Thursday, March 21, 2019


Torn regarding emoji’s. On one hand, they are a crutch for emotion communication in media where we don’t take the time to craft words that do the job. On the other, they serve as important tools for reducing ambiguity in the low bandwidth medium that is text conversation.

Thursday, February 28, 2019

Recognition is prejudice

Recognition is prejudice. The act of categorization, of labelling, the very capability that allows us (or any mind) to understand anything in the world, is inherently exclusionary. Who/what can or can’t be what. Limiting of possibilities. Thinking sideways allows us to fight against this.

Saturday, February 2, 2019

Song by Song

Song by Song is a podcast about Tom Waits. At least one episode for every song, starting from the beginning.

hello Martin and Sam!

I'm a bit behind, so this comment is in regards to an episode from a few months back. 

I appreciated both your and Vera's thoughts on Ruby's Arms, Big Time [157], particularly in regard to the duality between the toxic masculinity embodied in the story and the quality of the song itself. 

For me, that same feeling of contrast (almost cognitive dissonance) between unfortunate lyrics and beautiful music occurs often with Springsteen, e.g. Born To Run (as Sam says, those 80's singer-songwriters!). The critical side of me reacts to the "warped idea of romantic hero figures" and escapism, but at the same time, the music arouses such feelings of hope and freedom and action!

When I think too hard about this, my brain starts making dubious connections with Zen and existential struggle...

It's tempting and self-flattering to imagine I'm somehow evolved - "an artist", or in Fitzgerald's words "a first-rate intelligence" lol - to be able to consider both facets as true simultaneously. But maybe the more sane approach is to listen to music that doesn't have a compromise. I don't know. My favorite recently is tune-yards. 

Your show is wonderful!


Wednesday, January 23, 2019

Arrest Documentation

Usually, I'm perfectly OK with police action - pulling over traffic violators, deterring violence, and generally enforcing the laws designed to prevent society from devolving into a commons tragedy. But I notice that the attention of peace officers often falls more heavily, in ways that seem unfair to me, on those less privileged.

I've been struggling for the last few years with my instinct to just continue on my way when I see some action (police or otherwise) that doesn't sit right. It's easy to ignore other people's problems, and likely, if I knew all the facts, I would agree with whatever the officer was doing. But often (and this I acknowledge may be an anti-authoritarian bias of mine), I feel like the officer is overstepping their authority to stop and question citizens.

Examples that I have encountered where the opportunity to witness has presented itself and I have continued on my way:
 - cycling along Foothill Expy in the morning and passing a Latino man in a small delivery truck pulled over by an officer
 - cycling along Sunnyvale-Saratoga in the morning and passing a man on foot at a corner being questioned by an officer
 - driving along El Camino and passing a disheveled Black man at a bus stop being questioned by multiple officers

By walking by, I am tacitly agreeing to be a member of the society that hires police officers to take this particular action. I think it is too easy to assume we know what is going on there and we are OK with it. That police officer is an embodiment of myself deciding to use force.

I think we've given police too much authority to interfere in the activities of citizens that they think are weird or unusual (e.g. not having a shirt on, poking around in bushes, sitting on a sidewalk, yelling). Too broad of an authority to decide what constitutes a threat to public safety. To counter that, I feel motivated to use my privilege as a tall, white, wealthy, intelligent male to witness, document, and question police actions I encounter that don't feel quite right.

I would be thankful for a fellow citizen watching and documenting, if I were being arrested or questioned by the enforcers of the law.

Resolving this question of whether I should or shouldn't think twice, document, or even interfere, will only get better if I actually practice.

I live in Berkeley now, and frequently travel over to San Francisco, so compared to living in Sunnyvale suburbs for the last 5 years, I'm exposed to more frequent police action.

Here's a video I took on Saturday, January 12, 2019 at 4:37pm at the intersection of 5th and Bryant (37.7783, -122.3998) in San Francisco. Please contact me if you have concerns about this video being public.

