post by Grayden
Invest to Give
There have been a few discussions in the EA community about the merits of donating now versus investing and donating later. I think it is a very valuable question to answer, but I believe some have used flawed logic to conclude that we should donate later. In this article, I make the case against delaying deployment of funds to the distant future. I also suggest a novel approach, which I call Keynesian Altruism, that involves donating during global economic downturns to maximise impact when returns from investment are at their lowest. Using this logic, I conclude that now (August 2020) is a good time to be deploying (i.e. spending) funds.
I identify as an effective altruist and have earned-to-give for many years. I have studied advanced finance at a leading university and work in the asset management industry. All views are my own and unrelated to my employment.
In this piece, I have assumed a basic level of finance, statistics and economics knowledge. I have tried to be focused, concise and readable rather than exhaustive, rigorous and academic. I have also tried to avoid duplicating the contents of the articles to which I refer.
Invest to Give
A number of people have written some very thought-provoking articles asking whether we should invest money now and give at a later point:
The case for investing later [EA · GW]
Giving now vs. later [EA · GW]
The logic made in these article essentially goes as follows:
1) You can make positive inflation-adjusted returns by investing in equities;
2) On average the amount of money you will have in x years will be higher in real terms than you have today;
3) By deploying the money later, you will be able to buy more.
The proponents of this are very upfront about one-tail risks:
1) An existential event could cause complete loss;
2) The values of the fund may drift over time (my favourite example is the Communist Party of Great Britain, which disbanded in 1991 and became Democratic Left then New Politics Network then Unlock Democracy in 2007, hence within 16 years, the cause had changed from communism to advocacy for participatory democracy through a written constitution);
3) There are diminishing returns on deployment of capital and it is likely that EA will have more resources in the future;
4) We may (or may not) be living at the most influential time in history.
These are very well thought-through considerations that I broadly agree with.
The first issue I have with this is that it treats investments in equity as if they are risk-free. When you invest in equities, theory (CAPM) tells us we get paid two components: (1) a risk-free return + (2) a premium for beta. Component one, the risk-free return, is essentially compensating you for time. This is sometimes positive in real-terms and sometimes negative in real-terms. In short, if you invest in super low-risk investments (e.g. Treasuries), you may or may not have more money in x years in real-terms. Based on data up to 2011 compiled by Robert Shiller, the real risk-free rate of return has averaged -0.2% over the last 5 years (geometric mean), -0.1% over 10 years, +1.2% over 20 years, +1.9% over 50 years and +1.2% last 100 years, so typically more often positive than negative, but recently more often negative than positive.
Component two, the premium for beta, is essentially saying that you get paid for taking on risk that the market cannot diversify away, and this component should be ignored because it is a return you get for selling risk not selling time preference. There is a market for risk, just like any market for goods. You can sell risk (get paid to take it on) or buy risk (pay someone to take it off you). Like all markets, there is an equilibrium at which the market clears and this determines the price (for the finance theorists out there, this is the market risk premium multiplied by the forward volatility). When you buy equities, you are being compensated for taking on risk that can’t be diversified away. Essentially someone is willing to pay you around 5% (the market risk premium, with beta=1) to put your money through a probability function that spits out a roughly normal distribution with a standard deviation of 20% (the long-run normal level of the VIX). There is real economic value in minimising risk and real economic value to have investments that hedge you (i.e. having more money when the economy is weak and you need it); that’s why people are willing to pay for it. When you invest with leverage, you are just selling more risk. Because, at least in theory, you could get the return in a very short period of time and still get compensated for it, i.e. high leverage, short duration, this return is not driven by patience. If I short sell a AAA government bond with maturity of 1 year and use the cash to buy a 1-year forward on a stock or commodity, I will expect to make on average 5% of the notional capital at work, but (subject to no counterparty risk) I don’t need to invest any money at all. On one day in a year’s time, I will receive an uncertain amount of money and have to pay a certain amount of money, which will be either an infinite per-annum gain or an infinite per-annum loss if you insist on thinking about risky returns on an annualised basis. I’m not saying don’t invest in equities (for what it’s worth, I think EAs should invest in high-beta portfolios), but I am saying it’s not a like-for-like comparison.
