mindstalk: (Default)
I like doing botecs or Fermi estimates, and I also like doing them in reverse, framing a number I already know. I'll be doing the latter here.

Say I want to estimate the GDP of the USSR, not trusting their numbers. I'll posit some facts: population of 290 million, rounded to 300 million (I assume they're less likely to lie about numbers of people.) Subsistence agriculture GDP/capita of $500/year. Modern US GDP $50,000/capita -- but we're talking about 1991, 30 years ago. I know US productivity growth has been meh, so let's say the US was $30,000 back then.

So, Soviet GDP will be a GDP/capita (equivalent) estimate times population. From childhood reading I think I also know something about the Soviet lifestyle and economy: concrete apartment buildings with steam heat and electricity, supermarkets, subways, cars, aircraft carriers. Also expenses like ICBMs and a space program. Yeah, they often had the inferior version of things, but an ugly clunky concrete apartment is still a lot of resources.

So what are some GDP/capita estimates? $1000 seems too low, barely above African poverty, if that. $5000? Sure. $15,000? That's half the US of the time -- we also know the Soviets were a lot poorer than America, so it shouldn't be *higher* than that. So 5k-15k, for a GDP of $1.5 trillion to $4.5 trillion. And for a single figure, I like taking the geometric mean, so $8660/capita, and GDP of $2.5-2.6 trillion.

Wiki says $9200 for the USSR, and Trading Economics gives $36-37K for the US of the time.

As for Wal-Mart, how much could it be making? How many people does it sell to, and how much? I know it started in the US, and failed to expand into Germany, which suggests it has tried to expand. It sells to lots of people, but not everyone. Famously it sells to poor people, so they can't be spending *that* much on it. Many people may go there for groceries and regular household expenses, which suggests $200-400 per person, or $2000-5000/year. How many such regular shoppers? 50 million is certainly a lot in US terms. The US alone can't be 300 million, but maybe it they expanded a lot abroad it could be.

So, 50 million * 2000 = $100 billion/year. 300 million * 5000 = $1.5 trillion. That's a big range! But unavoidable when you don't much. Geometric mean is $390 billion. Actual number is $514 billion. Not bad. The real figure suggests, at $3000/year per person, 171 million people. Wiki says it's in 27 countries and multiple brands like Asda... though this raises the question of whether the $514 billion was for the whole conglomerate or just "Walmart". Wiki suggests the former, whew!

USSR vs. Wal-Mart

2019-Mar-05, Tuesday 14:37
mindstalk: Tohsaka Rin (Rin)

1989 GDP: $2.7 trillion in 1989 dollars. By this calculator that would be $5.4 trillion today, and that's not touching nominal/PPP issues (or fake Soviet statistics issues).

1990 population: 291 million, with 152 million workers.

Economic activity: everything from farming to space probes.


2018 sales: $500 billion in 2018 dollars. Or $514 billion, for the fiscal year ending in Jan 2019.

2019 employees: 2.2 million.

Economic activity: a whole lot of super-sized grocery stores and distribution trucks.

Why am I posting this? Because people are praising some book that one reviewer says claims "we are now surrounded by companies and organisations that are as large or larger than the USSR at its apex", and I want to inoculate people against bad ideas. Given that 10% of my USSR GDP is still bigger than Wal-Mart...
mindstalk: (atheist)
I haven't tried reading the proof yet, but there's an alleged trilemma: correctness, cost-effective, decentralized, pick two. It's plausible, sound very similar to the CAP theorem in distributed computing and databases. (Correctness, Availability, Partition-free -- basically if your nodes are partitioned, you can either shut down and stay correct, or be available and risk conflict when the nodes are re-connected.)

There's a list of various paradoxes about cryptocurrency: more users make it worse (more congested), quadratic total storage costs, conflict between users and miners... https://bankunderground.co.uk/2018/11/13/the-seven-deadly-paradoxes-of-cryptocurrency/

I can suggest some math. Say a blockchain is reasonably successful and has 100 million users (Bitcoin aims at competing with the global financial system!) doing an average of one transaction a day[1]. That's 1e8 transactions in under 1e5 seconds, so over 1000 transactions a second. The top blockchains, bitcoin and Ethereum, can handle around 5 transactions a second. See a problem? And a really global system could plausibly need 100,000 transaction a second.

