Career Update Aug 2020

Feedback from STM

In the last essay I had looked at the impact of Data Science (DS) career vs an EA career. I hadn’t looked at the lifetime calculations which seems to be a better indicator. For example, 8 years of grind from 30 and then ending up with a high-pay job, might still be better than taking a less risky but lower-paying job.

Also, at the end of the essay I concluded that I was going to switch to DS in the coming 6 months and then get back to focusing on EA. This I did without really looking at what sort of impact it had on the lifetime impact.

With this essay, I hope to address the above issues.

Introduction

Many people I know in FAANG are making 300k TC after 4-6 years of experience, be it in antenna design or Data Science or CS. Having seen examples of Jeff Kaufman being able to donate 50% (150k), there seems to be “great” potential with a FAANG job.

On the other hand EA also claims to have lots of impact. With Milan Griffes’ estimate here, on how just as an entry level researcher he is able to make contributions to the extent of 93k dollars, it seems to hold a lot of promise.

I am thus intrigued by both these paths as options for a “high-impact-career” and would like to compare these two over a life time and particularly for my scenario.

Note: There are 2 ways many people I know take, to get to the US. People take the masters route, and then work towards an H1B visa which allows you to work there. The other route is the L1 route. In this case you work in say Amazon in India (for e.g.,) and then try to move to the US on an L1.

With this essay we look at the following scenarios:

1. Start with a DS job, find a masters in the US and somehow get into FAANG and donate X% ever year. This also includes the fall back option of coming back to NL for whatever reason (recession, getting fired etc.) and continuing DS in NL for the rest of my life.

2. Start with a DS job, and instead of taking the masters route to the US, I will try the L1 route. The rest is the same as in 1.

3. Try, in the coming 4-5 years to become a GiveWell Researcher while the fall back option being to pursue DS with L1 and if that fails, back to Netherlands in DS.

4. Try in the coming 4-5 years to become a Charity Entrepreneur, and the fall back option being to pursue DS with L1 and if that fails back to Netherlands in DS.

Why GiveWell?

GiveWell is a “highly respected” organization among the EA community. Their impact is very clear with their cost-effectiveness estimates and their money moved to other organizations. This allows me to compare impact with other scenarios. If for example I took MIRI or Open Phil, then I really don’t have such estimates to make comparisons.

GiveWell can in-theory be substituted with anything else that holds similar or larger impact.

4-5 years is arbitrary

From here I get some estimates of how long people spent trying to move into EA be it in research or entrepreneurship. It is in the order of 2-5 years. If I don’t get in, in that time then I assume I should stop trying.

Why DS?

I am going to focus on DS instead of CS, for DS involves “critical thinking”, working with statistics, working with claims, hypothesis testing etc.–things that seem to help become a better EA researcher. Also Earning potential wise, it appears there is not “much difference” (looking at the average salaries of 137k$). So we stick with DS.

Why US?

Earning potential is quite high in the US. Earnings are substantially larger than here in Netherlands (or Europe). I know DS people with 4yoe in Apple California earning 300k$ TC and others in Walmart with 300k$ TC. While here (Netherlands) people seem to be earning 78k-103k$ as per Glassdoor. With salaries as in the US there are people like (Jeff Kaufman) who are able to donate 50%. This amounts to about 150k$ per year with a 300k$ TC.

Why Research, why not operations?

I think I am capable of doing research more than operations. Also, the journey to an EA job in research seems like it can be done from home behind a computer. For e.g., there are people who got a position in EA orgs by writing at the EA forum, critically evaluating other orgs’ work, and working on themselves on the side.

However, the path doesn’t seem to be direct for operations work. If you look at Tanya’s profile you see that she has some different types of management experience for a few years and then turns to FHI for a position. I would need to take some chances and try some management positions and try to be best at it and maybe that pays off. Not sure what “high-impact-option” to take up if the operations route fails.