The second issue I have is that the basis for it assumes inflation as a baseline. However, the cost of a product or service does not increase at the same rate as the cost to make somebody better-off, i.e. people in the future are likely to be wealthier so giving them the same amount of money (cash programming) or stuff (e.g. malaria nets) will be less impactful in the future. This difference is real GDP growth. Given the choice of deploying 10 malaria nets today or 11 malaria nets in 100 years’ time, I would definitely choose the former. In 100 years’ I hope and expect that everyone in the world can afford a malaria net and our philanthropy is targeting needs higher up the hierarchy. Even if scientific development means we have a malaria net that is twice as effective, I would still rather deploy 10 today. Historically, global real GDP growth has been at least 3% p.a. (and over the long run, lower income countries tend to be above that due to economic convergence). If your investments are not growing at greater than 3% p.a., then your outcomes will be shrinking even if your outputs are growing.
The third issue is that the return (in nominal terms) may be taxed or expropriated. If an individual delays donations, any gains will be part of their personal taxable income / capital gains. While charities normally have some tax efficient advantages, they are not exempt from all tax. For example, until 2016 in the UK charities had to pay tax on dividends but not capital gains. Of course, if doing at large scale, it would likely be possible to choose an offshore tax domicile, but this may also increase the risk of expropriation. It’s also worth noting here that for charities in the UK to maintain a level of capital above what can be justified as reserves, they need express permission in their deed of trust, as this would be considered an endowment. Even then, the original capital amount must be spent within 21 years to stay within the law.
In conclusion, it is more appropriate to compare the risk-free rate (typically -1% to 2% p.a. in real terms) with GDP growth (typically 3% p.a. in real terms), rather than comparing equity returns (typically 5% to 7% p.a. in real terms) with inflation (by definition 0% in real terms). That is even before you get into complications of implementing this strategy in practice: tax, expropriation risk, charity law, existential risk, value drift, diminishing returns and potentially diminishing influence.
While I disagree with delaying deployment of funds to the distant future, that’s not to say I insist on everything being spent now. There are short periods of time when the numbers can make it worth delaying deployment. This is due to economic cycles.
Basic macroeconomics tells us there are times when the economy is working faster than it should (overheated) and times when the economy is working slower than it should (usually defined as a recession). There is a lot of positive feedback in the economy; when demand for stuff is high, businesses make record profits, employees get paid well to retain them and capital is needed for growth. During a recession, the opposite happens and spending on advertising, capital equipment and R&D often falls.
Most governments do their best to reduce the magnitude (i.e. difference between peak and trough) of cyclicality through monetary policy. This involves lowering interest rates, which reduces the return you get from investing (this increases asset prices and explains why the S&P 500 is up year-to-date despite the worst pandemic in 100 years) and encouraging you to spend it instead. Essentially governments are subsidising spending because not enough people are doing it.
Another tool governments have in a recession is fiscal policy. This basically means they themselves are spending to fill the gap. Arguably if you are going to build long-term infrastructure assets (previously highways, now fibre optic), the best time is when lots of people are unemployed, so talented people are delighted to work for less than it would have cost to employ them previously. However, governments themselves are also faced with a reduction in income (less tax due to lower profits / incomes) and an increased in expenditure (more unemployment so more social security payments). This means if they are to boost spending during a recession, they need to be happy to run a deficit and need to be self-disciplined enough to run a surplus during good times. This sort of fiscal policy is known as Keynesian Economics. In my experience most governments do it to some extent, but few do it properly.
So what about the third sector. Just like their private and public sector peers, income for charities also typically falls in a recession. Charities are already making redundancies as a result of COVID-19, even though ironically their services are more required than ever now. There is also a structural cost to reduce a workforce: redundancy costs; other costs of restructuring (e.g. adjusting office space); then hiring and training new people when you return to growth. I believe that not only should charities / foundations run at a deficit during a recession (like a smart government), they should even consider growing to meet the increased need of their beneficiaries. I call this Keynesian Altruism.
From a practical perspective though, this takes guts for a charity to do as it means running 'unsustainably' (as in expenditure exceeds income) sometimes for a period of up to 5 years. It also requires planning in advance: charities don't normally have the reserve levels they need to do this (and as I mentioned above, charity law actively prevents it). The onus often falls on the donors, but they can be equally bad. Many endowments deploy the income they receive from investment income, which of course falls during a recession. I disagree with this approach. In my view, now is a very good time to reduce the size of an endowment. For individual donors it is harder, but if you are truly altruistic, there is no better time to donate, even if your job security may be lower today than a year ago.
Returning to the financial theory we discussed in the previous section, right now the risk-free rate of return is solidly negative (as at 10th Sept 2020, US Treasuries have a -1.28% yield over 5 years in real terms), but the spending power is higher now than it was last year. The price of impact has also likely increased because there is now more need for help.