There are some blockchains -- bitshares, EOS -- that claim such capacity, though I've also seen people say that they're not really decentralized, and one study claimed they didn't even measure up to their claims. https://www.prnewswire.com/news-releases/whiteblock-completes-industrys-first-eos-benchmark-testing-and-blockchain-investigation-300742130.html

But let's say some system can handle that many transactions. What's the storage? 20 bytes each for the From and To addresses, say 10 bytes for the amount, 50 bytes per transaction. At the small scale, each node needs to add 50 kB a second to the distributed ledger (which is what a blockchain at heart is.) No big deal. 5 MB at the big scale... still doable.

But imagine bringing up a new node after three years of such activity. 1e8 seconds in three years, so 5e12 or 5e14 bytes -- 5 or 500 Terabytes to download so that you can be a miner too. Eep!

I can imagine a way around that: a blockchain that every N blocks produces an explicit balance sheet, so that you don't have to go through the entire blockchain history to figure out how much someone has. I don't know if that's viable, such balance sheet blocks would be extra-attractive to attack, but I can't say it's unviable either. OTOH, how big is such a sheet? 100 million users, 20 bytes address per user, 10 bytes balance per user, 3 GB. If 3 billion users, 90 GB. Which has to be shipped around and validated by the mining nodes in the time it takes to make a block -- 10 minutes for Bitcoin, 10 seconds for Ethereum, IIRC. So we cut down on the total storage cost but need some really fat bandwidth.

[1] 30 per month. My own records show an average of 2 expenditures a day, or 60 a month. An average worker's month might see 2 paychecks, 1 rent or mortgage payment, 2-3 phone + utility bills, 4 grocery payments, 1 gas or transit pass payment... we're up to 10 a month right there.
mindstalk: (thoughtful)
(Forward links: street land use and Google Street View browsing).

Back in college, I found a newspaper article talking about the decline of US cities (or not, of a few) and it gave population densities. Having read Jane Jacobs and turned into a wee amateur urbanist, I memorized the numbers. I still know them. But of course they were all in people/sq. mile. Since I'm on a one person campaign to get more comfortable with the units used by 96% of the human race, I thought I'd type up the numbers in /km2, for my better retention, with a lot more places, significant to me or friends, added. And then I'll do various botec/Fermi modeling, to try to show what's going on on the ground.

Density data )

Models )

item values

2018-Jun-30, Saturday 11:46
mindstalk: (Default)
Used A/C: ~$1/lb
My new A/C: $3/lb
my old bike: $5/lb
very fancy bike: ~$150/lb
typical new car: $10-20/lb
cheap laptop: ~$70/lb
my laptop: $170/lb
cutting edge smartphone: ~$2000/lb

eating out costs

2017-Nov-30, Thursday 14:50
mindstalk: (Default)
There's a deli near my office where you can get a breakfast for $5-6. Say $5. 6 of them would be $30.

From the store, a dozen eggs is under $2, decent bread is under $5, call it $7 including butter costs. 6 sausages can be $6-9, so $13-16 for six almost-equivalent breakfasts.

Almost because I think the deli breakfasts include potatoes, which I'm not cooking fast for breakfast anyway. Unless I microwaved frozen hash browns. Another $3-5 for that.

Of course, this deli seems unusually cheap for Boston. Probably more common to be $8 menu, $10 after tax and tip, after which you're looking at $60 for 6 of those.

The same place sold me a basic pastrami sandwich for $7. From McKinnon's I can buy pastrami for $8/lb; a quarter pound in a sandwich would cost $2, under $3 including bread. I don't know how heavy the deli sandwich was, though it probably included a pickle. Again, it's cheap; other places would be $8-12.

MBTA passes math

2017-Apr-30, Sunday 08:11
mindstalk: (Default)
Say you're a regular commuter, taking transit at least twice a workday. 10 trips, which would cost $22.50 if you're using a CharlieCard. A 7 day pass is $21.25, so it totally makes sense to buy one, then ride the T whenever you want. Even if you somehow had a 4 day workweek, having a couple more trips would be likely.

Four 7 day passes would be $85; a monthly pass is $84.50. So that makes sense too. Or does it? Say you have three weeks of vacation, and leave town for them; maybe you'd save money by just cycling 7 day passes, and skipping the weeks you're gone.

I approached the math from a couple different angles, but this presentation seems best: a month pass costs about the same as 4 weeks, so 12 monthly passes covers the year for the cost of 48 weekly passes. Even if you skip 3 weeks, you'd still have to buy 49 passes... plus covering that extra day (or two, if leap) in the year. So go monthly!