0:50 - the man is injected with something by paramedic (perhaps a sedative as he does not seem to verbalize as much for the rest of the video). verbal discomfort, shock, and anger.
1:10 - lots of "I'm sorry, I'm sorry," as if in pain
7:40 - lifted onto a stretcher
8:40 - interview with bystander witness. "That should have never happened".

I was curious to follow up on this, so I went looking for some police reports. Conveniently, SF has a public database of police incident reports that is updated every few days. There is a visualization tool online, but I found it easier to use the map already provided to narrow in on the incident by location and date/time. The database entries do not have much detail, but I was able to find the Incident Numbers for the incidents most likely to be the one I saw. The time listed was about an hour off of what I observed.

To get additional info about the incident in order to label my video, I filled out a couple Request for Incident Report forms and emailed them to the SFPD. They got back to me in just a few days with the full report as written by the arresting officer.

The Incident Report (# 190029689) for the arrest of Mr. Foster is linked here:

Interesting to note the difference in narratives between the officer's report and the first-hand witness I interview in my video. The officer reports that he was kicked by Mr. Foster, whereas the witness reports that Mr. Foster was just sitting on the curb. The officer reports no complaints of pain, whereas the video clearly documents discomfort.

I feel that I still don't have enough information to pass judgement on any actions in this video. Except for the fact that I strongly do not approve of using an injection without consent.

But I feel that taking the video was worthwhile and I'll probably do it again.

Thursday, September 6, 2018

Primary Production

How fast could plants in a transparent box starve themselves of CO2?

In the spring, we sprout lots of little tomato and squash plants in a plastic tray. Maybe 60 sprouts in a 2 square foot tray. Initially, we keep the tray indoors, but when the sprouts are a little bigger, we leave it outside so they get more sunlight. Sometimes we put a transparent plastic lid over them to make a little greenhouse to keep them warm on cold days. The lid makes a weak seal around the tray, and I wondered if could be starving the plants of CO2.

As a first order of magnitude check, we would expect CO2 consumption to be a modest fraction (maybe 1/3?) of the total existing mass of the plants over the course of maybe a week because the CO2 is being consumed to create plant material. It seems reasonable that sprouts should grow by 1/3rd every week. The plants in one mini-greenhouse box probably have a mass of (1g x 60) = 60g. So very roughly we would expect CO2 consumption to be somewhere around 20g per week, or about 2g per day.

Ok, lets get more precise. What volume of CO2 is consumed (and O2 released) by a certain mass of plants per day?

If we can assume the box is producing at an average rate somewhere between tundra and grassland (the sprouting tray looks like a little fell field, but it is probably growing faster than one), we can say its mean NPP is about 300 g/(m^2*yr). If our box is 0.2m^2, one day should net about 0.2g in primary production (300g * 0.2m^2 / 365days).

I'm going to use glucose as a carbohydrate representative substance for all of primary production. In photosynthesis, every 6mol of CO2 (44g/mol) yields 1mol of sugar C6H12O6 (180g/mol). So 0.2g (0.001mol) in primary production of sugar requires about 0.3g (0.006mol) of CO2.

Ok, so the plants in the box consume about 0.3g of CO2 per day. Pretty close to our initial estimate of 2g per day, as far as rough estimates go.

Now, how much CO2 is in the box?

Partial pressure of CO2 in dry air at sea level (760torr) is about 0.3torr. So 0.04%. PV=nRT, so for every 100 mol of air, 0.04 mol of CO2. Volume of air in the box is about 20L, so at 1kg/m^3 or 1g/L, we have 20g of air. Air molar mass is 29g/mol (mostly N2), so we have about 1 mol of air in the box, and maybe 0.0006mol of CO2. Which at 44g/mol is 0.026g.

Ok, so there is only about 0.03 grams of CO2 in the box.

Wow, it looks like the CO2 in the box gets used up pretty quickly. Maybe in about 1 hour.

