Bi-Weekly "Metapolitics" Discussion
Details
I've decided to host our discussion at the Good Karma Café again. It's near the corner of 10th & Pine in Washington Square West neighborhood, and it's fairly easy to get to if you're using public transit. With SEPTA, take the Broad Street Line & get off at the Lombard South Station, and walk 2 blocks east on Lombard and then turn left on 10th street and it's up a half block. With PATCO, it's a bit more of a hike - get off at 15th & Locust stop and walk 6 blocks east on Locust to 10th, then turn right and walk 2 blocks south. For those who are driving, parking in the neighborhood can be tough to find. I'd suggest parking in the parking deck connected to the ACME at 10th & South.
Since we're using the cafe's space, they ask that each person attending the meetup at least purchase a drink or snack. Please don't bring any food or drinks from outside. If you're hungry enough to eat a meal, they have more substantial fare such as salads, soups & sandwiches which are pretty good and their prices are reasonable - check out the menu here: http://www.thegoodkarmacafe.com/our-menu/
--------------------------------------------------------------------------
WAGE STAGNATION, POVERTY & COST DISEASE
In our last discussion on the patterns of political violence, we discussed Peter Turchin's "structural-demographic theory" that predicts we'll see a rise in social conflict peaking around 2020, and he attributed this to 3 interlocking factors -- popular immiseration, elite overproduction & fiscal decline of the state. This meetup will focus on the 1st factor -- "popular immiseration" (i.e. decline in the public's material wellbeing) -- and try to determine the root causes. We'll focus on the issues raised by economists & political pundits in debates over 3 interrelated problems: wage stagnation, poverty, and the "cost disease" afflicting healthcare, housing & education.
The wage stagnation debate (which was hottest around 2011-2012) mostly focuses on allegations that average wages became decoupled from rising productivity in the mid 1970s, and that the capitalist elites have commandeered this income for themselves, causing income inequality to rise and making it harder for the lower & middle classes to afford basic consumer goods. Some pundits have connected wage stagnation in the U.S. with the rise of Reaganomics & Cintonomics enabling the 1% to siphon off all the productivity gains, or the U.S. leaving the gold standard which allowed the FED to debase the dollar with inflation. Others allege that it's a more global phenomenon connected to long-term technological stagnation outside IT industries, unemployment & underemployment due to automation & outsourcing of manufacturing jobs, and/or low-skilled immigration driving down wages & overburdening social services in the developed world.
LBJ's "War on Poverty" turned 50 in 2014, prompting several retrospective analyses of its anti-poverty programs and the unfortunate fact that the poverty rate (% of household earning less than half the median household income or about $26,000/yr or less in today's dollars) has remained fairly flat for the past five decades, fluctuating between 11-15%. Some economists & pundits have recently suggested that we switch from means-tested welfare programs (which have a high overhead due to large administrative cost) to a basic income guarantee that just sends poor people a set amount of cash each month to help them meet their basic needs. However, there's fears that this may just cause their cost of living to rise and primarily benefit their landlords & healthcare providers.
The "cost disease" debate arose this spring in response to Scott Alexander's essay in February and in retrospectives of the economist William Baumol's work following his death in May. The debate aims to answer the question over why the costs of many consumer goods have fallen over time, but several essentials have stayed high in price or increased substantially. From ancient times up to the early 20th century, the poor typically had trouble affording anything and often went hungry, wore ragged clothes & lacked shoes, and had very little in the way of household goods like furniture, cookware & tools. Today, the poor in developed countries are often not only not in danger of starvation but actually overweight (although still lacking in nutrition) due to an overabundance of cheap fast food, and they can typically afford cheap clothing & household goods from discount stores like Walmart & Target and warehouse clubs like Sam's Club, BJ's & Costco. They can often even afford electronic goods, like smart phones & flatscreen TVs, that were only owned by the wealthy a decade ago. Yet, the costs of healthcare, education and housing continue to rise and these necessities appear to be increasingly in need of government subsidies so that the not only the poor but also the working class & middle class can afford them. There's a variety of theories for why the costs in these 3 sectors continue to rise, and we'll explore them in our discussion.
https://secure.meetupstatic.com/photos/event/b/0/0/e/event_461505070.jpeg
We'll spend the last portion of our discussion on a "meta-topic" -- when & how should we apply tools like Occam's Razor and the Pareto Principle to complex social trends that have been attributed to a large number of factors? Scott's article in this section refers to the various theories about the factors behind the crime rate drop in the 1990s (which we discussed 2 weeks ago in the Skeptics meetup), but the general principles he discusses applies to all types of social & economic analysis. Scott Alexander & Stefan Schubert suggest that the probability of multi-factorial trends is fairly low, whereas Joshua Herring & Andrew Gelman partially agree with them but argue that the "root cause" of complex social phenomena may be something we can never really isolate & we may have to settle for complexity.