Though, having been using 7 day passes, I noticed that they actually shuffle forward. If I buy one on Monday morning, the next Monday I can leave a bit earlier and still use it, buying (or activating) my next pass Monday evening. And so on. The effect is that you end up covering 30 days for the cost of 4 passes, as each one picks up an extra "half-day" commute. And if you shuffled into buying a pass on a weekend, well, maybe you could skip travel that day and save an extra day.

Of course, there's a week's worth of 31 day months, so there's that -- you're not quite getting a month's worth for 4 passes.

It's nice doing estimations in my head, but at some point you have to turn to a calculator for precision. A year's worth of monthly passes is $1014. If you cover 30 days with 4 weekly passes, that's $85 per 'month', and $1020 to cover 360 days, with 5 more days to finagle. OTOH, if you can skip 3 weeks, you'd spend just $956.14 in a year, saving $57.75. Or $42.57, if you threw in 5/7 of another pass for the extra days.

Of course, that assumes you can maintain the shuffle. Weekends offer skipping a day, but a regular weekend thing might pin you down. Say I activate a pass at 8pm Sunday to go to Grendel's; the next week I might leave earlier, but I'd still have to activate a new one at 11:30 to get home. The week after that I could leave Grendel's a bit earlier, activating the next pass on Monday morning... okay, it still works, though Sunday feels a bit sticky due to the short 'commute'.

Of course, the monthly pass means not having to buy stuff every week, nor worry once a week about the timing of when you do things. OTOH, saving $40 to 60... well, it's not a ton, but it's not trivial either; 40/1014 is 4%.

Extra thought: if you really use the weekends on your one-week vacations, you could save another 2 days each, or 6 days total, in effect skipping another week.

As for me, if I had today off I'd probably just go monthly. Annoyingly, I probably have 4 or 5 trips to make today. Cash today and monthly tomorrow, or weekly today?


Meanwhile, the $12 daily pass is hard to justify unless you run around a lot. Even for a tourist spending $2.75 per trip via CharlieTicket, it costs more than 4 trips -- though if you're doing train/bus transfers that becomes a lot easier to justify, since the Tickets don't give a free transfer. But even then you'd d need bus/train, bus/train, and one more trip. For a Card user you'd need to make 6 independent trips to make a day pass economical. Most likely use case would be having to make multiple quick trips along a train line.
mindstalk: (thoughtful)
For those who haven't seen it, I'd like to shill last year's urban density post. Data and Fermi estimates, tasty!

A highlight of it is that you don't need that much height to get high density. Manhattan has 26,000 people per km2, but Paris has 22,000, with basically nothing over 8 stories. Brooklyn is at 14,000, twice as dense as San Francisco or Somerville, and my personal impression of it was that it wasn't that high, maybe wall to wall four story buildings.

Well, I finally went looking at random spots of Brooklyn in Google Streetview, and now I'm confused: I seem to have greatly overestimated the average form. Brooklyn Heights, right across from Manhattan, has a bunch of 12 story buildings (possibly office). I do see some wall to wall 4-5 elsewhere, also wall to wall 2-3. But also a lot of detached homes, even 1-2 story. It's not really obviously different from SF or Somerville (at least the parts I'm familiar with), yet has 2x the density.

Maybe it's a change in distribution? SF does have the Sunset, a district of 1 story homes on top of garages. Or in how many people are living per unit.

Or, hmm, back yards. I just switched to the Google Earth view of the last spot I'd checked, and there are none, just lots of smaller buildings in the back. Though I'm not sure if they're housing or garages. By contrast, in SF I lived in a 3 story Victorian, wall to wall, but half or even 2/3 of the lots were back yards, not that you could tell from the sidewalk.

Well, that was one spot; in a second, I do see back yards, some with swimming pools even, but they're 1/3 of the lot length.

OTOH I'm looking at Somerville now, and it doesn't look more generous with back space, though there's maybe more space between the buildings (mostly driveways.)

OTOH again, I just checked San Francisco, and it's what I remember. There's variation, but backyards in the Richmond are commonly half the lot, sometimes less, sometimes more. Ditto for the Sunset, something I never appreciated. Both have a layout where if you walk around the block you'll see nothing but building, but half the block is a contiguous (but property-divided, not communal) greenish-interior.

So compared to SF, I can see why Brooklyn is twice as dense: similar buildings but less yard space. Now I'm wondering why Somerville isn't denser, though... it doesn't even have major parks! But looking again, I think the spaces between buildings, plus greater yard space I now see elsewhere, may explain that.

As for Paris, I was partly wrong: there is a lot of green space, but largely enclosed by buildings like Brooklyn or San Francisco -- often enclosed by what looks like *one* building, in a private courtyard. I was basically right about the height, I haven't seen anything shorter than 4 stories, usually 5-8, though I did see something as high as 12.

