GiveWell actually benchmarks their charity recommendations against direct cash transfers and will generally only recommend charities whose benefits are Nx cash for some N that I don't remember off the top of my head. I buy that lots of charities aren't effective, but some are!
That said I also think that longer term research and investment in things like infrastructure matters too and can't easily be measured as an RCT. GiveWell style giving is great and it's awesome that the evidence is so strong (and it's most of my charitable giving), but that doesn't mean other charities with less easily researched goals are bad necessarily.
Certainly charities exist that are ineffective, but there is very strong evidence that there exist charities that do enormous amounts of direct, targeted good.
givewell.org is probably the most prominent org recommended by many EAs that does and aggregates research on charitable interventions and shows with strong RCT evidence that a marginal charitable donation can save a life for between $3,000 and $5,500. This estimate has uncertainty, but there's extremely strong evidence that money to good charities like the ones GiveWell recommends massively improves people's lives.
GiveDirectly is another org that's much more straightforward - giving money directly to people in extreme poverty, with very low overheads. The evidence that that improves people's lives is very very strong (https://www.givedirectly.org/gdresearch/).
It absolutely makes sense to be concerned about "is my hypothetical charitable donation actually doing good", which is more or less a premise of the EA movement. But the answer seems to be "emphatically, yes, there are ways to donate money that do an enormous amount of good".
And I think it's interesting that it flags it as confidently AI generated. I also got a whiff of AI and I'm never sure how to take confirmation bias from AI detectors - though that said, I've gotten a whiff of AI before and had this detector say it's confidently human.
Reading this I kept waiting for the... point? I feel like the whole thing was more like saying "UDP is like a convertible because you can strap a tarp on top when it rains". Like... sure? But that tarp is going to be crappy compared to a real roof. And the idea that that tiny layer is "the best of both worlds" is frankly ridiculous to me.
I recently set up a system for myself using some cheap NFC tags and Tasker that:
1. Has a "lock" tag next to my bad that scans when I place my phone down on my night stand
This disables all the apps on my phone that I consider distracting/harmful/etc.
2. Has an "unlock" tag that I scan next to the kitchen sink where I take my morning pill (could just as easily be in the bathroom for brushing my teeth etc).
This forces me to at least get out of bed before I can use time-sucking apps. It's a small thing, but baby steps is kinda what I was looking for, and it's noticeably decreased the number of days where I laze in bed for longer than I meant to in the morning. It's been ~2 months and still going well.
And then computed the annualized percentiles of growth over every 10 and 20 year period:
Percentiles 10 Year 20 Year
0.1 2.759994509 2.955813986
0.2 2.857204716 3.026836408
0.3 2.990680184 3.126358739
0.4 3.127881625 3.147020404
0.5 3.199826901 3.269223125
0.6 3.418119435 3.360219255
0.7 3.558537072 3.523213118
0.8 4.20459087 3.876319675
0.9 5.606452092 4.488515605
1 6.820735567 4.991026738
You're definitely right that they underestimate rent growth, at least if you're assuming that you should be making conservative estimates. Plugging some of these numbers in, I don't think this changes the overall conclusion of the post, but it does change the magnitude non-trivially so I think it's very worth considering. Thank you!
The NYTimes/NerdWallet calculators implicitly account for that in their logic - they track money you gain/lose from down payment/mortgage/interest/taxes, then selling the house at the end.
On the renting side, they only assume investing the money that isn't going to down payment/monthly mortgage payments, not investing the full value of the house.
My blog post here is just giving an argument that 2 of their parameters should be updated, then showing the result of that update.
I can definitely make this more clear, but this is specifically only changing the home appreciation and asset appreciation rates that produce that $1.5M change. I used the topline numbers at the beginning of the blog post, which is the national 20th percentile for housing growth and S&P growth. Everything else (local factors like price, rent, mortgage interest, property taxes) is held equal.
> I can't see how you're accounting for the cost/value of the use of the property
That's what the rent/buy calculators are doing! It's summing up all the cash flows for owning a property (down payment, mortgage, taxes, maintenance, etc, and then crucially selling it after 30 or so years) and for renting a property (rent, and investment income from money that would have otherwise went to down payment/mortgage), and telling you how the results differ.
All I'm doing is tweaking 2 of the parameters of these calculators: The rate the home appreciates in value, and the rate cash investments appreciate in value. Everything else stays the same.
Let's say I mistype and don't double the first "c", but otherwise type entirely correctly.
> acording to its archive...
This would be counted as having everything wrong except the first 2 characters, which doesn't feel like a good reflection of my accuracy.
I know this is a hard problem because I don't think there's any simple guaranteed way to re-align the string to account for a possible deletion or insertion, particularly if there are more mistakes in the following text, but finding and using some sort of accuracy-maximizing alignment would be great to have.
This is very neat. One piece of feedback and a gripe I have with a lot of these is that missed or extra characters throw off the entire next sequence and essentially require backing up to deal with them, as opposed to wrong characters which are fine to just be mistakes you move on from. It'd be great to have some detection for when the user is continuing that re-aligns their string.
I've been basing one of the biggest financial decisions in my life - whether to buy a house - in large part on NYT/NerdWallet Rent-Buy calculators. But when I dig in, it seems that the model is both extremely sensitive to home/S&P500 growth assumptions, and that their defaults aren't well thought through.
This site is my attempt to organize my thoughts on what reasonable defaults should be, and provides an interactive tool to explore housing and S&P500 growth historical growth rates.
That said I also think that longer term research and investment in things like infrastructure matters too and can't easily be measured as an RCT. GiveWell style giving is great and it's awesome that the evidence is so strong (and it's most of my charitable giving), but that doesn't mean other charities with less easily researched goals are bad necessarily.