about four years, the chance of a recession int he next month has reached 20% and climbing. But since World War II, the chance of a recession rises by comparatively little as a recession ages: it's maybe  2% chance of recession in the next month after four years, but still only a 4% chance of recession in the next months after 10 years.



In short, US business cycles in the last 70 years or so don't seem to have a natural lifespan.

However, the notion of predictable cycles was once very hot stuff in the economics profession. One classic exposition is the great economist Joseph Schumpeter's 1939 book on Business Cycles (an abbreviated version is available on-line here). Schumpeter suggested that the rise and fall of the economy could be understood through a mixture of three different kinds of cycles: short-run, medium-run, and long-run. The short-term 3-5 year cycles were called Kitchin cycles, and Joseph Kitchin argued in 1923 that they were based on variations in of psychological factors and crop yields. (If you want more, see Joseph Kitchen, "Cycles and Trends in Economic Factors," Review of Economics and Statistics, January 1923, 5:1, pp. 10-16.) The medium-run Juglar cycles were based on fluctuations in levels of fixed investment often stemming from waves of innovation, as first argued in an 1869 book by by Clément Juglar (a condensed English translation is available here). The long-run Kondratieff cycles happened every 50 years or so, give or take a decade or two, and were based on major technological shifts (an English translation of Nikolai Kondratieff's 1922 article is available here).  For example, Schumpeter suggests in his 1939 book that a Kondratieff cycle had run from the start of the Industrial Revolution in the 1780s up through 1842, when it was followed by what he called "the age of steam and steel" from 1842 and 1897, and then age of "electricity, chemistry, and motors" after about 1898.

But as Schumpeter was quick to note, the idea of three overlapping cycles wasn't meant to be definitive. He wrote: "There are no particular virtues in the choice made of just three classes of cycles. Five would perhaps be better, although, after some experimenting, the writer came to the conclusion that the improvement in the picture would not warrant the increase in cumbersomeness."

For the modern economist, this notion of maybe three or maybe five overlapping cycles, happening over maybe 3-5 or 7-11 or 40-60 years, sounds a lot like an attempt to impose an overall template pattern that isn't really there on an essentially random set of events. Sure, one can look back after the fact and analyze the proximate causes of recessions, like the Federal Reserve raising interest rates to fight inflation in the early 1980s, or the aftermath of the dot-com investment boom in the later 1990s, or the housing price bubble leading up to the Great Recession. But those proximate causes were not an inevitable cycle; instead, they were the result of other economic events and policy choices.

So the good news is that the US economy doesn't seem to be doomed by any mechanical law of aging recoveries to enter a recession soon. After all, there was a period between recessions in the 1960s that lasted 106 months and the another period between recessions from the 1990s into the early 2000s that lasted 120 months. But on the other side, the US economic recovery is far from bulletproof, and remains vulnerable to twists of policy and fate.

Will Peak Oil or Renewables Make Climate Change Moot?

Hopeful onlookers sometimes point to two possible escape hatches from the problems of burning fossil fuels. One escape hatch is "peak oil"--that is, the argument that production of fossil fuel resources is near or its peak. In this view, the impending fall in fossil fuel production might well bring higher prices and other economic hardship, but at least emissions from burning fossil fuels would drop. The other escape hatch is a large rise in cost-competitive non-carbon sources of energy, like solar and wind, but also nuclear and hydroelectric power. If these sources of energy undercut fossil fuels on price, then the economy could make a transition away from fossil fuels to an economy that used on cheaper and abundant energy from these other sources.