I still wonder about my Fermi models; they generally predict more people than we find. I might be overestimating land use, or underestimating non-residential use, or how much area is taken up by walls vs. the internal usable area. E.g., consider a "Main Street" model: 30% streets, 70% lots, half of a lot built up, ground story businesses, two residential stories on top. 700,000 m2 of residential area (700,000 * 1/2 lot use * 2 stories); at a rather generous 100 m2 per person, that's 7000 people per km2. Hmm, that's not far off from SF or Somerville, though my impression is that Somerville is short on local businesses and jobs, and neither has ubiquitous businesses like that. And 100 m2 is high... at 50, say, that'd be 14,000 people, more like Brooklyn, with entirely 3 story buildings and 50% open space (not counting streets.)
mindstalk: (robot)
Say an ebook is 2MB size; most are maybe half that.

5000 such books takes 10 GB, which is 1/6 of the 64 G SD card I added to my phone.

At 100 books a year, 5000 books is on the order of the number of books I'll read in the rest of my life, including re-reads. And that's with 1/6 the card and oversized files.

From another angle, say you can read 400 words a minute, and you read continuously for 100 years; that's 21 billion words in a century. Kind of a ridiculous upper bound, but let's run with it. At 100,000 words per book, that's 210,000 books, which at a more reasonable megabyte per file would take 210 GB. That's still overestimating file sizes -- I have a 200,000 word fanfic stores as a 1.2 MB ePub -- and reading time, by at least 3x -- so combing them puts us under 40 GB.

We're not at the point of having the Library of Congress in our pockets. We are at the point of having a lifetime's worth of reading in a pocket.
mindstalk: (science)
This will be more about crappy data than detailed modeling, but:

How much meat do Americans eat? How much do I eat? Good question, and I don't know! But a lot of the time, especially in my adolescence, I'd guess an average of 4 ounces for each of lunch and dinner. If lunch was lighter, like cold cuts, dinner was probably heavier, like steak or lamb chops or four drumsticks. Of course, sometimes I had PB&J for lunch, and often now I have falafel or hummus. OTOH, I'd doubt *more* than half a pound a day. So, half a pound is 180 lbs a year, 1/4 a day would be 90 lbs, I'd guess 135-180 lbs/year.

As for Americans, am I high, low, or average? I don't know. I know lots of vegetarians here, but statistically they're not that common; I've seen 5%, so take 10 lbs off the 180. Children often eat meat, but eat less, say 10% eat half as much, so another 5%, and we're down to 160 lbs peak. But honestly, a factor of two range is pretty good for a Fermi, I'm unlikely to improve it by pulling numbers out of my ass.

So, data?

Chicken council has a nice breakdown table; 2012 is 200 lbs of red meat + poultry, 14 lbs fish+shellfish, so 214 lbs. A lot higher than my estimate. A footnote says it's mostly retail weight, though fish is edible weight (so, not counting mussel shells?) and a few minor entries (turkey) are carcass weight.

NPR says 271 lbs. A lot higher! Except, then it has a breakdown chart, which shows about 55 lbs beef, 58 lbs chicken, 45 lbs pork, 15 lbs turkey, about 170 lbs total. That's a huge difference, and shows how much internal consistency checking the journalist did. The latter numbers also match up well to the Chicken Council, except for chicken where they had 80 lbs.

The Meat Institute claims 6.9 oz/day for men and 4.4 oz for women, for an average of 5.7 oz/day, or 129 pounds/year. Or less if we should exclude children. They also say the US produced about 95 billion pounds of the usual land animals, which is 317 pounds per person. They also say that the US exported 7 billion metric tonnes of beef+pork+chicken + "variety meats", which is 15,432 billion pounds. And that exports were 10-20% of US meat production. Ummm... to cap that off, pork and chicken shipments are said to be valued at about $5 billion, but beef exports at $800 billion.

So that's a whole lot of garbage.

The USDA has saner looking numbers... that's just a beef link, but it says 24 billion lbs/year beef, or 77 lbs/person-year of beef. So already that doesn't agree with the first two, though it's close to what the Meat Institute says.

The WSJ says the USDA says 71 pounds of red meat (including pork), 54 pounds of poultry, for 125 pounds/year per person. Which is close to the first Meat Institute number I cited. Their graph says 132 pounds, but that might include fish -- though 7 pounds fish is only half of the Chicken Council number.