The articles & videos you see linked below are intended to give you a basic overview of the major points in the wage stagnation & cost disease debated. For the 1st section, please try to watch both videos (a little under 9 minutes total) and read the 2 short articles. The 2nd section is also very brief - two short articles and two videos totaling just over 9 minutes. As far as the 3rd section, I certainly don't expect everyone to read all of the articles, since the 2 by Scott are quite long. If you're pressed for time, just read the short article by Tyler Cowen & glance at the OECD stats on healthcare spending. (As you can see, I wrote down some notes summarizing the main points under Scott's 2 blog posts so you don't have to read them if you don't have time.) The 4th section is also just 4 short blog posts which should be easy to digest, but it's a rather "wonky" topic so if you're pressed for time just reading Scott's post should give you the general gist.
NOTE: As usual, if you're interested, I've compiled a bibliography of articles & videos on the subject we're covering here which provide more detail: https://www.meetup.com/Philadelphia-Political-Agnostics/messages/boards/thread/50862044
IS IT TRUE THAT WAGES IN THE U.S. HAVE STAGNATED SINCE THE '70s AND HAVEN'T KEPT PACE WITH THE COST OF LIVING? IS THIS SOMETHING UNIQUE TO AMERICAN CAPITALISM OR IS IT A GLOBAL PHENOMENON?
- Gulzar Natarajan, "The Great Wage Stagnation" (short blog post)
http://gulzar05.blogspot.com/2011/11/great-wage-stagnation.html
Natarajan cites a report from the Resolution Foundation that shows that looked at 10 developed countries (US, UK, Japan, Canada, Australia, France, Germany, Denmark, Finland, Sweden) and found that all of them had experience wage stagnation in the 2000s, and some had experience it in the '80s & '90s as well, and that wages as a share of total employee compensation has been declining as well (i.e. more pay was in the form of non-wage benefits).
- Don Boudreaux, "The Real 'Truth About the Economy:' Have Wages Stagnated?" (video - 5:13 min.)
https://www.youtube.com/watch?v=s6FmhXQ32Wo
Boudreaux take issues with Robert Reich's video "The Truth About the Economy" that claimed, adjusting for inflation, most Americans' wages have stagnated for the past 30 years. Boudreaux has 3 counterarguments: (1) He outlines that using different measures for inflation yields different results, and while a CPI adjustment shows a 4% average wage decline over the past 30 years, the PCE shows a 10% rise in wages and GDP deflator shows a 18% rise. (2) He argues that once we adjust for inflation and account for non-wage benefits, we see a 26% rise in average wages between 1976 & 2006. (3) He emphasizes that income brackets aren't individuals -- i.e. most individual Americans have seen their wages increase over the years as they've learned new skills & advanced their careers, but more low-skill immigrants joined the American workforce and pulled the average wage down.
- Steve Horwitz, "Is the Cost of Living Really Rising?" (video - 3:31 min.)
https://www.youtube.com/watch?v=W8SLIt7xZxU
Horwitz argues that by comparing the average cost of goods & services to the average private sector wage, we can see that the cost of many of them has declined (food, household appliances). He also emphasizes that a lot of the comparable products are much higher quality - e.g. cars have gone up slightly in average price, but they tend to get better gas mileage & overall mileage.
- Bourree Lam, "The Surging Cost of Basic Needs" (short article)
https://www.theatlantic.com/business/archive/2016/06/household-basic-spending/485330/
A report from the Brookings Institute found that low-income households in the U.S. are devoting a greater share of their budget to basic needs compared with 30 years ago. Low-income and middle-income households (defined as the lowest and middle quintiles of the income distribution) now spend roughly 80% of their budget on housing, food, transportation, health care, and clothing. For low-income households, 40% of their budget went to housing—a 5.5% increase from 1984. Over the past 30 years, low-income families have seen their real expenditures decline by 4.5% whereas middle- and high-income families have spent more. And yet, as low-income families spend less on basic needs in real terms, because their real earnings have stagnated, the proportion of their budgets devoted to basic needs has gone up. This not only means less discretionary spending for low-income families, but it raises concerns that low-income families might be cutting back spending on food in order to compensate for rising costs in housing & health care.