What are the uncertainties in my estimates?
- area or volume of air in the box - probably correct within a factor of 1.5
- mass of plants in the box - probably correct within a factor of 3
- NPP assumption - desert is 90, grassland is 600 g/(m^2*yr) - so probably correct within a factor of 2 or 3.
- Assumption of glucose as a representative substance for all of primary production - the actual substance that gets made in the Calvin cycle is glyceraldehyde 3-phosphate (G3P). G3P is then transformed into sugars like glucose, which in turn are consumed to build the plant itself.

With these uncertainties, my worst case error is factor of 11 (3x1.5x2.5), so worst case bounds are from 5min to 11hours. The root sum of squares (RSS) error is a factor of 4, so the bounds are more likely from 15min to 4hours.

Make sure greenhouses get adequate ventilation!

Some supporting evidence:
“Many greenhouse growers have starved their plants for carbon dioxide in the attempt to conserve heat by limiting or eliminating air exchange in the greenhouse. Up to two full air exchanges an hour have been recommended for greenhouses to keep the plants and the equipment functioning properly.”

It's annoyingly difficult to search for anything related to greenhouses and CO2 and starving because climate skeptic articles overwhelmingly dominate the results.

Wednesday, May 2, 2018

Reserve Parachute Deployment

I wrote up this incident report after I deployed my reserve parachute while paragliding in the Owens Valley on 4/24/18.


I am a P4 pilot with ~200 hours since receiving my P2 in January 2016.
I have previously flown the Owens Valley for ~3hr in October 2017 and ~3hr in March 2017.
We arrived in Bishop Sunday evening after flying 3 days at Tollhouse. Light winds in the forecast for the first half of the week and reports from locals promised good flying.

Flew 80km (11:30-3:30pm) from Paiute with my friend, R, on Monday. We flew with three other pilots for the upwind, beginning leg of the flight. Conditions felt well within abilities (less than three minor <20% tip collapses, one minor <half chord frontal, <10mph SW winds). Flying without an oxygen system. Felt minor hypoxia symptoms (hint of headache) above 13kft during first half of flight. Countered with deep breathing and leaving the thermal. No identifiable symptoms above 14kft for the remainder of the flight.

I launched from Paiute on Tuesday at 12:45pm with R and two other pilots. Flew north, reaching Montgomery Peak at 2:45. Observed and reported wind were similar to Monday. No collapses or frontals. No identifiable hypoxia symptoms. Turned south after agreeing to attempt an upwind leg back to launch.

  • 777 Rook2 MS. High EN-B. Flat aspect ratio 5.6.
    • I’ve been flying this wing for about 14 months (~100 hours). No SIV on it. Last SIV (12 tows over 4 days) was 1.5 years ago on my previous wing - Hook3 (flat aspect ratio 5.4). Recently thoroughly cleaned to remove crud from tips.
  • Ozium harness. Flying for ~30 hours.
  • U-Turn Protect 38m^2 round PDA. Repacked when installed in Ozium.
  • SupAir Pilot helmet


What I remember:

Flying S in smooth air into a <15kph SW headwind on less than half bar, I was looking to push forward onto the windward face of Dubois where I had previously caught a climb. I felt that I was in a comfortable position, especially having seen another pilot cruising along the same ridge below me (half my AGL) only 4min beforehand.

Flying along in smooth air, with hands wrapped around brake toggles and touching C risers, I felt an instantaneous loss of pressure indicating a large frontal collapse. I released bar, and input on both brakes to minimize the frontal, but it was clearly way too late or insufficient. Hands up immediately after realizing I hadn’t stopped it. As it surged, I checked (don’t remember how much) and looked up and saw the wing for the first time: double cravat, deeper on left side, starting to auto-rotate CCW. Facing SE. With the depth of the cravats, my mind immediately made the decision that I would not be able to fix it with my perceived altitude and I reached down for the reserve, letting go right brake. I pulled and threw it as I spun through facing W. Found the handle quickly even in mittens and it came out with a strong tug. I pulled it out and let go in one motion. I failed to check to see where wing was. I felt the tug of it opening within the next few seconds. Realized at some point here that my helmet had come off (buckle released). I started hauling in hand over hand on left brake line to disable the wing. Looked down, saw my descent rate, and got my feet out of harness for PLF. I hit the ground facing W, most of energy absorbed through legs and butt cushion before rolling back and knocking my head. Minor flash but never lost clear vision.
(Landed 2:59pm at 13,400ft on nearly flat, packed scree).