Being a researcher doesn’t mean you can’t be a co-founder one day or grow into management positions in EA orgs. Research is only the start. It appears that researchers are capable of growing to MD level (higher impact than researcher) or even co-founder level positions.

In the coming sections we look at the lifetime calculations of the different scenarios.

Scenario 1: DS with Masters in the US

The above decision tree shows the different parts expected on the “DS-masters” route. I would start with a DS job in Netherlands, find my way into a university, pay my University related expenses, fight with the odds of getting fired, get an H1B and then hopefully a FAANG job, working till 65 without getting fired.

As a backup, in case I fail, I can always go back to Netherlands to pick myself up and continue DS there, albeit the smaller donation capability (shown later).

The Probabilities

The following table shows the different crossroads and probabilities (Success/Other) associated with them. And the last line repeats itself until I am 65 when I would need to retire.

The chance of Getting a DS job in NL in the next 1 year seems “high” to me. I have seen a couple of people who have moved to DS from TU Delft where they did their master’s in engineering in “sustainable energy” and moved to DS. A more relatable story is of this girl, who did bachelor’s in Mechanical (similar to mine) and then moved to a master’s in TU Delft and moved into DS within a few months of graduating. She seems to have done a Coursera course, R and some python assignments in the mean time.

Also, everyone works in DS, right from computer scientists to people who did their bachelor’s in Psychology and political science. This guy with a PhD in Electrical Engineering spent 8 months at a DS school in his weekends to transfer to DS in the US. As a result I think there is a high chance to Get a DS job in NL. I have placed this probability at 95%.

Finishing master’s and getting job (recession or not)

Everyone I know has managed to get into a US university of their capability in a time that is not recession and ended up staying in the US with a job. I would guess that this number is around 20-40 people. Three people from that lot did come back to India but that was because of not making the H1B Lottery (which is accounted for separately). Thus, I place the probability of finishing master’s and getting a job @ 70-80%. If its a recession like now, then the chances of getting a job after finishing master’s is assumed to be 0% which might be a “tad conservative”.

The world has had about 8 recessions in a span of 70 years. Chance of recession in a given year seems to thus be 8/70=11.4%. So taking this into account, the chance of “Finishing master’s and getting a job in the US (recession or not)”, is (1-11.4%)x(70% to 80%)+11.4%x0%=62% to 71%.

Paying back tuition fees:
Doing a master’s costs money in the form of tuition and living expenses while studying. As soon as one starts his education the debt opens. I don’t expect any scholarship. Hence, this is placed at 100%.

Not moving back due to recession or general firing

In the US it is quite possible that you loose your job in the time of recession or get fired from your regular job with little or no notice. This means that you have to scavenge for a job within the next 60 days and if you don’t end up finding a job you are kicked out of the country. This has bad implications for life time impact as this would mean going back to NL and working from there, where salaries are not comparable to that of the US.

A proxy for the “chance of getting canned during recession” seems to be in this website which tracks the startups that are laying off people. I assume the median % of layoffs in these companies as my “chance of getting canned during recession” in any company. This seems conservative, in the sense that not all companies are going to layoff people during recession (e.g., FAANG in the current pandemic). This is 22% according to layoffs.fyi. During recession I assume it is not possible to get a job within the 60 day period.

The “chance of moving back to Netherlands, as a result of loosing my job in non-recession time” seems to be “really low”. Of the people I know who are living in the US and working in tech, some of them have gotten fired, and all of them have bounced back within the 60 day period. I thus put the chance of moving back during non-recession time as 2% for a given year. (US unemployment rates during start of 2020 was <4%).

Thus the total chance of NOT Moving back to NL due to recession or some general firing is (1-11.4%)x(1-2%)+11.4%x(1-22%)=96%.

Getting an H1B Visa: The next hurdle to be crossed is that of Getting an H1B visa through the lottery process. There seems to be a 58% chance that a person with a switching background (mechanical to DS) will clear this hurdle ¹. This is including atleast 3 tries at H1B (i.e., 1-3 years of applying for the Lottery). The detailed reasoning and citations are in the footnotes ¹.