WHY AREN'T WE WINNING THE "WAR ON POVERTY"? WOULD SWITCHING FROM MEANS-TESTED WELFARE PROGRAMS TO A BASIC INCOME GUARANTEE HELP ALLEVIATE POVERTY?
- Derek Thompson, "The War on Poverty Turns 50: Why Aren't We Winning?" (short article)
Thompson assess the legacy of LBJ's War on Poverty and notes that the poverty rate has only mildly declined from 19% to 15% over the last 50 years. This may be partly due to lower-income wages falling behind inflation and rising cost of living. However, poverty is mostly a result not of low wages but unemployment & part-time employment. The poverty rate for full-time workers is only 3%, but for those who only part time it's 16.6% and for those don't work at all it's 33%. He also notes that those who look at the official unemployment statistics often overlook the fact that the overall "labor force participation rate" has retreated to its lowest point since the 1970s (63%), partly because America is aging and partly because people (mostly men) have dropped out of the workforce.
Thompson also notes that poverty is tied to the decline of marriage and the two-parent family. Among two-income married couples (61 million people), the poverty rate is virtually zero, and among marriages where one person works and the other doesn't (another 36 million Americans) the poverty rate is just under 10%. However, among the 62 million single-parent families, the poverty rate is quite high: 25% for single dads & 31% for single moms. This is tied to unemployment -- 41% of single parents (26 million households) don't work at all, and another 11% work only part-time. While a Columbia study found the growth of government benefits since 1967 has cut an alternative measure of poverty by 40%, it's more precise to say that Washington has been effective at helping working, two-parent families escape from poverty.
- Michael Tanner, "The War on Work" (video - 5:53 min.)
https://www.youtube.com/watch?v=1nN1HqAps4Y
Tanner's video summarizes the main points from a Cato Institute study that found a typical family with 2 young children under 5 can earn more in welfare benefits than from one parent earning minimum wage in 33 states (benefits range from $16L-$49K/yr). In 13 states, a single mother with 2 children can receive more in welfare benefits than she could earn working full time at $15/hr, and in 8 states it's more than the earnings from a $25/hr job. He also notes that the "workfare" requirements from the Welfare Reform Bill on 1996 have been decreased, and now nationwide fewer than 42% of welfare recipients are participating in "work activities" (working or looking for work, going to school or job training). He argues that welfare creates long-term unemployment by disincentivizing work, since people often lose benefits if they take work & earn more than the cutoff.
Tanner's study was criticized by Josh Barro, Rick Ungar & Elise Gould, said that it drastically overestimates the amount of welfare benefits the average recipient receives. Gould points out that in 2009, average transfer income for the lowest fifth of workers was $4,633 and average labor income was $12,871. (To be comparable with the Cato report, this doesn't include Medicare and Social Security income.) This means government transfers are far less than what Tanner claims, and labor income far exceeds government transfers for the lowest income group, meaning that real-world low-income families don’t feel so coddled by lavish welfare benefits that they don’t need to work.
- Matt Zwolinski, "Benefits of a Basic Income Guarantee vs. Welfare" (video - 3:30 min.)
https://www.youtube.com/watch?v=tUYbUKymgTA
Zwolinski points out that the federal, state & local governments in the U.S. combined spend about $1 trillion/year on anti-poverty programs (about $20K/yr for every poor person), but a lot of this money goes to administrative overhead rather than to welfare recipients. He suggests that if we scrapped all the means-tested welfare programs and replaced them with an unconditional basic income guarantee, He argues it's superior for 3 reasons: (1) it's simpler & easier for applicants, (2) it gives welfare recipients more freedom (e.g. they can save money & shift spending), (3) it treats everyone the same & is less susceptible to being gamed by special interests.
- Martin Farley "Why Land Value Tax and Universal Basic Income Need each other" (short article)
Farley argues that a universal basic income should be funded through a land value tax (along with a tax on other public resources such as licenses on minerals, airspace, spectrum bandwith, sea resources, public risk underwriting & government licences) for 3 reasons: (1) It's more secure - i.e. it is based on resources that are unlikely to disappear; (2) It doesn’t distort economic decision making; (3) It prevents the basic income from being captured by rent-seeking landlords & other businesses.