It has the interesting lines " These numbers factor in food loss at each level—carcass to retail weight, loss at retail such as spoilage and loss at the consumer level, such as plate waste. About half of the weight of meat is lost from the carcass to the consumption." I suddenly wonder if the "retail weight" includes bones.

Perhaps the 270 is carcass, the 170 is retail, and the 130 is plate. But, I'm having to guess and handwave, in fandom we call that fanwanking. In practice, the uncertainty of the reported numbers is as big as the uncertainty of my wild guesses, though the range is somewhat higher, 130-270.
mindstalk: (food)
Previously I'd tried estimating how much Americans spend on alcohol. Didn't directly estimate how much they drink, but I've now seen articles on that.


Log-normal curve. Yep, 30% don't drink, and another 30% have at most a drink every two weeks. But the average (calculated by hand from the graphs) is 9.8 drinks a week -- which is between the 8th (6.25) and 9th (15.28) deciles. Meanwhile the top decile is downing more than 10 drinks a day. At least, if we trust the government survey this is based on.

I'm probably in the fifth decile these days.
mindstalk: (thoughtful)
Number of cattle: 1.5 billion in the world.

Weight, I dunno. A newborn calf can be 25-45 kg, as much as a small woman on the upper end. 700 kg for an adult cow or steer, 1100 for an adult bull. I don't know how many are meat vs. working animals, and presumably meat animals don't stay as adults for long. Let me guess an average mass of 300 kg.

Say an average human is 75 kg. That means the biomass of cows today could instead be another 6 billion humans, for a total of 13 billion. They wouldn't even have to be strict vegetarian humans, just need everyone to replace beef with vegetables; chicken and pigs and such would still be around.

Of course, that replacement might not be trivial, if most of the cattle are eating grass; you'd have to turn pasture into farmland. But still.

(ETA: USA has 90 million cattle, suggesting we could feed another 360 million Americans.)


So, potatoes are crazy. The Dutch produce 44.7 tonnes per hectare. You can feed someone for a year with about a ton of potatoes, or less, so that's nearly 4500 people per square kilometer of potato farm. Maybe 5000. The Dutch are top, but other northwestern European countries also produce similar amounts. Italy and Spain are down at 25-28 t/ha, eastern Europe around 13, for only 1300-1500 people per km2. I don't know why the differences. Climate, water, fertilizer?

There's about 14 million km2 of arable land (out of 48 million km2 of agricultural land), so if that all was growing potatoes at similar rates, that could support from 18 to 70 billion people, vs. the 7.1 billion of today.

Why so much? According to Charles Mann, potatoes produce 4x the edible dry biomass that wheat does. Why that, I don't know; I'd guess being able to put most mass into the tuber, rather than stalk. Of course stalks can be fed to livestock or furnaces or industry, so wheat chaff can turn into useful things. But for straight food production I'm guessing root crops rule.

Another source gives 189519 hectograms per hectare for the world in 2012, or about 19 tons per hectare, vs. 4.5 tons/ha for rice. Maize is 4.8, other cereals less than rice, with wheat at 3; taro is 7.6, cassava 13. Those numbers aren't so far apart in calories: potatoes are 22% dry biomass, wheat 88% (Mann again.) The gap's narrowed since the 1760s, when Andrew Young found eastern England producing 1500 lbs/acre of wheat, but 25,000 lbs/acre of potato, for 4x the calories per acre. (That, or England's just better at producing potato.) If maize is like wheat, it's slightly more food per land than potatoes. (Though I think potatoes are a more complete food.) Of course, most US corn is fed to livestock, bringing us full circle.

Potato land is given at 193,000 km2. Maize, rice, and wheat add up to about 5.5 million km2, leaving a lot of arable land growing other stuff. Soybeans are another million.

Arable land is 9% of land. Scaling somehow up to 33% of land, and using the highest number, that'd support 250 billion people. Trillion person Earth would need 4x the highest national level potato yield. Challenging.


So, between a vegetarian diet and more optimal growing conditions, there seems to be room for a bunch more humans. Possibly a lot more humans.
mindstalk: (Earth)
In online debates, I've often seen some doom-laden people say "maybe solar can meet our day to day needs, but there's all the embedded energy in stuff, we'll never make that up." The classic example is the claim that solar panels don't pay for their own creation, which I'm fairly sure is false by now, if ever true. But today I'll talk about cars: how does the energy of a car compare to moving it around?