With this, all the probabilities used are covered for the DS Master’s path.

Effective Value

With calculations for 3 different types of impacts, and the above probabilities it is possible to complete the decision tree and end up with an effective value as shown here in Google sheets. The 3 different impacts are:

1. Failure impact: Lifetime impact calculations from the time of failure (@ above “crossroads”) until retirement.

2. Success impact: Lifetime impact calculations if we face success (@ above “crossroads”) all the time.

3. Additional significant costs that come into the picture such as tuition etc…

Failure Impact

The failure option for any point in the decision tree would be to go back to NL and pursue DS (“DS NL” in decision tree). This would be from the time of leaving US until retirement (and donate what is possible). The following variables are used to compute the DS NL impact.

NL is pretty stable. While US unemployment rates have gone up to 14% from 3.5% at the beginning of the year, in NL those rates have at max reached 4.5% during recession from 3.5% at the beginning of the year. It is thus assumed that I wont get fired till I retire, and will have a constant but slow growth rate (3% as observed where I work) in Netherlands.

Regarding donations, I think 15% is hard to do in the first ten years. I am currently able to do 10%. This leaves me with just over 15% for saving and vacations (36% taxes, 36% food and rent etc., 10% donation). Every Euro I earn extra from now on, is taxed at 50% along. Keeping expenses same and increasing savings and vacation expenses by 5%–between now and 10 years later–I estimate that I can increase the donation to 15% only after 10 years.

So in the first 10 years sticking to 10% and in the last 10 years going as far as 20% might be an option. I think 15% on average over lifetime should be doable with a bit of effort. Further optimization is possible but we move on.

In case I don’t get a US university for some reason, I am stuck in Netherlands as a result, then expected life time donation is around 560k$ based on years 32 to 65.

Success Impact

The following are the variables chosen to calculate the life time impact associated with working in DS in California:

I assume the average growth rate is 6% on average (based on 2016 to 2019 growth of Jeff Kaufman which is after ~8 years of experience).

In the first years I doubt I can do more than 10%. In the last 10 years of my career I guess I can do way more than 30%. I expect to make 300k TC 8-10 years after I start the job. I will have about 100k per month to spend in-hand after a 30% donation (based on this net-salary calculator). This seems to be quite a lot considering people in EA orgs are able to live on <100k gross salary in California. Maybe I can push it to 40%, like a higher limit. If I end up with success at all crossroads, then it is expected to have 4m$ to 5m$ in expected donations by 65, based on average donation of 30%-40%.

Additional Costs

Costs of program in a public university and the cost of living during that time is expected to be 70k for 2 years for a university like California State University. This I gather from someone who is doing their master’s there. I use this for now.

Lifetime Impact

Filling in the above values into the Google sheets shown here, estimates the Lifetime impact of the DS Master’s path. It is expected to vary from 713k$ to 865k$; with my “best-guess-estimate” at 731k$. The sheets value only shows the “best-guess-estimate”.

Scenario 2: DS with L1

The above decision tree shows the different parts expected on the “DS-L1” route. I would start with finding a company that would allow me to transfer to US in due course on an L1 visa. I allow myself about 4 years to mastermind this. Once I go to the US I would need to focus on plans to get to an H1B as the L1 is tied to the company and lasts for only 5 years. Once I am in the US I hope to also do an online master’s which would improve the chances of my H1B lottery (more on this later).

As a backup, in case I fail, I can always go back to the Netherlands to pick myself up and continue DS there, albeit the smaller donation capability.

The Probabilities

The following is part of the decision tree for what I expect to happen if I choose the “DS L1 route”.

