Which professions are paid too much given their value to society?

Many jobs have spillover effects on the rest of society. For instance, the value of new treatments discovered by biomedical researchers is far greater than what they or their employers get paid, so they have positive spillovers. Other jobs have negative spillovers, such as those that generate pollution.

A forthcoming paper, by economists at UPenn and Yale,1 reports a survey of the economic literature on these spillover benefits for the 11 highest-earning professions.

There’s very little literature, so all these estimates are very, very uncertain, and should be not be taken literally. But it’s interesting reading – it represents a survey of what economists think they know about this topic, and it’s surprisingly little.

Here are the bottom lines – see more detail on the estimates below. (Note that we already discussed an older version of this paper, but the estimates have been updated since then.)

JobMean externalities per jobPrimary sourceMethod
Research+$950,440Murphy and Topel (2006)Willingness-to-pay for longevity gains from medical research
Teaching+$130,706Card (1999)Returns to education in excess of teacher salaries
Engineering & programming+$18,720Murphy et al. (1991)Cross-country regression of GDP on engineers per capita
Operations & consulting$0Bloom et al. (2013)Randomized experiment measuring effect of consultants on plant productivity
Law-$31,200Murphy et al. (1991)Cross-country regression of GDP on lawyers per capita
Management-$64,800Gabaix and Landier (2008)Calibrated model indicating CEO pay captures managerial skill and firm characteristics
Finance-$104,000French (2008)Aggregate fees for active vs. passive investing

We calculated mean income for 2005 in an earlier article. We increased income by 30% to account for nominal wage growth since then.

The paper uses the expressions spillover and ‘externality’. An ‘externality’ is a technical term for a ‘cost or benefit that affects a party who did not choose to incur that cost or benefit.’ The authors of the paper call it an ‘externality’ when someone who buys a service does (or does not) benefit after taking account of the cost of purchasing it. This is a nonstandard usage, but fine for our purpose of assessing the overall social value of different professions.

The externalities per dollar were calculated from table 3 in the paper. This gives the externality share as a fraction of income. We divided this by the total income share for the profession to get the externalities per dollar of income. We then multiplied this by the mean income. The working is in the table below.

For research, the authors made three estimates of the externality share, and for management they made two (as explained in the quoted section of the two). We took the averages of these estimates, which is why the values we report are different from those that appear in table 3 in the paper.

Externality as share of economy income (estimate 1)Mean income (estimated, 2016)Income shareExternalities per dollar of incomeMean externalities per job
Research9.7%$107,7821.1%8.8+$950,440
Teaching6.9%$66,3003.5%2.0+$130,706
Engineering0.6%$127,9204.1%0.1+$18,720
Operations0$96,2003.7%0.0$0
Law-0.2%$343,2002.2%-0.1-$31,200
Management-4.05%$214,40013.4%-0.3-$64,800
Finance-1.50%$318,9334.6%-0.3-$104,000

How did the economists estimate the size of the externalities?

Here is more explanation of how the externality estimates were made in the paper:

To calculate the marginal social output from each profession, we draw on the literatures that estimate economy-wide externalities from various professions. Although we have done our best to faithfully represent the current literature, we emphasize that these estimates are highly uncertain extrapolations from heterogeneous and not easily comparable studies primarily aimed at different estimands than those we draw from them.

Our prior is that Coasian bargaining should eliminate externalities, so when these literatures do not offer a clear finding, we set the aggregate externality to 0. In the cases in which these literatures offer conflicting results, we adopt one value as a baseline and use an alternate value for sensitivity analysis. (Note: In the tables above, we just took the average.)

We ultimately care about marginal externalities rather than the average over the whole profession, but the authors didn’t try to estimate the difference between the two. One industry that might be good on average, but harmful if it becomes larger is finance.

Here is more info on the estimates for specific jobs (the papers used are listed at the end):

Arts Although some evidence, and a number of good theoretical arguments, suggest the arts generate some positive externalities, we are unable to find a plausible basis for estimating the magnitude of these externalities, and consequently assume 0 to be conservative.

Engineering The only study we found of externalities from engineering is a cross-country ordinary least-squares regression by Murphy et al. (1991). They investigate the impact of the allocation of talent on GDP growth rates rather than on GDP levels. To be conservative and fit within our static framework, we interpret these impacts as one-time effects on the level of output rather than impacts on growth rates. We multiply their estimate of the GDP impact of an increase in the fraction of students studying engineering by the number of students studying engineering according to the OECD to obtain an externality of 0.6% of total income.

Finance French (2008) estimates the cost of resources expended to “beat the market” by subtracting passive management fees from active management fees. Bai et al. (Forthcoming) show the informativeness of stock and bond prices (measured in their ability to predict earnings) has stayed constant since 1960, despite a vast growth of the finance profession documented by Philippon (2010). We therefore interpret the entirety of French (2008)’s estimates, which amount to 1.5% of total income in 2005, as negative externalities from finance.

Law Murphy et al. (1991) estimate externalities from law in the same manner they calculate externalities from engineering, and we apply the same methodology to yield a −0.2% externality as a percent of total income. Kaplow and Shavell (1992) present several models of why the provision of legal advice may exceed the social optimum.