Finally, while most of this article has focused on a quite complex, there are some basic practical steps that I think everyone can agree upon:
1) Cash sitting in a charity bank account costs money, so if you have lots of it, invest some;
2) Beware the extra administrative burden that comes with managing a charity with investment income (sometimes it’s better to have a bank account that pays no interest than one that pays 0.01% p.a.);
3) Taking time to work out where best to deploy funds clearly has value.
Comments sorted by top scores.
comment by MichaelDickens ·
2020-09-14T16:36:36.214Z · EA(p) · GW(p)
This article seems to be making two distinct claims:
- The standard arguments for giving later don't hold up.
- "Keynesian Altruism": It's better to give when the economy is weaker.
I believe these can be true or false independently. I want to expand a bit on the first claim.
You identify a lot of relevant concerns that I agree need to be addressed, and that often get ignored. I think that even after addressing them, giving later may still look better than giving now.
Are you familiar with the Ramsey equation (e.g., see this SEP entry)? The Ramsey equation states that, in an efficient market, where r is the risk-free rate, is the rate of time preference, is the rate of risk aversion, and g is the consumption (GDP) growth rate. The claim in RPTP Is a Strong Reason to Consider Giving Later [EA · GW] is that most market actors use a value of that's too high, which pushes up interest rates, and therefore "patient" actors should prefer to invest. (Right now, it looks like r < g. I don't know how to explain this. I did a little bit of reading on the matter and my impression is economists believe it shouldn't be true and it's a bit of a puzzle as to why it's true currently, but there are some potential explanations.)
You point out that donors need to worry about taxes and expropriation. That basically means . This is true for both altruists and non-altruists. But as long as most people have a pure time preference and altruists don't, altruists will have a lower than most people, and therefore will relatively favor investing. (I made an attempt to estimate the philanthropic discount rate here [EA · GW].)
Another thing you brought up is that most people don't invest exclusively in risk-free assets. The Ramsey equation does use the risk-free rate, but there's an extended version of the equation that allows for risk. The extended Ramsey equation (taken from here) is where g follows a normal distribution with standard deviation , and r and g are perfectly correlated. When accounting for risk, the same basic theoretical argument holds: impatient actors will push up interest rates, making investing look more promising to patient actors.
Of course, there's a case to be made that this theoretical model doesn't hold up (e.g., current risk-free rates seem incompatible with a positive pure time preference). I haven't seriously studied economics but my impression is economists generally believe this is a good model.
Replies from: Grayden
↑ comment by Grayden ·
2020-09-15T10:06:07.618Z · EA(p) · GW(p)
From a personal point of view, iff my utility curve is linear (i.e. losing 50% of my wealth would have a similar magnitude of utility change as gaining 50% additional wealth) and I know my date of death, then it would make sense to invest for as long as return on capital remains below GDP growth. I would be careful about saying "most market actors use a value of δ that's too high" because I think you can argue what they are doing is perfectly rational; if you're not sure if you'll reach retirement, you'll be less inclined to contribute to a pension (from a purely selfish point of view). Now we as altruists don't have to worry about the date of death because we are helping a pool of people into the future, who don't have to be alive today. However, we do have to worry about utility. To achieve the return on capital, we do need to take on risk. In general, wealthier people are able to take more risks than poorer people (utility functions are more linear at higher wealth). Altruists represent these poorer people (this point is more relevant to global health and development than animal welfare and long-term future), so should be sensitive to undiversifiable risks. In other words, I don't think it's obvious that we should be more patient (I'm talking in general terms, not about the specifics of economic conditions right now).
You can divide δ into (1) r or real risk-free rate and (2) - ηg or (beta * MRP). My subjective view is that the risk-free rate is too low and the MRP is too high. I think very few people think about their investments in the right way: "What level of return am I willing to accept to compensate me for volatility with standard deviation of x% (typically around 20% for the stock market)?". Most people subscribe to: "I'll do x% equities, y% corporate bonds and z% government bonds because that's what everybody else is doing". I personally invest 100% in equities for this reason. Furthermore, people are not flexible in how they behave (if you are familiar with the IS-LM model, I'm basically saying IS is steeply negative). In today's investment environment, everyone should be spending a lot more (including on charity) and saving a lot less, but that's not how people behave in practice. This is the reason why the real risk-free rate is so negative at the moment. Either way, the consequence is that you have to 'pay' a lot for a risk-free rate. Typically your money will grow not too different from inflation (and currently less) if you are not prepared to take any risk.
