Value Estimate/Funding Proposal for Canadian ODA Campaign

Bellow I lay out the case for a project that I believe may have a higher expected value then donating to GiveWell charities.

There are a number of points that I still feel uncertainty about, but I thought I would err on the side of getting feedback early. Let me know what you think!

Background

The Canadian Council for International Cooperation (CCIC) is an organization that brings together Canadian non-profits focused on international development. At their conference in October in 2017, they presented a pitch for their (as of then not-public) campaign to attempt to get Canada to honour their commitment to give 0.7% of their Gross National Income (GNI) to Official Development Assistance (ODA). If successful, this would represent an increase of approximately 9 Billion Canadian Dollars (CAD) per year to Canada’s ODA budget. [1] The campaign struck me as particularly compelling and well-researched, the consultants the CCIC employed to develop it seemed extremely knowledgeable. In addition, the Bill and Melinda Gates Foundation, which funded the campaign's creation, have renewed their funding to hire a new campaign manager. After reviewing the campaign in more detail, I have become convinced it would be worth directing EA funds toward this initiative.

My proposal is to develop and run an additional campaign office, comprised of four members of the Effective Altruist community (myself included) to provide support to the CCIC campaign. I believe bringing Effective Altruists together for this project will have additional benefits beyond this campaign. I note these at the end of the post. That said, the main focus of this post will be the expected value of the project itself.

Expected Value Calculation for this Project

My calculation of the expected value for this project has three variables.

1. The Counterfactual-Adjusted Increase in ODA if the 0.7% campaign is successful
2. How much more likely is it for the campaign to be successful if 4 EAs work on it full time
3. The Discount Rate between Canadian ODA and Equivalent money to GiveWell Charities

I provide details for each of these variables below. I include the number I think is most plausible for each, as well as my 90% Confidence Interval for each (that is to say, the range of estimates which I believe captures 90% of the probability space). If you want to generate your own expected value for the program using these variables, feel free to input your own estimates into this spreadsheet.

1. What is the Counterfactual Adjusted Increase in ODA if the 0.7 campaign is successful?

To calculate the total amount of increased ODA attributable to the campaign, we need to think about how much longer we would expect to wait for the ODA increase to happen if the campaign wasn’t successful.[2]

The shortest time we could expect would be 1 year, since government budgets are yearly. Given that a substantial increase to the aid budget is likely to only be announced during election years (or more precisely, the half-year leading up to an election), I think the yearly probability will be unevenly weighted in favour of election years. We can also calculate the background probability of an increase getting adopted during any given election cycle. Given the polling done on the issue (referenced in the report), I think there is a 20% chance that it will be adopted by a party during any given election cycle from now on in the absence of influence from a campaign of this kind. [3]

If we go solely on the baseline of 20% chance per 5 year election cycle, we should expect the counterfactual impact to be somewhere in the range of speeding implementation up by 25 years. That said, this does not capture the possibility of the program being implemented between election years, and if we think the probability per non-election year is 5%, then we get about 10 years counterfactually.

Given these considerations, I would say 10 years is a good estimate. My 90% confidence interval would be between 1 year and 30 years.

Throughout each step in this post I am going to note the number I find most plausible at the end of each section. As I noted before, you can input your own estimates for each of these numbers using this spreadsheet.

So 10 years, at 9 billion CAD per year, is 90 billion CAD.

Another consideration is how likely it is for the budget increase to be repealed. This is also a hard one to reason about in the abstract- there are only a few other concrete examples of other OECD countries increasing their spending to reach the 0.7 ODA target agreed to by the UN. The most analogous case I think is that of the UK, and it has updated me toward the view that once it is agreed to, there is a good chance that there will be cooperation amongst parties to keep it in place. To err on the side of conservatism I will say there is a 50% chance that the budget increase will be canceled, even in the case where the campaign is successful. We can take the additional conservative (and simplifying) assumption that we can ignore whatever increase in spending would have taken place before the program was canceled. So we reduce the 90 billion figure by 50% and get:

45 billion CAD in increased ODA payments

2. How should we expect 4 EAs working on the Campaign full time to effect its probability of success?

Like all of these variables, I find this a very difficult one to reason about. I assume that the CCIC and the Gates Foundation would only invest the resources (personnel and money) if they thought there was at least a noticeable chance that the campaign would succeed (they could of course be wrong in this). I believe that their belief here counts for something, but I am also cognizant of the fact that most campaigns fail, and so my calculations for how much we can make a difference here will be based on some healthy skepticism.