My wing was near me and mostly disabled and deflated already. I felt a tug from the reserve, which I then jumped up to disable by pulling on the apex line. Pulled everything into a pile and sat on it. I pulled out my radio, removed the headset plug, and radioed landed safe under reserve, probably within 2min of landing. Acknowledged by R who was about 2km away. I sat for 5min relaxing to stop shivers and ate a cookie. Pulled out my phone which luckily had service and checked the Mount Dubois descent route which someone had posted to Mountain Project. I did not want to descend the steeper, taller, and more unstable CA side. I texted my intention to walk down the NV side to the three other pilots (at 3:09). Relaxed for a little longer and then started packing up. Sorting out the wing for the concertina bag, I had to pull lines out from a few cells away from midspan. I responded LOK (at 3:39) to a chat from friend who was watching my inReach tracklog. I considered looking for my sunglasses and deployment bag, but decided getting down quickly was more important. Started walking at ~4pm. The hiking and route-finding were straightforward. Some Class 2 scrambling along the ridge. Continued to communicate with my retrieve via cell and inReach as I descended. Reached Davis Meadows Ln (Middle Creek) trailhead (8,400ft) at 740pm.


Speaking to top local pilots afterwards, the likely cause of the sudden frontal was encountering downdraft turbulence on the lee side of a thermal. Need to be extra cautious pushing upwind on bar in sharp-edged conditions. Altitude over ground makes terrain rotor very unlikely as a cause.

I was surprised to find that my tracklog showed 1500ft AGL - I thought I was quite a bit lower. This means both that my AGL estimation is poor, and that my SIV confidence is lower than it should be. 1500ft should be sufficient to at least work on the wing for a little bit. On the other hand, I felt after landing that I had had just the right amount of time to throw, assess descent, disable the wing, and get into PLF before landing. It went by much faster than the ~60s it looks like I had between the blow-up and landing. Overall, I am very happy that my mind made the snap judgement to throw immediately.

My helmet unbuckled itself at some point during the descent. It is unclear to me whether this was caused by the wing risers or the reserve risers or something else passing across the buckle and unlatching it. I believe this is a problem with the helmet buckle design. This is a strong accusation to make, so I will elaborate. I know that the helmet did not get pulled forward off my head with the buckle still closed both because I saw that the buckle was open after landing, and because testing afterwards demonstrated that this is impossible to do without damage to the trachea, which I did not have. The helmet was still with me after landing because it was still attached by the headset cable. I am doubly certain that my helmet was buckled at launch both because of my usual preflight checks (which have never missed my helmet), and because my headset mic does not sit properly at my lips unless the buckle is secured (and I used the radio several times during the flight). I will be contacting the manufacturer and getting a different helmet.

My reserve is large and gave me a nicely slow descent at sea level during SIV. At 13kft, the lower air density (0.8 vs 1.2 kg/m^3) means a descent velocity increased by ~20% (rho proportional to v^2).

Both local pilots recommended afterwards waiting up at altitude until late evening and taking a sled ride down to the valley floor. This would certainly have been possible from the gently sloped terrain where I landed and I should have at least considered this option, but I think I made the right call descending on foot because I was quite confident that I could make it out before sunset with my degree of alpine experience.

Wing inspection/untangling the next day revealed that I had managed to pull in 8ft of left brake and that the wing was twisted 5x with respect to my harness. I also had to remove some deep line-overs.

It is very possible that I could have landed somewhere less perfect (steeper terrain) or been injured on landing. I had enough food and water and shelter to last the night, though I should have been carrying both a SAM splint and stronger painkillers in order to be able to move to a more sheltered or stable location if necessary. It was very reassuring to find out that friends were watching my inReach tracklog in this case. This would have been critical had I not been able to communicate my condition myself.