Finally, I do think value drift and diminishing marginal returns are very important points. Value drift is major simply because the world changes so fast. And in terms of diminishing marginal returns, I think the most important thing is that what we do today impacts the future. When you deworm a child, that's not just an "expense" for benefits in that year, it potentially improves their school performance and stimulates economic growth. I prefer to think of it as "investment". I think it's much more important to build out a safe framework for AI now than try doing it in 100 years' time (even with more resources).
comment by Robert_Wiblin ·
2020-09-15T14:50:59.885Z · EA(p) · GW(p)
The idea that charities should focus on spending money during recessions because of the extra benefit that provides seems wrong to me.
Using standard estimates of the fiscal multiplier during recessions — and ignoring any offsetting effects your actions have on fiscal or monetary policy — if a US charity spends an extra $1 during a recession it might raise US GDP by between $0 and $3.
If you're a charity spending $1, and just generally raising US GDP by $3 is a significant fraction of your total social impact, you must be a very ineffective organisation. I could not recommend giving to such a project.
I'd think such a gain would be swamped like other issues like investment returns, us learning about better charities in future, or the worst problems getting solved leaving us worse giving opportunities, and so on.
An exception might be if you independently thought something like GiveDirectly was the best option and wasn't going to be beaten by another option in future. Then giving money for dispersal during a recession in the recipient country might be, say, twice as good as giving it outside of recession.
There's a bunch of discussion of these issues in my interview with Phil Trammell.
Replies from: Grayden
↑ comment by Grayden ·
2020-09-16T21:54:39.164Z · EA(p) · GW(p)
Thanks for your comment.
I'm not advocating it because of the fiscal multiplier. That would be the cherry on the cake.
The first simple step is simply to say don't cut back expenditure because shrinking and regrowing an organisation is costly. Most charities (though EA ones are somewhat atypical) see their income reduced during bad times. And since most charities think in bland terms of x months of reserves, this means their expenditure fluctuates as well. This is an not efficient way to manage an organisation. In good times, build a buffer, so you can keep going during bad times. Just keeping expenditure flat would be a major step in the right direction.
Of course you can take it a step further. There is another cost argument, which is that it is cheaper to do stuff during bad times. When unemployment is high, you can get talented people more easily. So even if the benefits are the same, the benefit/cost is higher. The fact the benefits may be higher, not just that the fiscal multiplier may be higher, but that fulfillment of basic human needs may be worse, is a bonus, though it probably only applies to Global Health and Development causes. I wouldn't use a Keynesian altruism strategy simply for this.Replies from: Robert_Wiblin, Ramiro
↑ comment by Robert_Wiblin ·
2020-09-17T11:52:52.969Z · EA(p) · GW(p)
Yep that sounds good, non-profits should aim to have fairly stable expenditure over the business cycle.
I think I was thrown off your true motivation by the name 'Keynesian altruism'. It might be wise to rename it 'countercyclical' so it doesn't carry the implication that you're looking for an economic multiplier.
↑ comment by Ramiro ·
2020-09-20T22:52:53.033Z · EA(p) · GW(p)
I wonder if exchange rates volatility during global recessions (usually, the US$ dollar and the Euro rise in relation to national currencies in developing countries) would add another point, at least for charities located in the developing world.Replies from: Grayden
(Personally, since my job is very stable and opportunities for investments scarce, I have been increasing my own donations to account for my declining consumption)
↑ comment by Grayden ·
2020-09-23T09:22:36.302Z · EA(p) · GW(p)
That's a very interesting point I hadn't considered. Yes, if the expenditure is in emerging markets, your money likely goes even further during global recessions
comment by ShayBenMoshe (shaybenmoshe) ·
2020-09-14T12:48:26.061Z · EA(p) · GW(p)
Thanks for posting this, this is very interesting.
Did you by any chance try to models this? It would be interesting for example to compare different strategies and how would they work given past data.Replies from: Grayden
↑ comment by Grayden ·
2020-09-14T16:11:18.475Z · EA(p) · GW(p)
I haven't. I think the key debate is whether the theory could work in practice, rather than whether the theory holds. In terms of modelling, I think it would be hard to quantify the benefits as the variables (in particular: (1) the cost of downsizing and then re-scaling an organisation, and (2) change in marginal CPLSE with respect to a change in GDP) are inherently difficult to measure. Do you have any thoughts about how we could do it?Replies from: shaybenmoshe
↑ comment by ShayBenMoshe (shaybenmoshe) ·
2020-09-16T07:03:54.437Z · EA(p) · GW(p)
I agree that it isn't easy to quantify all of these.