One of the details of the current campaign that makes me confident that EAs can have a disproportionate impact by supporting it, is the fact that CCIC is currently extremely time constrained in terms of people to run the campaign. The previous campaign-manager’s contract has ended, and it has fallen on the current communications officer to run the campaign (and they are only able to allocate about a third of their person-hours on the project). CCIC is hiring for a new role, and one of their responsibilities will be to help run the campaign (but this will represent only 50% of their time). Meanwhile CCIC has done a significant amount of work setting up the infrastructure for this campaign (following from the advice from the consultants) and they have gotten buy-in from their members. CCIC’s member organizations, which include the likes of Oxfam, Save the Children, and World Vision, have a postcode-mailing list of 1.3 Million Canadians. In other words, 1 in 30 Canadians have signed up for contact from one of these organisations. These organisations significant reach, and their thousands of volunteers, are an incredible resource, that likely won’t be fully utilized without coordination. Given their current plans, CCIC will only have 0.5 to 0.8 Full-Time Equivalent hours (between two staff members) working on the campaign (0.5 of which will be a new hire that they are still conducting the hiring round for). I don’t believe this will be sufficient to fully tap into this network. If you want to know more about what projects the EA staff would be working on, you can see my current best-guess here.

So with that context in mind, what effect would we expect 4 additional full-time EAs to have on the campaign? Here I am going to err on the side of being non-conservative, and say that a well run campaign by a centrally located organisation, with pre-existing access to policy makers, politicians and aligned NGOs, can exert a noticeable difference on the chances of a policy like this coming about. Furthermore, given the current staffing situation, I believe an injection of 4, highly motivated, strategically inclined, parallel staff, will substantially improve the prospects of this campaign. I have a 90% Confidence Interval that 4 EAs could increase the probability of success by somewhere between a one in ten thousand (0.0001) chance to a one in a hundred (0.01) chance. I will land on the logarithmic middle result: that it will increase the chances of the policy being adopted by one in a thousand (0.001).

45 billion CAD in increased ODA payments x 0.001 (increase chance of success from EA involvement) = 45 Million CAD

3. The Discount Rate between Canadian ODA and Equivalent money to GiveWell Charities

Next we need to compare the value of 1$ going to Canadian ODA to 1$ to an average GiveWell Charity (if we are using EA funds, and EA talent, focused on Global Poverty causes, this is the relevant opportunity cost).

The first thing to note is that even under optimistic timelines, the increase in ODA spending to its full amount (0.7 GNI) will take 5-10 years, and so if we are optimistic that governments will continue to improve in their ODA allocation decisions, we can expect the ratio from ODA to GiveWell to be more favorable in the future than it is now.

That said, I am somewhat pessimistic that things will improve too dramatically in the next 5-10 years, and I am also fairly pessimistic about the average quality of ODA spending. I would say that, on average, ODA funding is around 5% as effective as average GiveWell donations. My reasoning for setting it as 5% is that roughly 15% of the budget is already spent on evidence based health interventions, many of which overlap with the interventions pursued by GiveWell Charities (This includes 800 Million to the Fight Aids, Malaria and TB fund). That said, it is unclear, from my shallow investigation, what percentage of this budget allocation funds the interventions favored by GiveWell charities. Furthermore, the vast majority of ODA does not go to these public-health interventions, and given the core EA observation that most of the effect comes from the tail, missing these interventions could be akin to giving up almost all of the impact. So if we assume ODA spending is about a third as effective in these areas (because most is being spent on non-top interventions in those areas) [optimistic], and we assume that all non-health aid has no effect [pessimistic], then we arrive at 5%. As I do a more in depth analysis of Global Affairs ODA spending for my PhD dissertation, I will be able to give a more thorough breakdown, but for now this is my best (extremely rough) guess of what the ratio should be.

45 Million CAD in ODA x 0.05 (GiveWell:CAD ODA Ratio) =

2.25 Million CAD in GiveWell Donation Adjusted Value

If the cost of hiring 4 EAs to improve the campaign is 125,000 CAD (see below), this would represent a 18x multiplier compared to the same money going to GiveWell charities. It would lead to a net 2.1 Million Canadian Dollars in GiveWell-Donation-Adjusted Value being generated.

These numbers change slightly based on the opportunity costs of the people involved, and the effects it has on their career trajectory, but I think in most plausible scenarios the absolute amount, and the ratio, still seem very promising. I discuss this point further in the final section.