WHAT'S CAUSING THE "COST DISEASE" IN HEALTHCARE, EDUCATION & HOUSING? IS "COST DISEASE" A UNIQUELY AMERICAN PROBLEM OR A GLOBAL PHENOMENON?
- Tyler Cowen, "This Economic Phenomenon Is Making Government Sick" (short article)
https://www.bloomberg.com/view/articles/2017-01-18/this-economic-phenomenon-is-making-government-sick
Cowen gives a basic overview of the "cost disease" afflicting healthcare & education and focuses on 2 possible factors behind it: (1) noncompliant patients who don't maintain their health (and slacking students who don't study) and (2) the "ratchet effect" in government spending (i.e. Wagner's Law) due to the public's aversion to losing benefits they've become accustomed to.
- Scott Alexander, "Considerations on Cost Disease" (long blog post)
https://slatestarcodex.com/2017/02/09/considerations-on-cost-disease/
Scott present the statistics and shows that in the past fifty years, education costs have doubled, college costs have dectupled, health insurance costs have dectupled, subway costs have at least dectupled, and housing costs have increased by about fifty percent. US health care costs about four times as much as equivalent health care in other First World countries; US subways cost about eight times as much as equivalent subways in other First World countries. He presents several possible causes that may explain at least some of these rising costs: (1) Baumol's cost disease - i.e. natural rise in wages in labor intensive industries; (2) inflation (and difficulties in adjusting for it); (3) market failures due to the complexity of consumer choices, (4) government inefficiency in running public schools, mass transit & hospitals; (5) indirect government intervention through regulation; (6) increased regulatory complexity due to liability concerns; (7) decreased public risk tolerance; (8) free riding by those who default; (9) increases in non-wage compensation for public employees.
- Scott Alexander, "Highlights From The Comments On Cost Disease" (long blog post)
https://slatestarcodex.com/2017/02/17/highlights-from-the-comments-on-cost-disease/
Scott presents excerpts from various economists, journalists & rationalists who commented on his cost disease article. They attribute cost disease to a variety of factors: administrative bloat (John Cochrane); replacing cheap “inferior” goods with expensive “superior” goods whose quality is difficult to judge (David Manheim); increasing institutional risk tolerance & organizational complexity (anonymous hedge fund manager); gov't subsidies & entry barriers, market failures & information asymmetry due to lack of transparency (Noah Smith); gov't subsidies & entry barriers, consumer ignorance, and the US's clunky regulatory state (Megan McArdle); gov't subsidies & overregulation, growing risk aversion (Scott Sumner); market failures (Sohois); wealth extraction in markets that are complex, opaque, highly-regulated, unavoidable, limited-quantity, or intermediate (Andrew Swift); rising litigation & insurance costs and regulatory burdens (Habu71, John Schilling, Alex Zavoluk, CatCube); hidden recession since the 1970s ameliorated by manufacturing productivity gains (Doug); gov't subsidies & risk aversion (B. Kearns); diminishing returns from servicing marginal customers (fc123); gov't subsidies & entry barriers (Luke Hamilton); capitalists appropriating surplus value (Ben Wave).
Scott points out that most of these explanations fall into 3 general categories and he argues all are insufficient: (1) "Superior good substitution" ignores that many poor people can't afford superior goods & would prefer low-cost alternatives, (2) the Marxist concept of "surplus value appropriation" can't explain why the cost increases are in nonprofit industries such as schools & hospitals, and (3) "Administrative bloat" can't explain why non-bloated alternatives don't arise in industries with fewer entry barriers, such as colleges, private schools & ambulatory clinics.
*Note: Scott's essay was published at The American Interest, and Adam Garfinkle responded to it and suggested 4 more causes of cost disease: (1) "Cost-Pull Progress" (similar to Jevons' Paradox) where new technology increases demand for expensive products & services, especially in healthcare; (2) Baumol's Cost Disease driving up labor-intensive industry employment (rather than wages) leading to Administrative Bloat; (3) a Service Industry Labor Market Glut due to women joining the workforce, (4) Increase Internal Transactional Costs (such as "overlawyering") due to diseconomies of scale resulting from industry consolidation.