Here is something you could do, which unfortunately does not take into account the changes in charities operation at different times, but is quite easy to do (all of the figures should be in real terms).
- Choose a large interval of time (say 1900 to 2020), and at each point (say every month or year), decide how much you invest vs how much you donate, according to your strategy (and others).
- Choose a model for how much money you have (for example, starting with a fixed amount, or say receiving a fixed amount every year, or receiving an amount depending on the return on investment in the previous year).
- Sum up the total money donated over the course of that interval, and calculate how money you have in the end.
Then, you can compare for different strategies the two values at the end. You can also sum the total donated and the money left, pretending to donate everything left at the end of the interval. Or you could adjust your strategies such that no money is left at the end.
comment by PeterMcCluskey ·
2020-09-13T22:47:51.013Z · EA(p) · GW(p)
- Cash sitting in a charity bank account costs money, so if you have lots of it, invest some;
But the obvious ways to invest (i.e. stocks) work poorly when combined with countercyclical spending.
Charities are normally risk-averse about investments because they have plenty of money to invest when stocks are high, but need to draw down reserves when stocks are low.
Replies from: Grayden
↑ comment by Grayden ·
2020-09-14T08:25:15.299Z · EA(p) · GW(p)
That's a good point and I don't think I was particularly clear in my post. I will have a think about whether I can rephrase in a way that keeps it concise.
I'd like to separate my response into two issues: (1) liquidity (cash vs. Treasuries) and (2) risk tolerance (Treasuries vs. stocks). On liquidity, I think it's a good idea to keep a few months of expenditure in cash to ensure you can access it in an instant. Depending on your size, you may get some interest paid by the bank, but it's very unlikely to keep pace with inflation. However, anything you don't need at short notice can be invested in risk-free assets (e.g. short-dated US Treasuries), which have a better chance at keeping pace with inflation (with the usual caveat that the benefits have to outweigh the added admin).
Risk tolerance, i.e. whether to invest in stocks (maybe even with leverage) rather than Treasuries, is another topic and lots of smart people have written previous stuff on this, e.g. here. This is where the practical difficulties I mention come in. You need to be willing for income (including potential gains and losses on investments) and expenditure to be going in opposite directions, potentially over a number of years. Certainly, if a charity has 6 months' expenditure in the bank, I wouldn't recommend putting 3 months worth in stocks. But if a charity has 10 years' expenditure in the bank, I think it needs to realise how much that is costing it. If it puts 9 years' expenditure in stocks, then with even a bad market crash, it will still have 5 years' expenditure.Replies from: PeterMcCluskey
↑ comment by PeterMcCluskey ·
2020-09-15T02:31:36.122Z · EA(p) · GW(p)
10 years worth of cash sounds pretty unusual, at least for an EA charity.
But part of my point is that when stocks are low, the charity won't have enough of a cushion to do any investing, so it won't achieve the kind of returns that you'd expect from buying stocks at a no-worse-than-random time. E.g. I'd expect that a charity that tries to buy stocks would have bought around 2000 when the S&P was around 1400, sold some of that in 2003 when the S&P was around 1100 to make up for a shortfall in donations, bought again in 2007 at 1450, then sold again in 2009 at 1100. With patterns like that, it's easy to get negative returns.
Individual investors often underperform markets for the same reason. They can avoid that by investing only what they're saving for retirement. However, charities generally shouldn't have anything equivalent to saving for retirement.
Replies from: Grayden
↑ comment by Grayden ·
2020-09-15T10:18:37.331Z · EA(p) · GW(p)
I think what you are referring to is an Anti-Nightingale. If you always sell after a market crash, you will most likely (as in mode, not mean) have poor returns, but that doesn't change the expected value from investing. The odds of a roulette wheel never change, but you can change your strategy to give you a >50% chance of coming away with a profit. My strategy will give you a >50% chance of coming away with an underperformance of the market, but will not change the underlying odds.
Another trap some people (including professional investors) fall into is "buying the dip". It feels natural to expect that when the market is low, the future expectations must be higher and it must be a good time to invest. In a perfect market (not a given!) this is not the case. In fact due to government responses (lowering the interest rate), returns should actually be lower. In very practical terms, this time last year you might have expected a 6% return from investing in the S&P 500 for one year. Right now, that 6% might be 5.5% because interest rates are lower.