Rough Budget

This represents the top-end of what I think would be worth funding for this project. Depending on the communities interest in the project, and availability of funding, we could run on a reduced budget.

The funding provided would be for 4 EAs (myself and 3 additional EAs) to work on the project full time from late February to the middle of November (The Canadian Election is late October).

Standard Budget:

• 3 EAs on 9 month contracts, paying 3500 CAD/month, for a total of $94,500 CAD
• For Me: A TA buyout for Summer (4 months), at 1500/Month, and the above contract for the final three months (3500/month) for $16,500 CAD.
• In addition I would set $1000 for consultant fees (to work with the consultants who designed the campaign) and legal fees (to insure that all strategies that are adopted by the campaign follow the legal parameters for charities involvement in political action).
• I will also budget $1250 CAD per month for an office downtown, close to the CCIC offices and Parliament Hill x 9 months = $11,250 CAD.
• Total cost= 123,250 CAD

The three hires will ideally be able to move to Ottawa, but I will also happily consider individuals who would prefer to work remotely.

Reduced budget [2 staff, no office]:

• 1 EA on 9 month contracts, paying 3500 CAD/month, for a total of $31,500 CAD
• For Me: A TA buyout for Summer (4 months), at 1500/Month, and the above contract for the final three months (3500/month) for $16,500 CAD.
• In addition I would set 1000$ for consultant fees (to work with the consultants who designed the campaign) and legal fees (to insure that all strategies that are adopted by the campaign follow the legal parameters for charities involvement in political action).
• <remote work: no office>
• Total cost= 49,000 CAD

I am unsure how to calculate the relative effectiveness of the two budgets: If we assume that there is a lot of value on the margin of adding additional staff time, then since we have half the staff, we should expect half of the probable effect (reducing the expected value by 1 Million CAD). On the other hand, if we expect that additional staff won’t add much marginal value, and almost all the value is captured by the team existing, then doubling staff may only improve base chances by, say 10%, So the lower budget would affect the probability by 0.0009 rather than 0.001, reducing expected value by 200,000 CAD (and only reducing it net 120,000 CAD when considering the cost of the higher budget).

I believe there are additional benefits to this project, beyond it increasing the probability of the policy being adopted. The first (perhaps selfishly!) is that it will provide me with a number of key contacts in top anti-poverty organisations, as well as to MPs that I wouldn’t otherwise have access to. In the case of a successful outcome, this could translate to significant career capital. In a non-successful but well run campaign, this could also be something that would contribute to my career capital. This would also be true of the other EAs working on the project, and since I would be selecting for candidates who are interested in pursuing EA policy in Canada, I suspect this would increase their life-time impact as well. I also believe that this will have information value for EAs, as it will be a concrete case of how campaigns work. Throughout the campaign, the team will be reading up on and synthesizing the key literature on how to run a successful campaign, and as part of our debrief of the project, we will summarize our main lessons, so other EAs can learn from them, and apply them to future campaigns. I say a bit more about these external effects of the project in the final section bellow.

[More Speculative] - Some More Thoughts on the Indirect Effects of the Project

I have focused primarily on the direct effects of the project. These direct effects might be either significantly improved or diminished by the project’s indirect effects. I will try and say something in a bit more detail about three indirect effects. My views on each of these are weakly held.

1. Opportunity Costs

The evaluation did not consider the opportunity costs for the individuals involved. Since I will be targeting EAs, there is a plausible case that the opportunity costs for these individuals are high. An initial way to generate the opportunity cost could be to calculate the amount of money that the EAs could have donated, had they pursued other opportunities (such as earning to give). Obviously, if they individual had an extraordinary earning-to-give potential (either right now, or by skilling up to increase lifetime earnings), then this could have a significant effect on the value calculation. For example, if someone could have increased lifetime donations by 500,000 CAD by taking a different course of action over the 9 months, and if the team was made up exclusively by four such individuals, this would reduce the counterfactual effect of the campaign to nearly zero. Clearly for most EAs, their earning-to-give impact won’t be so significant, but since EA is currently far more talent constrained than funding constrained, distracting a, for example, top-tier researcher or operations individual for those 9 months, from other higher return projects, could have equivalently significant negative effects on the counter-factual. To mitigate these downsides, I will be aiming to (1) select EAs whose immediate opportunity costs are not extremely high (i.e. they are not trading this off with significant skilling up or EtG potential), and ideally (2) I will be targeting EAs whose career capital will be improved by working on this project.