Management Two strands in the literature offer competing views on the externalities of management. According to the first strand (Bertrand and Mullainathan, 2001; Malmendier and Tate, 2009), chief executive officer (CEO) compensation shifts resources from shareholders to managers in ways that do not actually reflect the CEO’s marginal product. Piketty et al. (2014) argue that 60% of the CEO earnings elasticity with respect to taxes represents this rent-seeking behavior, implying the negative externalities from management are 8.1% of total income. The other half of the literature argues market forces can explain CEO compensation (Gabaix and Landier, 2008) and suggests that therefore externalities are 0. Most managers in our sample work at lower levels of firms where the problems of measuring marginal product highlighted by the critics of CEO compensation are less likely to apply, so we take the figure of 0 as our baseline and use the −8.1% figure in sensitivity analysis.

Medicine We could find no literature estimating the externalities of (non-research) medicine and so set the externality to 0 to be conservative.

Operations This profession is comprised of consultants and IT professionals. Bloom et al. (2013) conducted a field experiment to determine the causal impact of management consulting on profits. They interpreted their results as consistent with the view that consultants earn approximately their marginal product, and thus we assume no externality for consulting.

Real Estate We could find no literature estimating the externalities of brokers, property managers, and appraisers and so set the externality to 0 to be conservative.

Research Our baseline estimate for the externalities from research comes from the value of medical research, measured in terms of people’s willingness to pay for the additional longevity this research makes possible. Murphy and Topel (2006) estimate the annual gains of medical research equaled 25.72% of GDP from 1980-2000. Traditional GDP accounting does not capture this externality, in contrast to our model, so we divide it by GDP augmented with this externality to obtain .2572 1+.2572 = 20.5%. Although this externality may be the largest externality from academia and science, this estimate is still conservative in assuming no gains accrue from other research fields. An alternative measure of research externalities comes from the literature that calculates the social returns to R&D. Jones and Williams (1998) suggest the socially optimal amount of R&D activity is four times the observed amount, which we loosely translate into a three times externality or 5.6% of GDP. A narrower benchmark for this externality focuses only on the externalities of universities to profits made by geographically proximate firms as studied in Jaffe (1989). We use his estimates to calculate a much smaller 3.0% externality, which we use as a lower-bound estimate in our sensitivity analysis.

Sales Although an extensive theoretical literature argues the welfare effects of advertising can be positive or negative (Bagwell, 2007), we are not aware of any work attempting a comprehensive estimate of externalities, and therefore, as with medicine, we use an externality of 0.

Teaching We calculate the social product of teaching as the impact of an additional year of schooling on aggregate earnings of all workers in the economy. The spillover from teaching is then this social product less the annual earnings of all teachers. As our estimate of the effect of a year of schooling on earnings, we use a 10.3% gain, which equals the midpoint of the numbers collected in Card (1999)’s review. Because teachers earn 3.4% of economy income, we use a spillover from teaching of 6.9% of economy income. We also compute the aggregate effect of teaching on earnings using Chetty et al. (2014)’s measure of teacher quality and its long-run impact on eventual student earnings. We use the ratio of total teacher pay to its standard deviation in our data multiplied by the social product Chetty et al. (2014) estimate for a standard deviation in teacher quality to obtain an aggregate effect equal to 10.2% of economy income. This figure leads to a spillover of 6.8% of economy income. Given the similarity between the two spillover estimates and the fact that the estimate based on returns to schooling is more easily interpretable in the aggregate, we use the Card (1999) number as our estimate.

How many extra externalities will result if you take a job in the industry?

The estimates given are supposed to apply to additional ‘marginal workers considering joining an industry. However, if you join an industry, the additional externalities will probably be less.

First, diminishing returns have been ignored, and, in established industries like these, diminishing returns will probably mean that additional workers will have less impact – positive or negative – than the average.

Second, by taking a job in the industry, you won’t lead to a whole extra person’s salary worth of income earned in the industry. We could model you taking the job as an increase in the labour supply for that industry by one. This will slightly decrease wages in the industry, and won’t lead to a whole extra worker employed (the amounts will depend on the supply and demand elasticities in the industry). This means the income earned by the workers in the industry will increase by less than a whole salary. If the externality-income ratio stays the same, then by working in the industry, you change externalities by less than what’s in the table. (Though note the externality-income ratio could easily change – it seems more natural to tie it to the revenue of the industry than worker income.)

On the other hand, if your salary would be higher due to unusually good personal fit, then you’d need to increase the estimates.

To do a full analysis, there are other adjustments we’d want to make as well (see our upcoming article on replaceability).

What conclusions can we draw from these estimates? Donations are more important

The main point that strikes me is that, with the exception of research, all the estimates are fairly small relative to salary. And, that’s even if we ignore the points in the section right above.

This suggests that at least for “normal jobs”, your donations to charity are a more significant component of your social impact.