Say the cas lasts 100,000 miles, and is particularly efficient at 40 mpg, so uses 2500 gallons in its lifetime. A gallon is about 4 liters or 4 kg of water, so we're talking about 10,000 kg of gasoline. The car probably weighs 1-2 tonnes, toward the lower end if it's getting 40 mpg. Even if the car were made of air-synthesized gasoline, it'd still be a small fraction of the lifetime energy cost, and it's not, it's made of stuff like "turn iron oxide into iron". Ideally speaking, nothing in a car is going to compare to the energy density of gasoline, though it's possible processes aren't ideal.

A factor of 10 is close enough to be worth more precision. A gallon of water is actually 3.7 kg, and oil is lighter than water -- 70% the density, even. So 2.6 kg/gallon gasoline, and 6500 kg for the total. And this gives 1.3-1.6 tonnes for the weight of compact to midsize cars. So, the lifetime gasoline weighs 4.3x as much as the car, and is still probably a lot more energy intensive.

Another thing that gets brought up is moving parts around in shipping and manufacture, how components of something might have made a few trips across the Pacific or world among them. Mexican iron ore to China to become steel to become a car in the US, say. For other goods that might be significant, but here we're talking about a car, which by assumption already moves 100,000 miles under its own inefficient power. If we grant a round trip across the Pacific, of maybe 19,000 miles, that's still 1/5 the distance, and container ships are far more energy efficient than a car's engine. I'd guess 10%, making the parts-transportation energy more like 2% of the car's lifetime total. Which suggests it might be significant for things that aren't cars. Then again, Without Hot Air says shipping can be 1.5% the energy of road transport, far smaller than even my guess.

Wait! One last check. Without Hot Air again says a car's embedded energy is 76,000 kWh, which would be 2.7e11 Joules, or the equivalent of 6750 kg of gasoline. That's way more than my estimate, directly comparable to the amount burned moving the car around. Wikipedia says the same thing, but it's quoting the same source (Treloar et al.) OTOH, says "the Union of Concerned Scientists (UCS), pointed out that a common life-cycle assessment calculation is that 85 percent of embodied energy use associated with a conventional vehicle’s life cycle is attributable to operation and 15 percent is attributable to manufacturing and disposal". This says "on average, every kilogram of steel you add to a vehicle will add about 5 kg of associated carbon emissions." and "Total embodied energy can account for 15 to 30 percent of a vehicle's total emissions over its lifetime." and gives a table of energies, such as 38 MJ/kg for galvanized steel, which is pretty much the same as gasoline. Aluminum and plastics or rubbers are even more. And this gives a longer table, with slightly but not hugely different numbers.

So that's a range of estimates, from maybe 1:6 to 1:1. Looks like my naive chemistry failed and stuff is a lot closer to gasoline density in manufacture than I thought. Not surprising for plastic (it's solid oil) but I thought steel would be cheaper than that. Of course, at less than 40 MPG the notional car would be burning more gasoline, but still; for cars of that weight that's likely no more than a factor of 2.

Except for one last complication: I was using the chemical energy density of the physical gallon of gasoline. Total energy cost of operating a car would include mining the crude oil and refining it into gasoline, which adds at least a bit.
mindstalk: (Default)
How many driving instructors are there in the Boston area?

1 million people, 10,000 turning 16 every year and mostly learning to drive at some point, 10 hours to learn to drive, 1000 hours of instruction time, so 100 students a year and 100 instructors. If it takes less time to learn or more is available I could see the number dropping to 25; I don't see it going to 400.

But! Most families have cars, do most people learn to drive from their parents? That could cut business by 90%, leaving 10. There may also be adult in business in the form of buying off driving infractions with lessons; if 10% of the adults take such a lesson at some point in their life, then that could support another 10 instructors.

The web will show me driving schools, not individual instructors; how many instructors per school? Basic driving instruction doesn't have much economies of scale: one student behind the wheel, one car, one instructor in the passenger seat. Having a business could be as simple as putting out an ad an showing up, unless there are regulations about having a car the instructor can take control of. But there could be some team up under brand name or sharing office expensives, I'll guess 2-3 instructors per school on average.

I'm ignoring motorcycles or commercial vehicles, just looking at basic cars.

Yelp lists 10 schools in the area, missing one I know of, but including a motorcycle school.