Before I say a bit more about career capital, it is worth considering what a reasonable estimate for the opportunity cost for individuals (ignoring career capital) could be. My very rough guess is that 9 months of an EA’s time is worth roughly 100,000 CAD. This is a weakly held view, but I arrived at it by considering that it is likely the average of the non-tail earning-to-give amount for 9 months of income [I assume that being with extraordinary earn-to-give potential would not apply], and the view that many EAs can work to fill talent gaps that the community values at around 100,000 CAD for 9 months [Again, conditional on the sort of candidate I would be aiming to hire]. This 100,000 CAD figure sits in the middle of my 90% CI of between 10k and 1 million. If we take this number, then the total counter-factually adjusted cost for the full budget is raised to 524,000 CAD. This means the project (ignoring its other indirect effects) would only be a 4x multiplier, and would only net 1.5 Million GiveWell-Adjusted dollars. If we take the upper bound of my CI, then this could be a counterfactually harmful project. That said, since I will be actively selecting for EAs who either have lower opportunity cost, or gain useful career capital from the experience, I do not believe this negative outcome is a likely one.

2. Career Capital

As I noted above, I believe that this project will have beneficial effects on my career capital (so long as I continue to focus on Canadian Development Spending). The network offered by the CCIC is quite extensive- it extends to the major NGOs in the development space, as well as to members of Parliament, and to the Minister responsible for Canadian ODA. In the process of driving the campaign forward, the EAs working on the campaign will have the opportunity to interact with senior members of the CCIC and senior members in organisations in their network (Oxfam, Save the Children, etc.). These contacts will be valuable, not only in securing future, more senior, roles in the industry, but also in having the potential to get a better sense of how to get new policy enacted in the Canadian context (and thereby how to get more EA-leaning policies to pass). In addition, if the campaign is successful, it will be a strong piece of work experience for future campaigns, and could therefore translate into political capital for partisan politics. I find it difficult to put a dollar amount on these benefits. My guess is that if this could, for example, execrate the EAs working on this project’s seniority, or improve the probability of them landing a more influential position by a significant degree, this would have a noticeable impact on their overall career trajectory. If we imagine that an EA focused on a career attempting to implement EA policies in Canadian ODA, will in expectation influence 10 million GiveWell Adjusted CAD in spending over their career, and this improves their career trajectory by 1%, then it would be worth 100,000 CAD. My CIs here are wide, and my overall position is weakly held, but if we think that EAs on this path can have a significant impact, than anything that even has a minor effect on their career capital, likely adds a significant amount of expected value to the project.

3. Information Value

The final benefit of the project I will consider is the value of the information it provides to the rest of the EA community. Since there have not been many EA projects of this kind, what we learn from the campaign, and the benefits of working with organisations like the CCIC might inspire other projects in the future. The value of the information could come either in inspiring other campaigns that otherwise wouldn’t have existed, or in improving campaigns that would have happened in either case. As I mentioned above, one of the goals of the project will be to actively read and synthesize details on how to run an effective campaign, and to apply these to our work. We will also have an analysis at the end of the campaign to outline what we learned, and what we think we could have done better, so that other future projects of this kind could be improved. I have a very hard time calculating a specific number for what the information value of a project like this could be. If the project goes well, and it reveals information that implies that other similar campaigns could go well, it could lead to other multi-million-dollar-in-expectation projects to exists that otherwise would not exist. Conversely, if the campaign goes very poorly, this could serve as important information to either help other campaigns do better, or deter other EAs from pursuing this sort of project in the future. Depending on the quality of the information derived from this project, and its use by the EA community in the future, it is possible that the information value of the project could exceed its direct effect. That said since I am deeply uncertain about the nature of these potential future projects, I am reluctant to set any specific dollar value here.

Footnotes

[1] The current budget is 0.24 GNI, so when we multiply Canada’s current GNI by 0.46 we get approximately 9 Billion CAD

[2] It’s worth noting that the full increase won’t happen all at once, it will mostly likely increase something like 15% a year until we reached the full 0.7 target, that said, since other counter-factual changes follow the same scale-up timeline, we can ignore it for sake of simplicity

[3] On the one hand, an increase the ODA budget is consistent with what Canadians tell us in polls, on the other it’s not a vote moving issue, and has polled well for several decades without any party incorporating it into their platform. So 20% might be a bit optimistic, but since that reduces the amount that can be counted as counterfactual impact for the campaign, I will keep the number high to er on the side of conservatively evaluating the potential counterfactual impact of the campaign