Things I would like to have done differently:
  • Checked the surge harder and earlier, especially with right brake as I saw it rotating left.
  • Not looking at my wing enough - I think I should have been looking back at it even earlier (while it was behind me).
  • More SIV. More confidence in clearing poor wing configurations.
  • I was using the wrong phone number to text my plan to my retrieve friend. Luckily the other comms channels worked as intended so this was not a problem.
  • Considering that I did not check which direction I threw, I think I got fairly lucky with my reserve toss. I should have taken the time to look. I am also strongly considering flying with 2 reserves in the future, especially in strong air.
  • I do not think hypoxia was a strong contributing factor in this accident based on the simple cognitive monitoring I was doing. But I think flying at high altitude requires even greater vigilance of the mental state, so I plan to use better personal checks for hypoxia like more complex mental math on a regular, perhaps alarmed, basis.


In order to fly the Owens, I need to be able to fix what I had with the altitude I had. I made the right decision to throw in this case. But I need a lot more practice getting out of bad configurations. I’ve scheduled more SIV.

Tuesday, March 6, 2018

VR sword haptics, part 2

Another option for a VR haptics sword is to use jets mounted to the tip of the sword create reaction torques. In spacecraft design, these are called cold gas thrusters. They work on the same principle as any jet engine or rocket - mass ejected from a system causes an equal and opposite momentum change in the system - except there is no combustion. A minimum of three nozzles (or maybe just one plus a direction actuator) mounted equally spaced in a plane normal to the sword would be able to generate torques about any combination of two axes.

What kind of pressures and flow rates would be required to generate the required thrusts? For an ideal nozzle, the imparted thrust equals the mass flow rate of the jet times the velocity of the jet:

F = ṁv

For supersonic flow from a diverging nozzle, jet exit velocity and mass flow rate are dependent on the inlet pressure, the nozzle throat diameter, and the type of gas. Hot, low density gases have better performance, but compressed air at room temperature is certainly easiest to use for this application. For a 200psi (14atm) inlet pressure, a 4mm diameter nozzle throat will allow the flow rate required to produce about 20N of thrust, which at a 1m lever arm of course means about 20N*m felt at the handle.

What would a system designed to generate these flows and pressures look like? Ideally, very little mass would be attached to the sword, especially the tip, so that swinging it around doesn’t take too much effort, and so that the jets have less momentum to counter. A proportional control (throttleable) air valve rated to the required pressure and flow rate costs maybe a few hundred dollars and weighs maybe 0.5kg, if well designed. A few of these would be too heavy to place at the tip, but a pipe running from handle to tip would be able to conduct the air with minimal pressure drop and keep the jet response time on the order of 20ms.

A tank large enough to hold enough pressurized air for even a minute of gameplay would be too large and heavy to be mounted to the handle, so the sword would have to be connected to an external supply hose. For continuous operation, a large stationary compressor and tank would be required - think large 6ft (2m) tall machine shop setups.

So, like the gyroscope approach, a reaction jet haptic sword would probably be too expensive for the consumer market. However, a reaction jet haptic sword would be significantly simpler to design and prototype, though it has the disadvantages of being tethered and very loud.

Animal Suffering

I’ve never given vegetarianism or veganism serious consideration. Nor seriously engaged with an adherent of either about their motivations. It seemed commonplace enough to be unremarkable - a respectable preference that wasn’t for me. What little thought I’d given to the idea could be boiled down to the argument from the natural order - many animals are carnivores, so why should we restrict ourselves? While exploring the effective altruism movement last year, I came across the first chapter of Singer’s Animal Liberation [1] where he lays out the core of the philosophical argument for minimizing animal suffering and was recently intrigued enough to read the rest of the book and engage fully with the ideas.

The core of the argument asks us to acknowledge that, just as there is no difference between the human races when it comes to being human, there is no difference between humans and other animals when it comes to the capacity to suffer. Just as being human entitles one’s interests to equal consideration among all humans, so having the capacity to suffer entitles a being’s interest in not suffering to equal consideration among beings. “Pain felt by an animal should be given the same weight as the same amount of pain felt by a human.” If we acknowledge this, then we cannot morally cause that pain in an animal for the purpose of a mere preference in taste.