In this path I need to first find an L1-able company. A company that will allow me to move to the US on an L1-visa. There are examples I have come across, across fields and countries. There are people who went all the way to JP Morgan in New York on an L1 after a few years in India and SWe engineers from India who are now in FAANG. Atleast for mechanical background, people from ASML Netherlands regularly work in the US for a few years on L1. It seems “accessible”. I am really unsure how to approach such a thing for DS and from Netherlands. I have to look into the possibilities more for DS people in Netherlands. Say 50% chance–much lesser than the chance of “finding a job in the US after a master’s” (80%). This is modeled as 2 chances for finding an L1-able company (as seen in the decision tree).

Even if I find an L1-able company the USCIS needs to approve it. This approval rate declined to 72% in fiscal 2019. This is used in the decision tree.

It appears that you increase your chances of getting an H1B from 42% to 58%¹ (over 3 attempts) if you have a “US accredited Master’s”. Even in the L1 path I would like to pursue an online master’s atleast to increase chances of staying in the US. Hence the Paying back tuition debt “crossroad”.

With an L1, there seems to be some push back from companies trying to take you in through an H1B. A friend while applying to companies while he was in L1 informs me that companies didn’t want him because he was on an L1. This was apparently so, as the company can only apply in April for H1B and furthermore needs to wait until a decision is received a few months later before that person on L1 is able to start. A way around this seems to be to not inform the companies of your visa status and focus on doing really well in the interview and informing them later about the visa. To factor this in, a multiplying factor of 80-100% is used at crossroad: “Finding H1-able company”.

The rest of the probabilities are similar to that in the previous scenario.

Effective Value

The exact same impact calculations from the DS master’s path on failure impact, additional costs and success is used here as well. The decision tree in sheets can be found here.

Lifetime Impact

Filling in the above values in the Google sheets estimates the life time impact of the DS L1 path to vary from 642k$ to 875k$, with my “best-guess estimate” as 642k$.

Scenario 3: Working as a researcher

The above decision tree shows the different parts expected along the “GiveWell Researcher” route. I would start by applying to GiveWell and working on improving myself in the next 4 years. Hopefully this gets me into GiveWell. From there on it can be anything such as growing in GiveWell or even starting up.

The Probabilities

The following table shows the different crossroads and probabilities (Success/other) associated with them. And the last line repeats itself until I am 65 when I would need to retire.

The general chance of getting into GiveWell or for that matter any EA org is quite low just looking at the number of people applying to it. For example, the acceptance rate for a GR at Open Phil was less than 5% of the people who applied in 2018. At the moment I am sure I am not GiveWell material. I didn’t get two internships with Happier Lives Institute and Charity Entrepreneurship which informs me that I atleast need to work on myself. However HLI informed me that my internal validity assessment, for one of their tests was “one of the best”, so all hope needn’t be abandoned yet. In a few years I can “build the skills” through DP hopefully and see where that gets me. I guess based on this, that there is a 1% chance of me getting in now, a 20-30% chance that I get in after 2 years of work, and a 60% chance that I make it in 4 years from now.

The chance that I don’t have to move back to NL and pursue other options either due to recession or general firing, is assumed to be the same as was the case with DS master's and DS-L1 routes. So a 96% chance.

Effective Value

To compute the effective value we need the following in addition to the probabilities above.

1. Success impact: Lifetime calculations if faced with success all the time in the decision tree.

2. Failure impact: What happens when I don’t get into GiveWell. What is going to be the impact?

Failure impact:

I assume that if I fail before 40 years I still can give DS L1 route a go, and if that fails, then ‘DS in NL’. Post 40 I just assume that I can only do ‘DS in NL’ for the rest of my life. These numbers are calculated based on the same variables as estimated in DS-master's and DS-L1.

Success Impact

Lifetime impact of working in an EA org is calculated based on GiveWell’s 2015 impact post. GiveWell moved 110m$ in 2015. But the amount that is to be associated with the researcher is much less. How much is it? The values used to estimate this are shown in the table below.