Sonia's makes teens take 12 hours behind the wheel, 6 in observation, and 30 in the classroom. [this seems standard] $450 for all that, or $30/hour for road practice. Indicates the law does require a two-brake system. No indication of how many people are involved.
Love's has one instructor.
Brookline has "a handful" of instructors.
North Quincy review mention 4 names; the website says "Collectively, we have close to 50 years of driving experience." so can't be too many people.
Safety Auto claims to have service a wide range of immigrant communities, and taught 3000 students in the past 2 years, vs. my rough 100/year estimate for one person. So maybe 7-15 instructors.
City Auto claims 50,000 students since 1998 or about 3300/year! 15-33 instructors?
D&D totally unclear, plus reviews call it a scam that makes you fail your test.
Metro: one review about parking in a handicapped spot, dead website.
Boston DS: reviews make it sound like one guy. Website gives us director, manager, and CFO, but not instructors.

Google lists some more:
Natural: no info
Newton: father/daughter team, photo of three cars.
A-L&L: no idea, though name and photo of two cars sounds like a two-person thing
T-Guide: no idea
Success: "thousands in the past 12 years". 3 cars.

Then Google Maps has a bunch of small dots.
Friendship: one guy?
International: I remember thinking a husband-wife team, don't know why.
Arlex: family-owned.
Canto: "Our staff speak English, Portuguese, Spanish, Cambodian(Kymer) and other African languages." That would imply at least a few polyglots, except "other African languages" makes me suspicious, as none of the ones mentioned are African.
Henry's Everett is family-owned.

Okay, there's not quite 40 dots on the broader map, I won't go through them all. But 100 instructors seems pretty plausible, in the end. I realized along the way that "most people learned from their parents" may not apply to immigrants; the two schools that seem potentially big have Chinese on their website or emphasize their immigrant service.
mindstalk: (CrashMouse)
The series continues! How much money do Americans spend on alcohol?

320 million Americans, I say 3/4 of them are adults, 240 million. That probably bleeds into the 18-21 set, but hey, a bunch of them drink anyway.

How much do they drink? That's hard. People I know include teetotalers, having a beer or glass of wine every other night or 2 of 3, and dropping $20 at the bar every Sunday night, not to mention other drinking. But I note that if you drink at all regularly then it's easy to spend at least $10/week: a beer or two a night, a glass of wine a night, two $5 cocktails on the weekend, one $10 cocktail. $10/week is $500/year is $120 billion. $5/week gives $60 billion, $20/week $240 billion.

$40/week would be $6/day and is starting to feel too high to me, so $480 billion is a strong upper bound. Half of $5/week sounds like half the population not drinking at all and half drinking $5/week, which seems too low, so $30 billion is a strong lower bound.

We might try a more complex model, but e.g. if 1/3 don't drink, 1/3 drink $10, and 1/3 drink $20, that averages out to $10 for the lot and we're back at $120 billion.

Intrusion of actual data: a poll I saw a while back claimed that in fast 1/3 don't drink at all. If that's true, and if the other 2/3 drank at $10/week on average, that'd be $80 billion.

If 1/3 don't drink, 1/3 are casual at $5/week and 1/3 more serious at $15/week, that's $80 billion.

$20/week is $1000/year which is like 1/40th of the median household income though 1/20th of a two person household. Seems like people could be spending 5% of their income on booze. Check: $240 billion is less than 2% of GDP.

There's expensive wine out there, might it tilt the numbers? If 1% drink a $100 bottle of wine every night then that's $36,000/year and I don't believe this, the 1% aren't that rich. It would also be a total of $86 billion. I conclude the actual consumption of $100 wine is not going to be significant.

Okay, so we've got a wide range of $30-480 billion, a tighter one of $60-240 billion, and my gut instinct favoring upward of $120, except for that big 1/3 don't drink figure which pulls it down sharply.

That's my guesses. Care to try your own before reading further?


So, let me look stuff up.

http://www.bls.gov/cex/csxann10.pdf says $400 is spent... per "consumer unit" on alcohol. 2.5 people per unit, so $160/person. Well, 2 people over 18, so $200/adult. That gives about $50 billion.

http://money.msn.com/saving-money-tips/post.aspx?post=9113d6f5-1a32-4187-afb6-d7d5164df959 cites the same, and adds the money has shifted from retail to bars; this may simply reflect prices. Also "U.S. per capita consumption in 2009 was the equivalent of 2.3 gallons (8.7 liters) of pure alcohol, a lot less than the 2.76 gallons in 1981 but more than the 2.14 gallons in 1997."

http://www.drug-rehabs.org/alcohol-statistics.php is a lot more alarmist than the BLS and says $90 billion/year. No source is given.

http://articles.chicagotribune.com/2012-01-31/business/ct-biz-0131-liquor-export-20120131_1_liquor-sales-alcohol-sales-david-ozgo says "$59.24 billion alcohol industry. The data represents sales by manufacturers and importers, not retail sales" So consumers could be spending more at the bar and store (tax) but OTOH much of that number might be exports, too.

http://smallbusiness.chron.com/revenue-comes-selling-alcohol-34021.html says $90 billion but again gives no source... oh wait, list of references, which grounds out in this recovery site with no other references.