This argument took some effort to internalize, but I now accept its validity. The text addressed well the various conceivable misunderstandings or objections - for example, equal consideration does not imply equal treatment or rights, plants do not suffer in the same way that we do, modern factory farm conditions indeed inflict considerable pain - as well as my naive rationalization from nature. Unlike the carnivores, humans have the capacity to think morally. With our intelligence comes the responsibility to consider such ethical questions as minimizing suffering.

“I hope that anyone who has read this far will recognize the moral necessity of refusing to buy or eat the flesh or other products of animals who have been reared in modern factory farm conditions. This is the clearest case of all, the absolute minimum that anyone with the capacity to look beyond considerations of narrow self-interest should be able to accept.” [1]

Singer doesn’t attempt to make any strong arguments about the ethics of pain-free killing because killing animals is more ethically ambiguous than inflicting suffering. For me, it comes down to a question about the difference between human and animal self-awareness. What is the meaning of life for a cow? Does it have goals beyond the next meal or meaningful relationships? I like this thought experiment: is there a difference between three short pain-free cow lives and one pain-free cow life three times as long? “In the absence of some form of mental continuity, it is not easy to explain why the loss to the animal killed is not … made good by the creation of a new animal who will lead an equally pleasant life” [1]. I currently tend to agree, but not without considerable uncertainty. Although Singer warns that “if we are prepared to take the life of another being merely in order to satisfy our taste for a particular type of food, then that being is no more than a means to our end” [1], he also “can respect conscientious people who take care to eat only meat that comes from [animals with pain-free lives].”

The problem is that achieving this practically is difficult. Even “humanely certified” dairies (separation at birth, bull calves sold to slaughter, continuous pregnancy and milking) or “free-range” egg farms (aggressive laying schedules and feeding, beak trimming, male chicks sold to slaughter) or “painless” slaughterhouses (crowded feedlots, mistakes with the captive bolt gun) are not without significant suffering [2]. The pressures of the market encourage treating the animals as machines, especially when the consumer is uninterested in the details.

I find it is important to keep in mind the utilitarian equivalence of even seemingly minor suffering to pain we feel as humans. We can imagine the physically unpleasant moments of being farmed for flesh and wonder if we would inflict that on ourselves for the pleasure of eating the animal product.

Flesh and animal products truly free from inflicted pain might be ten to twenty times more costly to raise than factory farmed flesh. Very small herds, caring husbandry, uneconomical resource utilization, careful killing. An expensive delicacy. At what point is the taste worth the effort? Veganism is the obviously unambiguous moral choice, but I do not think I have the willpower for that just yet.

Because pain-free products aren’t readily available for sale, I must rely on some kind of metric to help me make choices to reduce the suffering I sponsor. Sufficient strides have been made at the vanguard of humane farming of animals that some products are available for which the embodied suffering is tied mostly to processes surrounding slaughter. For example, pasture-raised hens can have fairly pleasant lives until the hours before slaughter where they are crated up, shipped, racked upside down, and conveyored off to stunning. For this reason, animals that produce a lot of product over their lifetime, like dairy cows, rank very high (better) on a calorie per suffering metric [3]. If we assume equal suffering per death and assign equal value to the suffering of different animals, salmon and chicken flesh are roughly equivalent in calorie per suffering, eggs rank ~10x higher (better) than that, beef ranks ~10x higher than eggs, and dairy ranks ~20x higher than beef. Based on this metric, if one must consume animal products, one should should stick with dairy. And if flesh must be occasionally tasted, then beef is least harmful.