But there's yet another possibility, and it's the one laid out by Thomas Covert, Michael Greenstone, and Christopher R. Knittel in their article, "Will We Ever Stop Using Fossil Fuels?" appearing in the Winter 2016 issue of the Journal of Economic Perspectives.  In this outcome, supply of fossil fuels isn't going to run out in the next few decades, and alternative non-carbon energy sources aren't going to become cost-effective for enough uses in that timeframe to substantially reduce consumption of fossil fuels, either. One might wish it was otherwise. As the authors write: "After all, who wouldn’t prefer to consume energy on our current path and gradually switch to cleaner technologies as they become less expensive than fossil fuels? But the desirability of this outcome doesn’t assure that it will actually occur—or even that it will be possible." Because they believe that neither of the two escape hatches from the problems of burning fossil fuels are likely to be available, they argue that addressing issues like climate change and conventional air pollutants will require a strong policy intervention to reduce the use of fossil fuels.

When it comes to the supply of fossil fuels, an important lesson to remember that technological progress happens in many areas. It happens in solar and wind power, but it also happens in finding, developing, and extracting fossil fuels. examples include the discovery of  how to drill in ever-deeper water, as well as the more recent developments in getting oil and gas from tar sands and from hydraulic fracturing, As the authors write: "It is an empirical regularity that, for both oil and
natural gas at any point in the last 30 years, the world has 50 years of reserves in the
ground. The corollary, obviously, is that we discover new reserves, each year, roughly
equal to that year’s consumption." Here's a figure showing the growth proven reserves of oil and gas reserves over time.

"Proven reserves" is a specific term referring to reserves that are available at (more-or-less) current prices, and given current levels of technology.  Geologists also estimate fossil fuel "resources," which are the quantities of fossil fuels known to exist, but not economically viable--yet. The known resources are maybe 3-4 times the size of the "proven reserves. And then there are enormous other fossil fuel resources, like oil shale and methane hydrates, which are not currently counted as either reserves or resources, but technological  developments over time could bring them into the market as well.  As Covert, Greenstone, and Knittel write: "If the past 35 years is any guide, not only should we not expect to run out of fossil fuels any time soon, we should not expect to have less fossil fuels in the future than we do now. In short, the world is likely to be awash in fossil fuels for decades
and perhaps even centuries to come."

When thinking about non-carbon technologies, it would take a book-length manuscript to go through all the possible developments. The authors thus focus on a few key points. Global demand for energy seems certain to rise dramatically in the decades ahead with overall economic development in today's low-income and emerging economies. The question about non-carbon energy sources is not whether they will expand (spoiler alert: they will expand), but whether they will expand so quickly and dramatically that they undercut fossil fuels in a wide array of uses. This outcome may be desirable, but that doesn't make it likely or even possible. As the authors write:

[T]he International Energy Administration Agency (2015) projects that fossil fuels will account for 79 percent of total energy supply in 2040 under the current, business-as-usual policies, which already takes into account some rise in these alternative noncarbon energy production technologies. In the medium-run of the next few decades, none of these alternatives seem to have the potential based on their production costs (that is, without government policies to raise the costs of carbon emissions)
to reduce the use of fossil fuels dramatically below these projections.

The paper offers a few comments in passing about carbon capture technology, nuclear power, and hydro power, but the main focus is on solar and wind technologies as alternative methods of generating electricity, and on whether developments in battery technology will make fully electric cars viable. Overall, Covert, Greenstone, and Knittel write:

Our conclusion is that in the absence of substantial greenhouse gas policies, the US and the global economy are unlikely to stop relying on fossil fuels as the primary source of energy. The physical supply of fossil fuels is highly unlikely to run out, especially if future technological change makes major new sources like oil shale and methane hydrates commercially viable. Alternative sources of clean energy like solar and wind power, which can be used both to generate electricity and to fuel electric vehicles, have seen substantial progress in reducing costs, but at least in the short- and middle-term, they are unlikely to play a major role in base-load electrical capacity or in replacing petroleum-fueled internal combustion engines. Thus, the current, business-as-usual combination of markets and policies doesn’t seem likely to diminish greenhouse gases on their own.

