Introduction and Summary
We recently did a case study with Ramit (see the full case study write up here). He was wondering whether to start a medicine degree, with the aim of going into research, or to continue in finance doing earning to give, where he already has a job as a quant researcher earning in the range of $150,000 per year.
We did an in-depth comparison of the expected impact of the two paths to help him decide. The rest of this post contains our case study research notes on the comparison.
If forced to guess now, we lean in favor of earning to give, though we think it’s very high priority to gain more information. Ramit is going to try to better assess his degree of fit with medical research, perhaps by working as a researcher during the summer, and learn more about his earnings prospects in finance by making applications and speaking to a headhunter. We’ll review our decision when we find out more.
In the rest of the post, we explain our reasoning:
- We outline our general approach
- We define a number of factors to compare the two options
- We evaluate the two paths based on the factors
- We make our overall conclusions
Note that there were several potentially important issues we didn’t address, including job satisfaction and which path is best for career capital in careers besides finance and research.
Our initial guess at the answer
We state this guess so that you can see in which direction we’re most likely to be biased. Our initial guess was that earning to give is better.
Both career paths will produce a stream of benefits over time. Ideally, we could create a quantitative measure of these benefits and directly compare the expected stream of benefits produced in research with the expected benefits produced from the donations.
Unfortunately, there’s very little relevant data available, so explicitly going through this analysis would be highly time consuming. Moreover, pushing to make a quantified model too early can easily cause you to overlook important considerations that are not easy to quantify, or spend too much time evaluating intractable considerations (for more reasons why an overly quantified approach can fail, see this post by GiveWell). There’s also reason to think it’s better to do head-to-head comparisons rather than full impact estimates.
So, rather than immediately create a quantified model, we’re going to start with a more qualitative analysis: we’ll attempt to list the most important factors and heuristics to apply to compare the options, evaluate which path wins for each, and make an overall gut judgement. We’re interested in extending this analysis into a quantitative model in the future.
The factors we used to compare the two paths
- What’s the size of Ramit’s contribution in donations and research labor in each path?
- What’s the effectiveness of biomedical research compared to the top donation opportunities?
- Advocacy potential: Which path provides a better public platform for encouraging other high impact activities?
- Flexibility: Which is the more flexible path?
- Discount rate: In which path do the benefits accrue soonest?
- Value of information: In which path will the effective altruism community learn the most useful results about what works?
How did we come up with the factors?
Broadly, we’re looking for the factors that have the best combination of being important in determining the answer, relatively easy to evaluate and reasonably uncorrelated.
As explained in our career framework, we often find it useful to think of impact as a product with the structure: contribution x cause effectiveness. Contribution is what the person adds to the cause. Cause effectiveness is a measure of how far these resources go to create impact. Both of these factors are highly important and somewhat tractable to evaluate.
The first three questions are directly taken from this framework. (1) compares relative size of his contribution as a donor vs. a researcher. (2) compares the effectiveness of the causes Ramit could support in earning to give vs. medical research. (3) accounts for some part of his potential to make large indirect contributions (e.g. from persuading other people to do earning to give, or advocating for reform within research). (3) is less tractable, but highly important.
Answers to these questions would suggest a ‘static’ answer to which path would be better. Unfortunately, the situation is more complicated because we expect our knowledge about which cause is best to change significantly over time, and the benefits to accrue much later in the medical research path.
If Ramit becomes a researcher, he will probably only start making important research contributions to his field in 10-20 years.1 In that time, our best guesses about which causes are most high priority will almost certainly have changed. The more flexible your career path, the easier it is to switch which cause you’re working on to take account of this new information. Therefore, we compare both paths in terms of flexibility.
Furthermore, it’s generally better to have impact now rather than later. That’s because (i) solving problems now creates compounding benefits that allows us to solve further problems (ii) the overall space of opportunities to do good can be expected to decline in the future due to the general march of progress. (See here for an overview of the crucial considerations for charitable donations – a situation which seems fairly analogous). So, we’ll compare the paths in terms of how soon the benefits accrue.
A final extra consideration is the value of information produced for the effective altruism community and Ramit by pursuing the two paths. By doing either, we’ll learn something about what works which may be applicable to other people aiming to make a large impact.
We considered but didn’t include whether the effective altruism community is going to be non-funding constrained in the future. You might think that the support of major foundations like Good Ventures will mean that the best donation opportunities will get taken over the next decade, making earning to give much less important compared to direct work. We didn’t include this factor, however, because (i) it’s seems like a very uncertain judgement call (ii) if that happens, Ramit will be able to quit finance and do direct work (he might end up in a slightly worse position than if he studied medicine, but it doesn’t seem like a strong consideration).
How happy are we with these factors, and what has been left out?
We think they capture important, tractable elements of the expected value comparison, though wouldn’t be surprised if they overlooked a further factor or if there was a more tractable way to make the comparison.
We’ve done about 5 detailed comparisons of earning to give to direct work before, though not with academic research vs. earning to give.
An initial assessment of the factors
What’s the size of Ramit’s contribution in donations and research labor in each path?
In either case, he’d have the potential to make substantial donations.