GiveWell moved 110m$, but there were other agents involved as well. It doesn’t seem fair to attribute all the impact to just GiveWell, or the donors, or the orgs performing the interventions such as AMF or Give Directly.

There seems to be two schools of thought when it comes to assessing impact when multiple agents are involved. One is through shapely values as motivated here and the other is through counterfactual impact as explained here. Shapely values seems to divide the value among the different agents equally. Whereas, counterfactual impact captures how much impact an action “truly” has, compared to doing nothing. However using both seems not to be advocated in the links above. I prefer using counterfactual impacts as it gives the agent’s impact when it exists and doesn’t exist.

Shapely Values: Just like we divide the impact of the org among the different people working in the org, it seems appropriate that we should also divide the impact of the 110m that GiveWell moves, among the following: the Donors (e.g., Good Ventures), the middle men (GiveWell) and the intervention Orgs (AMF). The impact from the 110m$ seems to belong to each of these Orgs. I assume all Orgs have equal weight in this, for lack of any other better estimate. So we get a multiplying factor 0.33.

Counterfactual impact of the org itself

It “seems likely” that if GiveWell was not formed by the 2 co-founders in 2007 then someone else would have co-founded it (perhaps a few years later). The reason this is being said is because there are many orgs that today, try to work with “cost-effectiveness ideas” trying to do the “most-good” to identify interventions that are “worthy to put money in”. There are many orgs such as Animal Charity Evaluators (ACE), Open Philanthropy Project (OPP), Happier Lives Institute (HLI), Charity Science Health (CSH), Charity Entrepreneurship (CE), The Life You Can Save (TLC’S) etc., that formed between 2010 and 2020. Thus for the hypothetical situation I would imagine a “GiveWell-like” org being formed say 7 years later by some other people. The two scenarios we thus use are: Actual–> GiveWell forms in 2007; And hypothetical–> GiveWell-like org forms in 2014.

The counterfactual impact of the org is, how much more impact the actual scenario has over the hypothetical scenario. In 2007 the contributions in both scenarios are:

Actual Scenario: GiveWell moves 2.4m$ to charities with cost-effectiveness 7. (In 2018 cost-effectiveness was 18, so 7 seems conservative. Money mainly went to their current top charities like AMF all along).

Hypothetical Scenario (tabulated below):

• 2.4m$ moved to “other-orgs” with much lesser cost-effectiveness of 1. Give Directly has the same Cost-effectiveness.
• 292k$ not used to run GiveWell the 2 person organization in 2007, with cost-effectiveness of 1. GiveWell costs for 2 co-founders is based on GiveWell salaries for 41 people in 2018.
• 100k$ Earning-To-Give by the 2 co-founders which goes to orgs say with cost-effectiveness of 5. The co-founders started GiveWell in the actual scenario so I think they might have had more than cost-effectiveness of 1 for their money.

The scenarios are evaluated in this Google sheets until 2037 when it is assumed GiveWell ceases operating, in the actual scenario; and in the hypothetical scenario it ceases operating 7 years later 2044.

How much GiveWell moves (from 2007 for the actual scenario and from 2014 for the hypothetical scenario) is taken from it’s website. For example, in 2018 GiveWell moves 125m$. So, in the hypothetical scenario in 2025 the GiveWell-like org is expected to move the same 125m$, i.e., with a 7 year delay.

The outcome seems conservative as the cost-effectiveness estimates for the actual scenario is assumed to be a factor 2 atleast smaller for all years. Cost-effectiveness for 2018 is 18 while we assume throughout that the cost-effectiveness is just 7 for GiveWell moved money. Also, in the actual scenario, the impact of the people founding GiveWell-like org in the hypothetical scenario, is assumed to be 0 for ease of calculation.

In order to compare the two scenarios, we multiply the dollars with cost-effectiveness and add it up for each scenario, per year. We then compute the Net Present Values as of 2007 for actual scenario and hypothetical scenario. This allows us to see the value of starting GiveWell 7 years later. For the future years from 2007 we assume a discount rate of 5%.