Garfinkle says that a multi-faceted research project is needed to address cost disease which is too complex to unravel at this point: "It is obviously one thing to establish that a given would-be cause really is a cause, and another to assign relative weights to the factors established as contributing causes. And, as already suggested, the challenge may not be merely additive, since some factors may interact with and magnify other factors."
**Note: Oddly, one of the possible factors behind cost disease that didn't get mentioned in Scott's blog posts is that "costly signaling" with "positional goods" (status symbols) often creates a positive feedback loop that resembles an arms race. People have to spend more & more to maintain the same status (a.k.a. the "Red Queen effect"). Bryan Caplan has made this argument for college education, i.e. there's "degree inflation". Robert H. Frank has argued the interlinked nature of housing & public education and the way they signal status through exclusivity drives up costs in both. Robin Hanson has made a signaling argument for excessive healthcare spending, although he's not arguing that people tend to spend more on their own healthcare as a status symbol, but rather that many people advocate spending more tax dollars on healthcare to "show that they care" (i.e. social desirability bias).
- OECD.org, "Healthcare costs unsustainable in advanced economies without reform" (short article)
http://www.oecd.org/health/healthcarecostsunsustainableinadvancedeconomieswithoutreform.htm
Lest we think that runaway healthcare costs is something unique to the U.S., this report shows that healthcare spending has risen faster than economic growth in all OECD countries over the past 20 years. The report notes that public funds still account for around three-quarters of health spending in most OECD nations and this funding is heavily reliant on payroll taxes which will decline as their populations age, and so this jeopardizes the future of these nations' healthcare systems if reforms aren't enacted.
Another OECD report shows that the "cost disease" in housing costs isn't unique to the U.S. either. You can see in the chart below the percent of the average household's disposable income that is spent on housing and the US is the column 5th from the left at about 17.5%. Judging from the graph, only 2 countries have spend less than this.
https://secure.meetupstatic.com/photos/event/d/9/b/3/event_461695731.jpeg
OECD reports on higher education show that the U.S. spends roughly similar portion of GDP as most other developed nations on tertiary education, but more of this spending in the U.S. comes from private rather than public funds. Public and private spending on education continued to rise, even during the economic downturn of 2008-2009, with total investment increasing in 24 out of 31 OECD countries for which data are available. But while public spending on education as a percentage of total public expenditure remained at 13% on average across the OECD in both 2005 and 2009, it fell in 19 out of 32 countries over that period. Private funding, mainly from households, represents on average 30% of total expenditure on tertiary education. This proportion ranges from less than 5% in Denmark, Finland and Norway, to more than 40% in Australia, Israel, Japan and the United States, and to over 70% in Chile, South Korea and the United Kingdom.
HOW LIKELY IS IT THAT THERE'S MULTIPLE, UNCORRELATED FACTORS BEHIND THE "COST DISEASE" IN EDUCATION, HOUSING & HEALTHCARE? SHOULD WE ASSUME THERE'S A SINGLE UNDERLYING DYNAMIC LIKE RICARDO'S "LAW OF RENT" OR TURCHIN'S "ELITE OVERPRODUCTION"?
- Scott Alexander, "How Likely Are Multifactorial Trends" (short blog post)
http://slatestarcodex.com/2015/02/14/how-likely-are-multifactorial-trends/
At the end of the "Highlights From The Comments On Cost Disease" blog post, Scott linked his earlier essay on the likelihood of multifactorial trends, implying that he thinks it's unlikely that all or even a large number of the hypothesized factors behind cost disease can each be causing a small part of the cost increase. In this essay, Scott took issue with a Vox article about the 1990s crime rate drop that claimed there is no “smoking gun” and crime probably declined because of a bunch of reasons coming together perhaps about 10 factors each accounting for about 10% of the decline. If ten different factors caused the decline in crime, that would require that ten different things suddenly changed direction, all around the same time in the early '90s, which would be a pretty big coincidence. As a reductio ad absurdum, he suggests that we imagine it was ten million different factors, each accounting for one ten-millionth of the decline. Since there are about ten million criminals in the US, we could structure this as one factor per criminal. This would mean that, around 1994, each of America’s ten million criminals independently and coincidentally had a major life change that made crime seem less attractive. Scott suggests that this is ridiculously unlikely. Although he doesn't use this term, he appears to be thinking in terms of Occam's Razor and the "conservation of belief" - http://projects.csail.mit.edu/church/wiki/Occam%27s_Razor,and_the_Law_of_Conservation_of_Belief(ESSLLI)
- Stefan Schubert, "Multiple Factor Explanations Should Not Be One-Sided" (short blog post)
http://lesswrong.com/lw/kpj/multiple_factor_explanations_should_not_appear/
Scott says his reasoning in the above blog post is drawn in part from Schubert's classic Less Wrong essay that cautioned skepticism in the case of multiple factor explanations for trends that seem too one-sided. Schubert gives Jared Diamond's "Anna Karenina principle" for the rise of agriculture and Steven Pinker's multi-faceted explanation for the long-term decline in violence as examples of suspiciously one-sided trends. (Elsewhere, Schubert criticized Jonathan Haidt for claiming there were 10 causes of America's current political dysfunction.) He says there is something suspicious with only listing mechanisms that play in the one direction, and it's often easy to find important mechanisms that play in the other direction. In his view, countervailing factors on the other side of the scale leads to a better understanding of how historical trends have unfolded.