Twenty Years Since the Welfare Reform of 1996

Twenty years ago in 1996, President Bill Clinton signed into law the Personal Responsibility and
Work Opportunity Reconciliation Act, more commonly known as "welfare reform." The welfare reform of 1996 sought to "end welfare as we know it," as President Clinton had often stated. The Winter 2016 issue of the Journal of Policy Analysis and Management has a "Point/Counterpoint" exchange on the effects, which at least for now is freely available on-line, although many readers will also have access through library subscriptions.  The intellectual combat here isn't in the binary, black vs. white. fire vs. ice, war-of-the-worlds style. Instead, Ron Haskins takes the position the glass-half-full position in "TANF At Age 20: Work Still Works" (pp. 224-231), and then the team of  Sandra K. Danziger, Sheldon Danziger, Kristin S. Seefeldt, and H. Luke Shaefer takes the glass-half-empty position in "From Welfare to a Work-Based Safety Net: An Incomplete Transition" (pp. 231-238). The authors then offer a response-and-rejoinder to each other, as well.

In his overview called "Welfare Reform: A 20-Year Retrospective," Richard V. Burkhauser offers some reminders of the intensity of the rhetoric back in 1996 when the welfare reform bill was on the verge of being signed into law. On one side, here's Democratic New York Senator Daniel Patrick Moynihan:

“The welfare bill terminates the basic federal commitment to support dependent children. It endangers children with absolutely no evidence that this radical idea has even the slightest chance of success . . . The current batch in the White House have only the flimsiest grasp of social reality, thinking anything doable and equally undoable. As, for example, the horror of this legislation ...” 

And here's an opposing view, from Republican Florida Congressman Clay Shaw:

“When the Senate passed, with a good bipartisan vote, with half the Democrats joining the Republicans, I began to think the President would sign this bill . . . July 31st has got to go down as Independence Day for those who have been trapped in a system that has been left dormant and left to allow people to actually decay on the layers of inter-generational welfare which has corrupted their souls and stolen their future . . . It will work ..." 

In terms of nomenclature, the previous welfare program called Aid to Families with Dependent Children (AFDC) now became Temporary Assistance for Needy Families program (TANF). The change in name was mean to reflect a change in emphasis. AFDC had been an "entitlement" program, meaning that if you qualified for the program as a low-income family with children, you were entitled to the payment. The central change in TANF was that welfare became a work-based program. Recipients now had to either work or be preparing for work through education or job training to be eligible for welfare. In addition, time limits were placed on the length of time welfare could be received during a lifetime.

The adoption of TANF was not the disaster that some had predicted. Welfare enrollments did drop dramatically and work levels rose, as Haskins describes:

The welfare caseload, which had increased almost every year since the beginning of the War on Poverty in the mid-1960s, fell every year after 1994 before increasing slightly during the Great Recession of 2007. Over the six years between 1994 and 2000, the caseload fell by about 60 percent to a level roughly equal to the 1971 level. The decline in the rolls over this period was accompanied by a 16 percent increase in work by all single mothers and a 35 percent increase in work by never-married mothers, the subgroup of single mothers who were most likely to go on welfare. Meanwhile, the poverty rate among single mothers and their children fell from 44 percent to 33 percent, a decline of 25 percent to its lowest level on record.

Moreover, poverty rates among the key group of households headed by single mothers. Here's a figure from Hansen's paper. The top line shows the poverty rate in this group if you look at earned income only. The next line down shows the poverty rate if you include cash benefits including TANF, but also unemployment insurance, general assistance, Supplemental Security Income, and others). The next line down adds in the value of food stamps. The line under that adds the value of the Earned Income Tax Credit. The bottom line also adds income from other household members and government stimulus/recovery payments. Right after 1996 welfare reform, the poverty rate for this group declines, and at least according to the bottom line in the figure hasn't changed much since then.

But of course, nothing is truly simple in social science. As all the authors in the symposium point out, the welfare reform law of 1996 was operating in a broader economic and policy context. The economic context was the dot-com boom of the late 1990s. When President Clinton signed the welfare reform law in August 1996, the unemployment rate was 5.1%, and it would fall all the way to 3.9% by late in 2000. In other words, it was a good economic time to impose work requirements.