However, there is another metric I consider important - greenhouse gas emissions. Climate change is an immediate crisis that will cause great suffering to beings and damage to ecosystems on a scale much larger than animal farming (irrecoverable diversity loss, many more beings affected by 2-4 orders of magnitude [4]). Though harder to quantitatively compare, minimizing its impact by reducing greenhouse gas emissions is just as important as reducing farm suffering, if not much more. Studies of the relative environmental impacts of animal farming show that beef is ~10x worse than all other animal products [5]. This penalty knocks beef down to rough equivalence with eggs in the ranking above. 

My diet has changed considerably over the last ten years. From what was approximately the average American proportions, I estimate that my flesh consumption has decreased by 5-10x. Dairy and eggs have probably stayed about the same. Vegetable consumption has increased maybe 2-4x. These changes have been motivated solely by exposure to new foods, increased consideration of my physiological health, and consciousness of the impact that my choices have on greenhouse gas emissions. But now, the acknowledgement of the equivalence of the capacity for suffering among beings demands further change.

As I rarely eat chicken or beef already, it is easy for me to eliminate terrestrial flesh entirely. Eggs and fish consumption is next for reduction. Finally, I plan to slowly work on reducing dairy intake. This is possibly my brain forecasting and generating loopholes, but I think it is important to remember that because being perfect often distracts from other valuable things, drastically reducing is often almost as good as entirely eliminating.

Trying out this new perspective feels disconcerting - a disturbance in the norms I’ve been comfortable with. I catch glimpses of what people might have felt when deciding to treat slaves as humans. It helps me appreciate the struggle this idea continues to have in gaining credence in our culture. The core of the idea also seems to be hidden behind so many contrasting and confusing prescriptions, that the effort to examine it and form a considered opinion is a deterrent. It seems especially important to lay out clear cases and engage in debate.

I’d love to hear and discuss other perspectives and rationales.

[1] Singer. Animal Liberation. 1975. Chapter 1

Individual Real Estate Investment

I find myself confused at the temptation people have to try buying real estate as an investment. Typically, their plan is to do a bit of research into “hot” areas or engage with a property manager in order to select a property that they imagine has some special potential for either appreciating in value or generating large rents. Then they plan to buy the property, hold and maintain it for some time while collecting rent, and finally sell it for some total gain they expect to be better than if they had invested their money elsewhere. I don’t understand the attraction. And I’m confident the expected returns are poor relative to other investments. I see buying individual investment properties as analogous to buying individual equities on the stock market.

It is uncontroversial that stock picking is for suckers. Putting bets on what will win is fun and exciting, as I am told gambling is. It indulges the human brain’s overconfident capability for predicting the world. But it doesn’t pay off. The vast majority of actively managed funds are outperformed by broad market indices [1]. That is, an investment in a total stock market index fund almost always generates greater returns than careful picks by the investment industry’s top experts. An amateur trying to pick individual winners based on what she read in a few smart sounding articles cannot expect to perform better.

From an expected value perspective, an individual picking a particular real estate property is just as misguided as an individual picking a particular stock in order to “beat the market”.

An index fund provides an inexpensive, diversified investment option. It is designed to eliminate the high risk of picking individual stocks while still capturing the returns from the growth of a market as a whole. The real estate market has investment products designed for the same purpose - real estate investment trust (REIT) indices. A total market REIT index captures the returns from the growth of the real estate market as a whole. It is designed to eliminate the high risk of investing in individual properties.

If index funds are shown to be the best long-term holding of stocks, then a total market REIT is the best long-term holding of real estate. The only reason someone would gamble on individual properties is because they mistakenly expect to beat the total market index.

I spent some time trying to find a study on the real estate market similar to the report above on the stock market. A study that investigates whether or not the vast majority of real estate portfolios containing individual properties are outperformed by a total market REIT index. But either it hasn’t been investigated or I couldn’t find it. So we will have to settle for the argument from structural similarity.

A total stock market index fund returns the average performance of all the companies in the market, weighted by market capitalization. It follows that, over some period, there is a 50% probability that a random selection of stocks will outperform the index fund, and a 50% probability that a random selection will underperform. Despite this prior probability, greater than 90% of managed funds (selections of stocks chosen not randomly, but by intelligence) underperform the index fund. This is due to adjustments to the portfolios that hurt the fund (i.e. picking a promising replacement stock that then drops in value) as well as fees and transaction costs.

