What Do Commercial Donor Advised Funds Have in Common with Donald Trump’s Tax Returns?

Both Have Something Someone Doesn’t Want You to Know

Payout rates are a key battleground in the policy debate over donor advised funds. Recall again that every itemizer who contributes to a charity gets a charitable contribution deduction (subject to certain caps and other technical limits), no matter when the money is ultimately spent. So, too, every decedent’s estate that deeds money to charity fully escapes the estate tax on bequeathed funds, even if those funds aren’t spent for generations. Critics, like me, and Roger Colinvaux, and Ray Madoff, and lots of others, argue that this timing mismatch is troubling. Some DAF supporters — including Daniel in his post yesterday — argue in response that DAFs actually pay out pretty fast, so what’s the big deal?

These defenders have pointed time and again to DAF average (usually mean) payout rates, but mean payout is a meaningless number. For instance, commercial DAF sponsors say that DAF accounts pay almost 20% of their balance per year, on average. So what? Policies to encourage minimum payout won’t affect those who already want to pay more.

The question is: how much money is there in accounts that currently pay very little? How many dollars are locked up indefinitely? As DAF sponsors know well, that is the key question, and not one sponsor has ever voluntarily released data that would shed any light on it. This is the same dynamic as Donald Trump’s tax returns. Sponsors know that we will draw negative inferences from their refusal to release detailed payout information. Yet they persist in hiding it anyway. How bad must it actually be?

Here’s some suggestive evidence. At a conference Ray & I organized in 2015, the IRS presented data summarizing payout rates aggregated at the DAF sponsor level (remember that a single sponsor may hold one or many DAF accounts). Here’s what that distribution looked like:

As this graph shows, although the average DAF payout across all sponsors was quite high — a quarter of sponsors spent more than 23% per year — there were a large chunk of organizations that were spending essentially nothing. If the distribution of individual accounts within commercial DAF firms is similar, that is a very, very different picture than the rosy “DAFs pay 20%.”

You read Whatever Source Derived, so you must be a sensible person. A little reflection will quickly convince you that, even if we thought the industry aggregate payout were important, the mean value isn’t very meaningful. For instance, some people use their DAF account as an accounting device or as a tool for processing illiquid donations. These folks tend to spend every dollar in the fund each year. Obviously, that considerably inflates the average payout rate. When Bill Gates walks into a bar…

Also, there’s a question of how representative recent payout rates have been. At the moment we’re still at the birth of the DAF as a meaningful institution. Among private foundations, spending is vastly higher in the early years of the organization. I made this graph:

So, spending is more than twice as fast at organizations that have received funds recently. It’s likely spending at older organizations would have dropped even farther, but private foundations have to spend at least 5% of their investment assets annually. Hey, for older organizations, the average bumps right along that floor.

The graph also helps to rebut another point DAF supporters (again, including Daniel) often make, which is that imposing spending minimums could in effect create spending maximums. Maybe floors serve as a mental anchor to reorient the spending of more active spenders? If so, we sure don’t see that in the graph above. Here’s another one.

This is a graphical summary of a natural experiment the U.S. conducted on the effects of payout minimums on spending by organizations that weren’t directly affected by the minimum. As you can see, I’ve graphed average payouts among private foundations with living donors and those without (based on data assembled by Ben Marx of the University of Illinois). The vertical lines represent two policy changes.

Between 1969 and 1976, the minimum required payout was six percent. From 1976 to 1981, it was the greater of five percent or the foundation’s net investment income. After 1981, it settled into the modern five percent floor. If the “ceiling” hypothesis were correct, then firms that historically were spending more than the floor should have reduced their spending when the floor dropped. While predictions about 1976 are ambiguous, since they depend on firm’s investment returns, the 1981 change unambiguously dropped the cap for at least some firms.

Even a casual inspection of the figure suggests that there is little evidence the 1981 change reduced spending among the group of high-spending firms. As expected, the group of low-spending firms trends downwards a bit after 1981, consistent with the theory that these firms aim to spend as little as possible.

My colleague Ray Madoff and I further conduct regression analysis in which we formally test the impact of 1981 on firms that we calculate were spending above the floor at that time (and in which we can control for factors like the stock market and the obvious noisiness of these data). These estimates are necessarily imperfect because we lack firm-level investment return data, and so have no accurate measure of the floor facing firms before 1981. We can, however, compare the effects of 1981 on firms with living donors against those without. On average, 1981 increased the payout ratio among living-donor firms by about 1.5 percentage points relative to deceased-donor firms, and this result is highly statistically significant.[1] To emphasize, this was a year in which the floor unambiguously declined. If payout floors ever serve as ceilings, we see no evidence of that in our data.

So, long story short, take what you hear about DAF payout rates with a grain of salt. Essentially all the data we have are reported by an industry that profits more the slower their customers spend.

[1] Full results of these regressions are available on request.