The counterfactual impact is thus estimated to be 37% better than the hypothetical scenario as computed in the Google sheets.

Impact attributed to Researcher:

An entry level researcher’s contribution is expected to be 10% of a co-founder in 2015. GiveWell employed 9 co-founders that year. So we get a multiplying factor of 10%/9 on the total money moved in 2015 (110m$).

Over a lifetime I find it hard to believe that one will be stuck at 10% of the co-founder contribution. With time one would move up the ladder and so we assume a growth rate of 8% on this 10%. This implies that one will end up having an impact of 46% of the co-founder after 20 years of experience, and 100% after 30 years of experience. Which seems reasonable to me.

Counterfactual estimates for the researcher

The explanation is as follows: There are two scenarios Actual Scenario and Hypothetical Scenario as shown in the table below. In the Actual Scenario:

• AA to II people exist
• AA to FF work in EAO.
• GG to II do ETG.

When AA is removed, you get the Hypothetical Scenario:

• The whole chain is assumed to displace up.
• GG moves to work in EA and his ETG is lost unfortunately.

A, B, … I refer to the contribution of AA, BB, …. II in an EA job. Here we assume all EA jobs have the same impact, EA (makes the calculation simpler), and we quickly see that the Counterfactual Impact (CI) = A-G x EA + ETG by GG. It has an AA better than GG component and an Earning-To-Give by GG component.

Assuming AA managed to get a job in GiveWell, and GG was just not good enough to get into the meritocraticaly last research job in EA, then I expect AA to be atleast 10% “better” than GG. AA produces 10% more “impact” than GG for the same number of years worked per salary units. For lifetime calculations we assume variation between 5%-15% with best guess at 10%.

And GG being an aspiring EA is expected to donate 10% of his/her salary. If they are from US this would be in the order of 20k with a 200k salary let’s say. If they are in the EU it would be in the order of 3-10k (which I see as feasible). I assume for now that GG is from the EU and hence suggest that his/her base donation would be around 3.5k$ (Additional Counterfactual ETG Base).

To count the lifetime impact we need growth rates. Typical growth rates I see in Netherlands are 3%. I will stick to that for GG aka Growth rate Attributed to Counterfactual Earning-To-Give Base = 3%.

The following is the summary:

Lifetime Impact

Filling in the above values in the Google sheets estimates the life time impact of the GiveWell Researcher path to vary between:

Even if I consider shapely values and therefore have an additional multiplication factor of 33%, the values are comparable to the best-guess estimates of DS-masters (732k$) and DS-L1 (642k$).

Note: Varying only the counterfactual contribution from 5-15% captures the entire variation from 840k$ to 1.9k$. This goes on to inform an audien that the GREATER you are than GG the GREATER the impact.

Scenario 4:

The above decision tree shows the different parts expected on the “Charity Entrepreneurship” route. I would start by applying to Charity Entrepreneurship between age 30 and 34 while working on improving myself during that time.

If I don’t get into Charity Entrepreneurship in that time, then I would try to get into research in GiveWell and if that doesn’t work out, the back up options would be to go back to data science in Netherlands or attempt the L1 option.

If I get into Charity Entrepreneurship, I startup my own charity which on average is expected to last 4 years with a certain impact (according to Charity Entrepreneurship cost-effectiveness analysis), and then one has to move on to something else. Perhaps research/management in GiveWell/GiveWell-like org!

The Probabilities

The following table shows the different crossroads and probabilities associated with them. And the last line repeats itself until I am 65 when I would need to retire.

Currently I think there is a very low chance of getting into Charity Entrepreneurship and becoming an incubee. I didn’t manage to make it to their internship. Hence the 99% chance of being rejected when I apply now. Let’s say over time I can develop skills and credentials needed for it.