A total market REIT index returns the weighted average performance of all the real estate investment companies in the market. Just as in the previous case, a random selection of real estate properties has at best a 50% probability of outperforming the total market index. Equivalently, a real estate investor starts with a 50-50 chance of underperforming an index. Combined with the lesson from the stock market, where the attempt to try to land on the outperforming side of that 50-50 ends up doing the opposite, it appears that investing in individual real estate properties is unwise, as the investor can choose to take the guaranteed average through a REIT and walk away.

As an aside, I was curious to look at historical data. This paper [2] finds that global long-term (1870-2015) mean returns from real estate (including both capital gains and rents) are similar to returns from equities (stocks). The paper also breaks down the results by nation and “medium-term” - for the USA over the last 70 and 40 years, total market returns to stocks have been a couple percentage points higher than those to real estate. But this is only tangentially relevant here.

An objection might be raised that because a total market REIT invests only in real estate investment companies on the public market, it does not capture returns on the housing market as a whole. While this is true, the returns on a total market residential REIT should be comparable to returns on housing [2].

The most common objection to this argument for placing individual stocks and properties in the same risk/reward class is that the motivated investor can discover some type of special information about a property (ostensibly because it is a tangible thing that can be plainly inspected and evaluated) that is much more difficult or disguised in the case of a stock. Perhaps they have noticed a trend in housing prices in the area, or an office park is about to be constructed nearby, or they heard there may be a treasure chest buried in the yard.

The implicit assumption is that the situation is comprehensible. That the investor has a sufficiently accurate model of the adjacent world - of all the major factors that might significantly affect the value of the property - that she can invest with some confidence. But this is classic overconfidence effect in action - an underestimation of the complexity, the randomness, the noise in the world. Maybe that upwards trend was actually an overvalued bubble, or maybe the increased traffic around the office park makes the neighborhood less desirable for families. The same cognitive biases that cause stock pickers to fail apply to real estate investors.

Full time fund managers have access to as much, if not more, publicly available information about the companies they evaluate as a real estate investor has about economic and social forces that govern housing values. There is no such thing as a simplified market playground or some kind of home court local advantage.

Even in areas with booming markets, there is still large local variability. Say an area happens to experience growth rate that exceeds the market average by 2% (something that roughly represents what the SF Bay Area has achieved over the last 25 years). With a relatively large standard deviation in annual property value changes of 9% (typical for the area) [3], only about 58% of properties beat the market average and 42% underperform, assuming a gaussian distribution in values. This same standard deviation means a 16% chance of underperforming the market by greater than 7%. Underperforming the market index doesn’t necessarily mean that the property is losing value, just that an investor could be seeing greater returns if she had taken the easier path of investing in a total market REIT index.

What about the advantages of leverage that are easily accessible in real estate through a mortgage? The potential returns on a leveraged investment are indeed higher, but the potential losses are equally large. Leverage merely magnifies the position - both risk and reward. It doesn’t affect the conclusions above. And besides, the companies held by a REIT hold plenty of leveraged positions on their books already, so there is nothing special about being able to hold an individual leveraged position.

I’ve been assuming the goal of an investment is to maximize returns. The other reason that someone might choose to gamble on individual properties or stocks is because it is fun. Index funds are boring. Some people enjoy the effort of research and the thrill of a bet. But it should be clear that they are spending money for that enjoyment. Just like I spend lots of money on my paragliding hobby.

[1] S&P Global. SPIVA® U.S. Scorecard. 2017.

Thursday, March 30, 2017


A nice interactive map displaying data from HOLC "security maps" of the 1930s.

It should not be surprising, but it is illuminating to read such blatant discrimination on seemingly mundane government reports.

"DETRIMENTAL INFLUENCES: Infiltration of colored residents. There are now about twelve families scattered over the area indicated"

"Infiltration of undesirables/dark Portuguese/Oriental store-keepers/Orientals: serious threat/subversive races exist"