Catherine Wolfram's excellent blog post about the proposed "Green New Deal" (GND) nudged me to write out a few thoughts. I recognize that we do not have a clear definition of what policies would be bundled into such a GND. At its heart, it must be a Keynesian large scale expenditure of federal $ to jump start the green economy.
When government spends $, balanced budget conditions require someone's taxes to rise. I will not comment further on this issue here. Let's focus on the winners from this increased government expenditure.
Can empirical economists working in the Chad Jones, Paul Romer endogenous technological change literature, estimate the marginal increase in patents as a function of new government expenditure? Patents can be measured both in terms of quantity and quality. David Popp has done some of the best work here. Will our scientists now be able to finance new equipment and launch great new projects because of the GND?
In past work, Austan Goolsbee argued that NSF budget expansions raised the salaries of scientists as they are inelastically supplied in the short run and a demand increase for basic science thus raises their wages.
The GND can be thought of as a "Big Push". Such a "Big Push" can switch the equilibrium an economy achieves in an economy where there are fixed costs to launching new green ventures. In the presence of fixed costs to creating a production line, entrepreneurs will only enter the "green space" if they anticipate that there is a market for new devices (such as solar panels) that they will produce. If the federal government commits to the Green Big Push, then more companies will enter this space and some will succeed. An open question here is whether the Federal Government can resist the urge to "pick winners". Who in the Congress will make the allocative decision concerning which firms will receive GND subsidies for their own nascent idea?
If the Congress could delegate the allocative decision of $ to experts, then would more voters be comfortable with launching the GND? David Weinstein's past work on MITI in Japan shows that this agency has not been effective at "picking winners" in its industrial policy approach to boosting the Japanese economy.
As Catherine discusses in her article, a carbon tax would direct economic activity towards the efficient allocation of resources. Will a command and control Keynesian expenditure increase achieve the same outcome?
-
Productivity report cards have been posted for academic economists. Without being too modest, here are my rankings;
overall (#340)
last 10 years (#141)
h-index (#315)
environmental (#11)
I think these imperfect indicators provide some benchmarks for judging whether tenured academics are still hard at work.
My current peer group
-
Jan3
Designing a Carbon Tax and Dividend Scheme When People Differ by Place of Residence, Occupation and Investment Holdings
I support introducing a large carbon tax to mitigate climate change risk. Economists recognize that such a policy will introduce price effects (substitution effects) that will help mitigate the climate change challenge. We also recognize that this policy will introduce income effects that will lower the real incomes of several different sets of people in our economy. This blog post will discuss 3 sets of people who would be adversely affected by carbon pricing. The anticipation of such a negative income effect creates an incentive for those affected groups to oppose the carbon tax. I have documented these political economy effects in this 2013 paper and this 2015 paper.
The 3 Groups
Suburban and Rural Residents --- Imagine if you live in center city San Francisco. You do not drive much. You live in a small apartment. You use little air conditioning. Your carbon footprint is tiny. See my 2010 paper with Glaeser. If you moved to rural Kentucky, you would drive much more and live in a much larger home that would use more oil for winter heating and electricity for summer air conditioning. Given that local power is generated using coal, the carbon tax would have a much larger bite for you. My point is that one's initial location in the United States (both which metro area and where within the area you live) plays a key role in determining whether you bear the incidence of the new carbon tax.
Fossil Fuel Industry Workers -- -As the carbon tax grows larger, jobs in coal mining and oil refining will diminish. These people will be unemployed for a while. Labor economists have documented that there is duration dependence such that a spell of unemployment makes you less likely to easily transition to another job. Unfortunately, we have learned from James Heckman's work that job training for middle aged people is rarely effective in helping them transition to new careers. My work with Jonathan Eyer helps to explain why states that mine coal have sought to preserve these jobs with special policies.
Investors in Fossil Fuel Supporting Assets --- If you own a $800,000 house in Houston or if you own $100,000 worth of Exxon stock, the increase in the carbon tax will lower the value of these assets. Why? Houston's economy is partially based on fossil fuel exploration and extraction. So, is Exxon. A carbon tax induces substitution away from these activities and lowers the value of assets tied to these industries.
So, note that there are at least 3 broad constituencies who have pocket book reasons for opposing carbon pricing. A well meaning economist who seeks to devise a "cap and dividend" program must figure out how large a $ check to write to each American household. The challenge here is that there are 330 million Americans and each differs along the 3 factors I listed above. It would take quite a sophisticated model for the government to figure out for each person --- where does Matt Kahn live? What industry does he work in? What assets does he currently own? With the IRS micro data, a research team could know #1 and #2 above but not #3.
If cap and dividend cannot be implemented, then opposition to carbon pricing is likely to continue.
A good mechanism design scholar should figure out a mechanism for each American to truthfully reveal her "carbon position". That would be a good research project!
It is important to note that the carbon tax raises a set of difficult counter-factuals; how much will Houston home prices decline by if there is a significant carbon tax? How much will coal miner unemployment increase if there is such a tax? Will displaced coal miners find an equally good job 8 months later? What transition costs will be imposed on them? Without knowing the answers to these questions, how does one devise "the right" dividend for such affected individuals? Without such a well structured (and credible) dividend, these individuals will oppose the Pigouvian policy.