As a medical researcher, we estimate he could expect to earn at least $80k per year, and potentially up to $200k if he conducted his research as a doctor.2 Ramit intends to donate most of what he earns above $60k. So that’s $40-$140k of donations per year.
In finance, we’d estimate his expected earnings are in the $200-500k range.3 That implies donations of $140-$400k per year. In addition, in finance he’ll gain an additional 6 years of earnings (the time he’d spend studying medicine), which adds another ~$10k per year averaged over your career, increasing the donations to $150-$400 per year.
Overall, in finance he could donate an additional $100-350k per year, whereas in research he’d also be contributing directly. The question is, could we expect these extra donations to have more impact than the value of his labor invested in biomedical research?
One way to gain traction on that question is to ask whether medical researchers would prefer his labor or the extra money. Our initial guess was that researchers would prefer the labor of people with high potential as researchers, and that you might be such a person.
This was borne out when I asked several researchers (see here and here, and stay tuned to see two further unpublished interviews) whether they’d prefer a good researcher working for them or an annual grant of several hundred thousand. They said they’d prefer a good researcher working for them. One of the researchers also thought that Ramit was likely to be a good researcher. His reasoning was that he seems highly motivated, smart and would have the valuable combination of an MD and programming skills, and that someone like that can get their own funding relatively easily. (see our upcoming overview of biomedical careers for more).
If this is true, then Ramit’s labor is worth something like 2-10x the extra donations in support of medical research.
In reality, he wouldn’t have to donate to medical research, he could donate to something else, which brings us to…
How effective is biomedical research compared to the top donation opportunities as a cause?
We think biomedical research is likely to be a high priority cause, but we overall prefer prioritisation research and promoting effective altruism (see our upcoming list of best guess causes for more). Note that this is a controversial judgement call, and could easily change. To see our reasoning, see our overview of biomedical research (upcoming).
If Ramit was supporting these causes through donations, then we’d prefer donations by about a factor of 10 to informed support given to biomedical research (though note that this is a complete judgement call). If he was supporting a more normal charity through donations, then we think that’s not obviously much better than supporting biomedical research.
So, if Ramit supports the top causes with his extra donations (and indeed he’s willing to consider options like funding GiveWell), then we think they may be worth about 10 times support given to biomedical research. So let’s combine this with the previous result.
Finance: donate an extra $100-350k to causes with effectiveness of 10 -> overall worth 1000-3500k units
Research: produce labor for medical research valued at $500k (depending heavily on Ramit’s ability) with effectiveness of 1 -> overall worth 500k units
Given how uncertain we are of each factor, overall I’d say that if you would donate to the top causes and you have high potential to be a good researcher, the two options are close to equal, though I slightly prefer earning to give.
Gaining more information seems highly valuable. We recommend (i) shadow a medical researcher and do informational interviews (ii) try out being a research assistant (ii) apply to medical school (iv) speak to people about how much you might earn in finance (v) ask a headhunter how much you might earn (vi) apply to top finance jobs.
Advocacy potential: Which path provides a better public platform for encouraging high impact activities?
We’re highly uncertain about this factor. In favor of earning to give, he’ll probably meet more wealthy, influential people as a high earner and a donor. In favor of medical research, he’ll have academic credibility and more access to top scientists. Doing earning to give is also more unusual and impressive, which makes it easier to get attention and means that you provide an example of an unusual way to do good (whereas there’s lots of well motivated medical researchers).
Overall, we’d guess the two paths are about equal on this factor.
Flexibility: Which is the more flexible path?
Earning to give is clearly more flexible, because donations can be switched between organisations very easily. With research, you’re fairly confined to stay within medical research (although we’ve found that within medical research you can be fairly flexible over your career). That creates a danger that biomedical research becomes much less effective over time, and you’re less able to switch into something more effective than if you were doing earning to give.
How much is this extra flexibility worth? If he took the research path and decided later in his career to stop supporting biomedical research, he could drop out into medicine and start pursuing earning to give.
Doctors who can pursue medical specialities can earn around $300k p.a. averaged over their career.2 We asked one doctor whether starting your career in research and then becoming a consultant later would reduce your earnings. They thought probably not, and it may even increase your earnings since it could help you become a high level consultant. Another researcher pointed out that it was difficult to make the jump later in your career, so, overall we think we can assume relatively high earnings if he drops out of research, but caveat that the option could go away after age 40.
Supposing he can drop out into high paid medicine and donate $200k per year, then the finance path (in which he’s donating $100-350k) is only a little better, so this consideration only slightly favors finance.
Discount rate: where do the benefits accrue fastest?
Earning to give clearly wins, because:
- With the medical research path, he’ll lose 5 years spent at medical school, then an additional 5-10 years as a junior researcher before he starts making a substantial contribution (and another 20 or so years until your peak impact1)
- He’ll also incur substantial medical school costs. The median debt of medical school graduates in 2013 was $170k.3 This is roughly equivalent to another 2 years of donations.