What I see in existing co-founders is that they have a “top notch” education (like economics, philosophy master), already have experience leading teams in social work and other, and/or have started local EA chapters somewhere in the world, e.g., Varsha who is running Suvita, Michal who is running Policy Entrepreneurship Network.

Let’s say I manage to support an existing EA group in Netherlands, and work on finding volunteer positions at CEA to gain their confidence for other roles, or even take up roles in EA research via unpaid internships. Even then, with the massive pool of people (2000 applications for ~20 positions), it seems hard (1% acceptance rate) to get into Charity Entrepreneurship. Hence the 15% (1-85%) and 25% chances at age 32 and 34 (attempts 2 and 3 respectively).

In case I do make it as an incubee, I would be starting up and working on the charity for 4 years (which is the estimated lifetime of an average charity that Charity Entrepreneurship supports). After which I would need to pick up something else. During the time of working on the startup charity, I assume similar probabilities as done before, 96% chance of success of completing that year and getting the impact for that year.

After Charity Entrepreneurship charity, I could perhaps work in research organizations like GiveWell. In Scenario 3 I assume a probability of 20% and 60% to get into GiveWell in attempts 2 and 3. Here I assume similar probabilities for attempts at age 37 and 38 (35% and 60%). At age 37 I expect a higher chance of getting in GiveWell having already started a charity and worked on it.

Effective Value

To compute the effective value we need the following in addition to the probabilities above.

1. Success of GiveWell

2. Failure case when neither Charity Entrepreneurship nor GiveWell works out.

3. Success of Charity Entrepreneurship

Success of GiveWell is already known from the previous scenario. Failure would lead to either DS in NL or DS L1 based on the age. For the Success of a Charity Entrepreneurship charity consider the following:

There is a cost-effectiveness analysis from Charity Entrepreneurship as to how much “counterfactual impact” a co-founder could have. They claim it is 100k$ given to Malaria consortium (top GiveWell charity) or 1.5m$ given to Give Directly per year. I am not convinced with this and explain why, below. The value I expect it to be is 15k$ given to Malaria consortium.

For this counterfactual impact the Actual Scenario is founding of the Charity Entrepreneurship Charity; Hypothetical Scenario is when the Charity Entrepreneurship Charity is not founded.

The cost-effectiveness analysis thus has 4 components to it to compute counterfactual impact of the Charity Entrepreneurship Charity:

• Estimated impact due to Charity Entrepreneurship Charity (add)

  • Estimated size of Charity Entrepreneurship charity or revenue to an “average” charity is 1.46m$.

  • Estimated average cost-effectiveness of their charities is 7.

  • Estimated impact due to Charity Entrepreneurship Charity is 1.4mx7 = 10m$ per year per charity donated to Give Directly.

  • All the values are from Charity Entrepreneurship’s calculations.

• Estimated loss in impact due to taking away funding from other charities (subtract)

  • On average Charity Entrepreneurship charities redirect 26% from GiveWell top charities, 8% from Give Directly, 36% from non-GiveWell etc.

  • Cost-effectiveness of GiveWell top charities is 18 as of 2018 (15.8 according to Charity Entrepreneurship), Give Directly is 1, and non-GiveWell orgs is 0.38.

  • Average cost-effectiveness is 26% x 18 + 8% x 1 + 36% x 0.38 ... = 5.3.

  • Estimated loss in impact = 1.46m$ x 5.3 = 7m$ per year per charity donated to Give Directly.

• Estimated loss due to one-time costs of Charity over entire lifetime (add)

  • This is assumed to be 100k for a 2 person charity started in the UK.

  • Unfortunately the cost-effectiveness analysis seems to suggest that this 100k would be given to Give Directly if there was no Charity Entrepreneurship Charity. I disagree that cost-effectiveness = 1.

  • I think the cost-effectiveness of this 100k would also be 5.3 (just like the previous case).

  • We get 100k/4 x 5.3 = 123k$ per year per charity, donated to Give Directly.