Finally, you should note that I have implicitly assumed that people are income maximizers. In reality, utility maximizers will differ along another dimension. Some are more concerned about the impact of climate change than others. For those who worry about climate climate negatively hurting our world, they require a smaller dividend check in return for their supporting carbon pricing. The revelation mechanism would also have to yield this key unobservable to help the carbon tax designer figure out the efficient allocation of the carbon revenue that is recycled back to the people.
-
The NY Times reports that iPhone sales are way down in China. The recent 40% decline in Apple's stock price suggests that investors were unaware of the emerging trouble that Apple now faces. How would an industrial organization economist explain these recent facts.
A traditional explanation is that iPhones are expensive and are a luxury good in China. As upper middle class households are feeling the pinch of a weakening economy, it makes sense that some would cut back on luxuries and would purchase cheaper cell phones than the iPhone. So, a traditional explanation would focus on income effects and the fact that people purchase fewer luxury goods when they feel poorer. The iPhone is a durable product and the Coase Conjecture on durable monopoly is also relevant here. Those Chinese middle class and richer people who love the iPhone have already purchased one. The "marginal buyer" now consists of two sets of people. Those who already own one and are thinking of upgrading and those who have never owned one. By definition, the latter group do not have the same "love" for iPhones or they would have already bought it. There may be some poorer people who love the iPhone but can't afford it. As the economy grows worse, they are even less likely to be able to afford one. In the first group here, the Coase Conjecture matters. As it tries to sell new iPhones, Apple is competing with itself. The new iPhones must be much better in quality per $ spent for existing customers to upgrade and purchase the new iPhone.
A non-traditional explanation --- When I travel to China, I see that the iPhone is a status good. Similar to the BMW or the Mercedes, people use their iPhone in public to signal that they are "high status". Veblen understood this. Given that Apple is an American company and President Xi is focused on Chinese nationalism, I can imagine that the "good status" associated with American products in China is declining. People in China may now want to conform and signal to others that they are "loyal Chinese". This endogeneity of status (i.e that one day an iPhone conveys status and soon later it does not), could be formally modeled using the Akerlof and Kranton economic model of identity.
If this second model of behavior is the correct one, then Apple will need to form a partnership with a Chinese company and make future phones and sell them through that company. Apple will not have to take this step if President Xi ends his "boycott America" phase.
These two different models could be tested if Apple starts to randomly lower the price of the iPhone in China. If sales do not increase sharply, then this favors the "non-traditional" explanation. I would also want to see micro data on whether current owners of iPhones now switch to a Chinese phone even as the price of the new iPhones decline. This would be strong evidence of a lack of brand loyalty and President Xi's nationalism would be the prime explanation.
-
The NY Times has published a very interesting piece about Facebook's effort to prevent suicide. Facebook must have coded up an algorithm for predicting a person's probability of committing suicide as a function of the content of one's posts to the platform. A statistician might ask if the R2 in predicting actual suicide based on "public thoughts" posted to Facebook is .75 or .0002. I will return to this point in a moment.
As discussed in the news article, Facebook is using its predictions of suicide propensities to alert local police about the impending possibility of this horrible event.
Statisticians and economists would hope that the NY Times would discuss Type I versus Type II errors here and the asymmetric loss function. Permit me to explain.
Ideally, Facebook's algorithm would perfectly predict who is at risk of suicide and who isn't. This would be a statistical model with an R2 =1. Such a model does not exist. In reality, Facebook will incorrectly classify some non-suicidal people as suicidal (Type 1 errors) and some suicidal people as non-suicidal (Type 2 errors). From society's perspective, we have an asymmetric loss function. It is worse to make Type 2 errors than Type 1 errors.
The NY Times article focuses on Type 1 errors and confrontations between the police (who have been alerted by Facebook about the impending risk) and the unsuspecting non-suicidal person who now faces the police. The Times is concerned that violence could occur as an unintended consequence of this intervention. Libertarians will be outraged that people are being sent to hospitals for evaluation based on the algorithm's predictions.
While I do not have access to Facebook's suicide prediction algorithm, I believe that the key parameter here is the goodness of fit of its statistical model. The worse is the R2 in its core regression then Facebook's effort is not serving society's goals. People are believing in the hype of AI and Machine Learning. If the actual output that these models produce does not actually predict low probability events, then people will not be classified correctly.
Facebook should have to present to academics its current statistical model as this will yield greater accountability and honesty about the quality of its forecasts.
To summarize, Facebook has a prediction model that it uses to make person specific predictions. It then alerts the local police to intervene hoping that it has correctly classified individuals and that the police can arrive in time to take action. If Facebook does not classify people correctly or if it triggers salient thoughts about suicide as the police appear at your door, then Facebook's efforts can backfire.
This raises the deep issue of how statistical models affect our life and the feedback model of how predictions affect the future by impacting our decisions.