The policy context was that the shift in welfare from AFDC to TANF was accompanied at about the same time by a substantial shift in other programs to provide work support. As the Danziger et al. team of authors point out, "For example, the Earned Income Tax Credit (EITC), which provides tax credits to families with children and low earnings, was increased significantly in the early 1990s, the minimum wage was increased in 1997, and access to medical care was expanded by the State Child Health Insurance Program of 1997." For an earlier post about how a Congressional Budget Office report on how government work-support programs largely offset the falls in welfare payments, see "Where Has Welfare Reform Taken Us?" (January 23, 2015).

The broad shift to work requirements as part of welfare remains popular, and on its own specific terms, successful. Both sets of authors in this symposium support the shift, which represents a real change from how welfare was regarded before 1996. As Haskins writes: "Low-income working families with children receive more help from government than ever before—and there is bipartisan agreement that this is good policy." But the shift raises some obvious questions. What about assistance to those adults who are disconnected from work and don't have children, or to parents who are disconnected from work and do have children?  And given that nearly half of all households headed by single mothers do not earn enough to be above the poverty line based on their own income, as shown in the figure above, what can be done to raise the payoff for low-skilled and low-wage work?

The Danziger, Danziger, Seefeldt, and Shaefer group writes: "We want to make the work-based
safety net more effective without returning to AFDC." Their opening essay lists four proposals along these lines, and the follow-up essay lists four more. I'm not endorsing everything on their list, but here are the eight items, in two groups of four:

1. Adoption of a public responsibility to provide work opportunities to those for whom employer demand is limited, especially when unemployment is high. This includes transitional jobs or public subsidies to private-sector firms, nonprofit agencies or government agencies.

2. Expanded child care subsidies.

3. Reducing barriers to TANF entry by requiring states to spend a larger fraction of block grant funds on cash assistance and raising the TANF block grant to reflect economic and demographic changes. 

4. Modifying the SSI [Supplemental Security Income] program by adding part-time or temporary disability benefits. ...

First, because the federal minimum wage has not increased since 2009, an increase would reduce earnings poverty for single mothers, who are disproportionately represented among minimum wage workers. The Congressional Budget Office (2014) estimated that a $10.10 per hour wage would raise earnings for 16.5 million workers and reduce poverty by about 900,000, while reducing employment by about 0.3 percent (about 500,000 jobs).

Second, Hoynes (2014) proposes to raise the EITC for families with one child so that it is equivalent to that of two-child families, adjusted for family size. This would increase the maximum EITC for one-child families by about 40 percent for those in the bottom two quintiles. ...

Third, Ziliak (2014) proposes to convert the Child and Dependent Care Credit from a nonrefundable to a refundable credit. For example, for children under the age of five whose families have less than $25,000 in adjusted gross income, the refundable credit would be $4,000 for the first child in a licensed facility and half that for a child in an unlicensed facility. Low-income families do not benefit much from the current credit because they have little taxable income.

Fourth, families with incomes below $3,000 do not benefit at all from the $1,000 per child tax credit and other low-income families do not receive the full credit because their income tax liability is lower than $1,000 per child. The Center for American Progress (2015) has proposed making the credit fully refundable. This would provide needed cash income for the disconnected and additional income for TANF recipients, particularly if the credit could be delivered on a monthly basis.

These changes are essentially incremental. For example, proposals like a substantial increase in the Earned Income Tax Credit for all recipients are not includes. Moreover, I don't think of all of these changes as being work-related: for example, making the Child and Dependent Care Credit refundable doesn't encourage work in any direct way. Nonetheless, I'm broadly sympathetic toward proposals that support the households of low-income workers, and especially those with children.