For instance, if we suppose making a random American $1 wealthier is worth one “unit” of impact, then I think donations to GiveDirectly are at least 20 times better, mainly because the recipients are about 100 times poorer than the average American. And I think Against Malaria Foundation (AMF) is about 4 times more effective than GiveDirectly, so it creates 80 units of social value per dollar of donations.

This means that, if someone earning $100,000 per year donates 10% of their income to AMF, they create $0.8m of social value per year, which is about as large as the largest externality estimates in the table.

Moreover, I think there are charities that are 10 times more effective than AMF, and if you earn $100,000, it would be possible to give 30% rather than 10%. Putting the two effects together, it’s plausible your donations could be another 30 times more valuable.

I also think the externality estimates are more likely too large than too small. My prior is that the externalities are close to zero, and the authors seem to have updated too aggressively based on speculative calculations. Moreover, I don’t think “dollars of wealth created” is a very good proxy for humanity having a good long-run future, which is the main thing I care about. I think it’s more important to focus on issues around technology and existential risk, and most jobs don’t have much impact on that.

This would mean that how much and where you donate is normally far more important than the “direct” impact of your job, which is an argument for earning to give. The exception would be if you work at a very high-impact organisation, such as AMF, you’re a good fit with research, you work in government or politics, or take other unusually important positions. Advocacy could also be a path that’s competitive with earning to give in terms of impact. There could also be exceptions if you take an unusually harmful career.

That aside, it seems true that most people could have more impact by earning to give. But, that doesn’t mean most members of our community should earn to give, because they might be able to take many of the even better jobs just listed.

What about the estimates for specific jobs? Unfortunately, we can’t draw many conclusions about these. The estimates are based on rough calculations in just one or two papers. I’d prefer to see a more thorough process where (i) we start with a prior estimate near zero, (ii) we consider theoretical arguments as well as empirical estimates and (iii) we update away from the prior depending on how robust the evidence is.

If this were done, my guess is that the ordering would be similar, though art and medicine would also have positive externalities, marketing might be a small negative, and the size of the externalities would generally be smaller (perhaps with the exception of research).

Read next: Which jobs are highest-impact?

Which are the 10 most harmful jobs?

Get our latest research in your inbox once a month

Sources

Murphy, Kevin M., Shleifer, Andrei and Vishny, Robert W. (1991). ‘The Allocation of Talent: Implications for Growth’, Quarterly Journal of Economics 106(2), 503–530.
French, Kenneth R. (2008). ‘The Cost of Active Investing’, The Journal of Finance 63(4).
Philippon, Thomas (2010). ‘Financiers vs. Engineers: Should the Financial Sector be Taxed or Subsidized?’, American Economic Journal: Macroeconomics 2(3), 158-182
Kaplow, Louis and Shavell, Steven (1992). ‘Private versus Socially Optimal Provision of Ex Ante Legal Advice’, Journal of Law, Economics and Organization 8(2), 306-320.
Bertrand, Marianne and Mullainathan, Sendhil (2001). ‘Are CEOs Rewarded for Luck? The Ones Without Principals Are’, Quarterly Journal of Economics 116(3), 901-932.
Malmendier, Ulrike and Tate, Geoffrey (2009). ‘Superstar CEOs’, Quarterly Journal of Economics 124(4): 1593-1638.
Piketty, Thomas, Saez, Emmanuel and Stantcheva, Stefanie (2014). ‘Optimal Taxation of Top Labor Incomes: A Tale of Three Elasticities.’ American Economic Journal: Economic Policy, 6(1), 230-71.
Gabaix, Xavier and Landier, Augustin (2008). ‘Why has CEO Pay Increased So Much?’, Quarterly Journal of Economics 123 (1), 49-100.
Bloom, Nicholas, Eifert, Benn, Mahajan, Aprajit, McKenzie, David and Roberts, John (2013). ‘Does Management Matter? Evidence from India’, Quarterly Journal of Economics 128(1), 1-51.
Murphy, Kevin M. and Topel, Robert H. (2006). ‘The Value of Health and Longevity’, Journal of Political Economy 114(5), 871-904
Jones, Charles I. and Williams, John C. (1998). ‘Measuring the Social Return to R&D’, Quarterly Journal of Economics 113(4), 1119-1135.
Jaffe, Adam (1989). ‘Real Effects of Academic Research’, American Economic Review 79(5), 957-970.
Bagwell, K. (2007). ‘The economic analysis of advertising’, Handbook of Industrial Organization vol 3 1701-1844.
Card, D. (1999). ‘The Causal Effect of Education on Earnings’, Handbook of Labor Economics vol 3 1802-1859.
Chetty, Raj, Friedman, John N. and Rockoff, Jonah E. (2014). ‘Measuring the Impacts of Teachers I: Evaluating Bias in Teacher Value-Added Estimates’ 104(9), 2593-2632.

Notes and references

  1. Lockwood, Benjamin B, Nathanson, Charles and Weyl, E. Glen, “Taxation and the Allocation of Talent” (April 25, 2016). Journal of Political Economy, Forthcoming. Available at SSRN: https://ssrn.com/abstract=1324424 or http://dx.doi.org/10.2139/ssrn.1324424.