• Estimated loss in impact due to co-founder Earning-To-Give (Subtract)

  • If the co-founders decide not to go through with the org then they ought to be doing something else. Here it is assumed they could be Earning-To-Give. Considering that they are already co-founder material, Charity Entrepreneurship assumes a salary of 100k and a donation of 50%. There is 100k per year available to be donated anywhere, assuming 2 co-founders.

  • Unfortunately they seem to be thinking that this 100k would also be given to Give Directly instead of the other more cost-effective ones available with GiveWell.

  • I think the cost-effectiveness of donations can be as high as 18 assuming the co-founders just hand over the money to GiveWell to do what it sees fit. This being the case means that the estimated impact of both co-founders doing Earning-To-Give is 100k x 18= 1.8m$ per year per charity donated to Give Directly (GD).

This is summed up in the following table:

Adding and subtracting as in the table above, we get average impact of a charity per year in dollars donated to Give Directly is 527k$. If we want to compare it to regular donations then, we should compare it with say dollars donated to GiveWell (cost-effectiveness = 18). Looking at this 527k$ in terms of dollars donated to GiveWell–i.e., dividing by cost-effectiveness of 18–we get 29.3k$ per year per charity donated to GiveWell. This is for 2 co-founders. For one co-founder this is 14.6k$ per year per charity.

Lifetime Impact

Filling in the above values in the Google sheets estimates the life time impact of the Charity Entrepreneurship path to vary between 914k$ and 1.97m$, with the best-guess estimate at 1.45m$. The main reason this is so high is because of GiveWell researcher route being followed after the 4 years of founding a Charity Entrepreneurship Charity.

Conclusion

The lifetime calculations are shown below:

It seems quite clear that pursuing the GiveWell route has very high potential even if we look at the lower bound of the estimate, which assumes I am only 5% “better” than the person being displaced at the end of the chain (GG), into Earning-To-Give.

Charity Entrepreneurship route and GiveWell route seem to have similar impact. This is not too surprising as for most part the Charity Entrepreneurship decision tree has GiveWell researcher decision tree in it. I feel more comfortable going into the GiveWell route, where in-theory I can work-from-home and improve my skills. Maybe later in life once I have some experience in research I can still consider founding etc., (like Milan Griffes or Peter Hurford).

But so far as the estimated lifetime impact (effective value), the GiveWell route seems to be the clear winner.

Opportunity cost of starting DS now vs later

I know DS in US is a cash cow and that is what I am chasing as a backup in-case for any reason it is not possible to get into GiveWell or GiveWell-type orgs (Open Phil etc.). While trying to get into GiveWell I think I will need a day job in the Netherlands and easiest is to continue with what I am doing. However I propose to already change to DS and work in DS while I continue to try to get into GiveWell. In the case that GiveWell doesn’t work out, at age 35, I would already have a few years experience in DS and can push towards the L1 route without spending time on learning DS afresh at 35. In order to see what sort of impact this would have on lifetime calculations we look at the following cases.

Case 1: Start GiveWell route now and at 35 start DS for the first time, in case GiveWell career fails (Google sheets: Opportunity Costs 1).

Case 2: Start the GiveWell route 1 year later, transition to DS in the mean time (Google sheets: Opportunity Costs 2).

Life time impact (best-guess-estimate) of case 1 seems to be: 1.363m$, while life time impact of Case 2 seems to be: 1.359m$. This is a very small number to take the difference seriously. For example, in case 2: I have not included the effect of having 3-5 years experience in DS when I have to pursue DS in L1 (when GiveWell doesn’t work out). The starting salary would be higher in US in that case. So Case 2 could potentially trump Case 1 then.

If the difference were in the order of 50k$, then I would seriously consider focusing on the better case. Now it just seems that both cases would end up with “similar results” and I have a strong preference for Case 1.

Stats

Total words: 8k
Total time: 102hrs Avg.time/day: 4hrs

Start date: 22-08-2020 End date: 15-09-2020

Footnotes