Breaking Down US Inflation Rates by Category

Since 2000, the Federal Reserve has focused on the Personal Consumption Expenditures price index for its primary measure of inflation, rather  than the better-known Consumer Price Index. (For the reasons behind this choice and distinctions between the measures, see this post on "Consumer Price Index vs. Personal Consumption Expenditures Index," from January 17, 2012.) The PCE price index can also be broken down into a bunch of price indexes by type of product, and comparing these subsidiary price indexes with the overall PCE price index offers some views on long-term patterns of what drives inflation.

The best-known breakdown of the PCE price index, and the one used by the Fed, is to focus on the "core" price index that excludes food and energy prices. The blue line shows the "core" index, while the red line shows the overall PCE index. As you see, they are much the same over time, but the overall index fluctuates more because it includes the comparatively volatile energy and food prices.



Just how volatile are energy prices? The price index for "energy goods and services' appears in blue. it's highly correlated with the rises and falls in crude oil, and as you can see it dances around considerably. (Everything is measured here as the percent change from a year earlier.) The reason that overall PCE price index is currently lower than the "core" index in the figure above is largely because the core index doesn't include energy.




Although the core PCE price index also leaves out food, the volatility in prices for food is a lot less extreme than for energy prices. The blue line shows the price index for food, with the overall PCE price index appearing in red.




Another way to slice up the overall PCE price index is to look at changes in the price index for goods and for services separately; moreover, we can separate out durable and nondurable goods, and further separate out some more specific categories of interest. For example, the price index for durable goods appears here in blue, with the overall PCE price index again appearing in red for comparison. What's interesting here is that inflation in durable goods prices has been consistently below the overall inflation rate for decades; indeed, the inflation rate for durable goods has been consistently negative since the mid-1990s. Part of the reason here is that technological change in some durable goods like those related to computing and information technology has been so extreme that goods have become cheaper over time; another part of the reason is the role of cheap imported products in holding down prices.



The blue line in the next figure shows a subset of the durable goods price index, which is the price index for the subcategory of "video, audio, photographic, and information processing equipment and media." Again, for comparison the overall PCE price index is shown by a red line. In this subcategory, the inflation rate has actually been negative for most of the years back to 1960, except for a period in the 1970s. Moreover, the deflation rates in this subcategory have consistently been around 10% per year and sometimes even lower for two decades.



In comparison, the price index for nondurable goods, shown below by the blue line, is much closer to the overall PCE price index. Nondurable goods includes energy, which helps to explain why the blue line fluctuates more than the red line which again shows the overall PCE index.



The price index for services, on the other side, tends to be somewhat higher than overall PCE price index, as shown in the graph below by the blue line often being a bit above the red line in recent decades. About two-third of consumer expenditures on on services, rather than goods.



Within the category of services, what are some of the major subcategories that are tending to keep inflation higher in this broad area? The blue line in this figure shows the inflation rate for housing and utilities (which includes both the price of renting a home and an imputed price when owners "rent" their own house to themselves) which tended to be lower than the overall PCE inflation rate shown by the red line in the 1970s and 1970s, but was often higher from the 1980s up through the early 2000s.



Health care costs in the PCE index include both what people pay out of pocket, and also the costs of health insurance premiums payed by employers on people's behalf. Inflation in those health care costs, as shown by the blue line, was consistently higher than the overall PCE index for most of the time up to the last decade or so.



The price index for education services includes what people spend on higher education, as well as private schools. Inflation in this area has been consistently above the overall average of the PCE price index.


I'll also toss in one component of education that isn't included under services, but rather under goods: the price index for educational books. For several decades, it has been rising more rapidly than the overall PCE index.


These sorts of figures and tables aren't the final word, of course. There are hard questions in measurement of inflation that have to do with making sure that when you measure price changes you also adjust for changes in quality, and also adjust for changes in the patterns of what people are buying. The PCE index arguably makes such adjustments  better than the Consumer Price Index, but the problems remain large.

But with those kinds of concerns duly noted, it seems fair to say that over the long-term, the process of US inflation is a balancing act between price inflation for services that rise at a higher-than-average rate and price inflation for durable goods that rises at a lower-than-average rate.