- Joshua W. Herring, "Causes of Many Moving Parts" (medium-length blog post)
http://theonlywinningmove.net/2015/02/15/causes-of-many-moving-parts/
Herring agrees with Scott that multifactorial trends are less likely than monofactorial ones; but he says he's not sure that this the various factors the Vox article mentions for the 1990s crime rate drop isn’t underlyingingly monofactorial. He argues that "general frustration with the crime level" could be the underlying – if admittedly unmeasurable – variable that caused most of the other factors (e.g. mass incarceration, more policing, gentrification, decline in crack use & gangs, less use of cash), although he also cites the general increase in information technology as another likely factor.
Herring argues that a lot of what seem like independent factors might actually stem from the a common source, in more or less roundabout ways. So, Scott Alexander is right that models that attribute trends to every available cause are unappealing, but he’s wrong that there’s always an alternative explanation that is measurable or even able to be articulated. Herring says the connections between factors he's hypothesizing may be of dubious scientific value in that they can’t be observed in any way. Since there’s just no way to talk about them meaningfully, at least not at the present level of scientific knowledge, we must content ourselves with complex causes. When you’re in a situation where the underlying causes are scientifically nebulous, you end up with what looks like a wild coincidence of unidirectional trends. And it’s worse for the fact that the underlying cause won’t have caused them all to the same degree, or in the same way. It’s just, there’s a "force," for lack of a better term, nudging everything in a particular way, and that comes out looking like a hugely coincidental jumble.
- Andrew Gelman, "Whither the 'Bet on Sparsity Principle' in a Nonsparse World?" (short blog post)
http://andrewgelman.com/2013/12/16/whither-the-bet-on-sparsity-principle-in-a-nonsparse-world/
Elsewhere, Gelman has expressed his opinion that Occam's Razor isn't a useful principle for social scientists because deliberately limiting the complexity of the model is not fruitful when the problem is evidently complex. Instead, if a simple model is found that outperforms some particular complex model, the appropriate response is to define a different complex model that captures whatever aspect of the problem led to the simple model performing well.
In this blog post, Gelman deals with the "bet on sparsity" principle (a.k.a. "sparsity-of-effects principle") which is similar to Occam's Razor in that it states that a system is usually dominated main effects and low order interactions. (The Pareto principle that 20% of the factors provide 80% of the effects is a crude version of betting on sparsity.) Gelman says this principle may be useful for the hard sciences, but he's less sure of its relevance for social sciences, where no effects are clean and no coefficient is zero, where every contrast is meaningful—but some of these contrasts might be lost in the noise with any realistic size of data. He thinks the solution may be that in a dense setting we are not actually interested in “recovering the underlying model" since the underlying model, such as it is, is a continuous mix of effects. If there’s no discrete thing to recover, there’s no reason to worry that we can’t recover it -- this is similar to what Herring said above.
Gelman does concede that sparse models can be faster to compute, easier to understand, and yield more stable inferences. He notes that some people say that a sparse model is less likely to overfit, but he objects that you can also get rid of overfitting by using a strong regularizer. However, he does think it's fair to say that a sparse model can yield more stable inferences, in that the inferences for the more complex model can be sensitive to the details of the regularizer or the prior distribution.
Bi-Weekly "Metapolitics" Discussion