His contribution in earning to give has a 12-18 year head-start compared to medical research. We suspect it’s generally better to contribute now rather than in 12-18 years time, because (i) contributions now have broad compounding benefits (ii) the general march of progress means that it’ll be harder to make a difference in the future (e.g. vaccine research will be less important, because we will have reduced the incidence of infectious diseases substantially through other means).
A 3% discount rate over the period, as used as a lower bound by the Copenhagen Consensus, makes earning to give about 1.5x better (roughly estimated as 1.03^15). Some experts in the effective altruism community have suggested discount rates of 10-20%, because they think that investing in meta-charity is an unusually good opportunity. A 10% discount rate, would make earning to give 4x better.
Value of information: In which path will the effective altruism community learn the most useful results about what works?
A significant number of effective altruist minded people are pursuing careers in finance, but we’ve only advised three people studying medicine, and none of those are definitely planning to enter medical research.
This suggests there could be significant value in Ramit entering medical research and writing about his experiences. On the other hand, in the 5 years it would take him to study medicine, we think it’s highly likely we’ll find other effective altruist minded people who are already in medical research, so overall don’t think this is a large benefit.
Summary of factor evaluation
|Factor||Which path wins?||What’s this based on?||Key uncertainties|
|Relative value of research labour compared to extra donations to top opportunities||Slightly in favour of earning to give||Several controversial judgement calls, expert opinion about value of labour and earnings estimates||1. Ramit’s potential to be a good researcher 2. The relative effectiveness of biomedical research compared to top donation opportunities|
|Advocacy potential||Unclear||Controversial judgement call||1. Could he persuade more people to do earning to give in finance? 2. Are there good advocacy opportunities mainly open to medical researchers?|
|Flexibility||Moderately favours earning to give||Earnings differential between finance and medicine||His expected earnings in finance and medicine|
|Discount rate||Strongly favours earning to give||Widely agreed upon facts about how long it takes to train as a doctor and expert opinion on the appropriateness of discounting||More detailed modelling of the flow of benefits over time|
|Value of information||Weakly favours research||Judgement call||Prospective growth of the effective altruism community|
There’s also ‘model uncertainty’ arising from the chance that we’ve forgotten an important factor.
If asked to guess now, we’d rate earning to give in finance slightly higher than medical research if Ramit donates to top causes. The main reasons are: (i) benefits accrue much earlier (ii) he’ll be able to support more effective causes now and to a greater extent in the long-term.
However, we’re highly uncertain of this assessment, so our top recommendation is to gain more information. In particular, much hangs on Ramit’s prospective ability as a medical researcher. If he’s not likely to be highly motivated as a researcher, then the case for earning to give in finance is much stronger. If he would be a good fit, then we’d recommend finding out more about your earnings prospects and doing more detailed modelling of the comparison.
Suggestions to Ramit for future work
Try to work out whether you’d be highly motivated in biomedical research e.g. become a research assistant over the summer, do shadowing and talk to researchers.
Interview several people about your earnings prospects in finance and medical research. In particular, to what extent can you do important research while being a highly paid consultant?
Wait for more in-depth analysis of biomedical research as a cause from GiveWell and CEA.
Find out more about whether effective altruist minded people have opportunities for outsized impact in medical research as advocates or by being more effectiveness minded.
Create a more detailed model of the discounted value of your donations and labor contribution in research.
Notes and References
Generally, academics in medicine only reach the peak of their career in their 40s or 50s, 20-30 years after they start. It seems generally agreed that researchers don’t make major contributions in their first 5 years of research after graduating.
“At the contrary extreme, the typical trends in other endeavors may
display a leisurely rise to a comparatively late peak, in the late
40s or even 50s chronologically, with a minimal if not largely
absent drop-off afterward. This more elongated curve holds for
such domains as novel writing, history, philosophy, medicine,
and general scholarship”
Age and Outstanding Achievement: What Do We Know After a Century of Research?
Dean Keith Simonton, Psychological Bulletin, 1988, Vol. 104, No. 2, 251-267
http://resources.emartin.net/blog/docs/AgeAchievement.pdf ↩ ↩
According to BLS Occupational Handbook, medical scientists earn a median of $80,000 per year.
For doctors, the United States Bureau of Labor Statistics ([U.S. Bureau of Labor Statistics](data: http://www.bls.gov/oco/ocos074.htm)) reports mean annual wages of $168,000 for family physicians in the U.S., more for higher-paying private practice (vs employment) and regions, $220,000 for surgeons. Other surveys report somewhat higher figures. Part of the difference reflects the distinction between wages and profit sharing/bonuses/other compensation forms.
According to the Medical Group Management Association’s Physician Compensation and Production Survey, median total compensation for physicians varied by their type of practice. In 2008, physicians practicing primary care had total median annual compensation of $186,044, and physicians practicing in medical specialties earned total median annual compensation of $339,738. ↩ ↩
The lower range comes from his current salary of $150k, which he thinks he could easily beat. The upper range comes from our experiences talking to people in quant trading jobs. We think Ramit would have a reasonable chance of succeeding in these jobs, and we know that the salaries start in the $100-200k range, then rise to over $500k in a couple of years. This estimate doesn’t represent our fully considered view on the matter, but is rather a quick judgement call based on what we know already. ↩ ↩