So, the most reliable seeming data is the BLS; unless there's a statistical mistake somewhere, we've got $50 billion, on the order of manufacturing sales, though it feels like that should be higher at the consumer level. $90 billion floats around rehab circles but has no good source.

Conclusion: $50 billion is outside my tight range though within the wide range. Americans drink a lot less than I thought, or drink much more cheaply than I thought. One of the sources repeated that only 65% say they drank in the past year, so $50 billion/160 million who drink = $312 per drinker, or $6/week. On the order of 4 bottles of beer on the weekend, or half a bottle of wine, or one cheap bottle, or 3 two-buck Chucks from Trader Joe's...

mindstalk: (food)
How many restaurants are there in NYC? 8 million people. Lots will eat almost always at home, lots eat out for lunch five days a week, a bunch eat out all the time. At first I thought 2 million might be eating out for any meal; then I remembered children and thought about income inequality, so dropped down to 1 million. Assuming an average restaurant can feed 30 people during peak meal time, that gives 30,000 restaurants.

Source one: 35,000 full-service + counter restaurants, plus another 4000 snack and specialty food bars. Within 33% of guess.

Quora, as seen in Google search results, says 24,000 listed as inspected. Within 20% of guess.

But the NY Post says only 8000. Way off, only within 10x of guess. Mention of Williamsburg (in Brooklyn) suggests they do mean the whole city, not just rich and dense Manhattan.

But this says 6000 additional bars and restaurants as a 47% surge allegedly due to smoking ban, so about 18,000 total. First source gave about 18,000 full-service restaurants and 2600 bars+nightclubs.

My estimate extended to the whole country would give 30,000 * 320/8 = 1.2 million. Though we might suspect NYC was high in restaurant use, both as a city in general and as NYC where many apartments might barely have cooking facilities. This says about 580,000 restaurants in the country, according to some market research firm, so cutting by a factor of 2 would have worked.

(Also: " A disproportionate number of the restaurants closed have been independently owned, leaving the field to chains, which now make up 46 percent of American restaurant locations.")

Conclusion: meh. Originally I thought "Boo-yeah!" with two estimates within 20% of my guess, but looking at more links, it seems that as with weddings, estimates very widely, from 8000 to nearly 40,000! and reliable data's hard to find. The first one gives the impression of having done more work so maybe being most accurate.
mindstalk: (Default)

How much money is spent on weddings in the US? I figure US cohorts are 4 million people per year. Assuming almost everyone gets married eventually, and that those who don't are made up for by re-married people, that's 4 million weddings a year. Guessing $10k per wedding, $40 billion.


A couple sources said $40 billion or $51 billion, and I felt very proud -- getting within a factor of 10 is a decent success for a Fermi problem, never mind a direct hit -- until I realized that weddings involve two people, so I should have figured 2 million weddings, and $20 billion. Oops! (In my defense, it was like 3am.)

The actual numbers quoted seem rather scattered, actually. Wikipedia says 2.5 million weddings and $40 billion, yielding $16,000/wedding, but the cited source no longer exists. Citing a different source it says $28,400 per wedding, but at 2.5 million that should give a total of $71 billion.

This says 2.1 million, spending a total of $86 billion, or $41,000 per wedding. It also claims $70,000 per wedding in Japan in 2005. Given that most sources have Japan GDP/capita at a fair bit less than the US -- especially in PPP -- I am skeptical of this source.

And this says "* The average wedding cost is $26,501, slightly more than a 5% decrease from 2009 when the average cost was $28,082 but up $8,000 since 2002." without giving total numbers.

(Going from $18,000 in 2002 to $26,000 in 2011 seems like a huge jump. 44% increase. Well, there is inflation... but 2% inflation over ten years gives 22%. 3%, 34%. 4%, 48%. But inflation's been low, at least since 2008...)

Business Week cites $51 billion, which would be $24,000 for 2.1m weddings or $26,000 for 2m weddings.


So, I thought this would be a trivial problem, where we Fermi something we can simply look up as an exercise and calibration of estimation, but actually we've got a scatter of numbers and estimation may help us trust some sources more than others. 2 million and about $25,000 seem the most likely numbers, for $50 billion; whether the $25,000 is a *typical* number or is badly inflated by big spending of the rich, or even inflated by selective reporting, is unclear. Example of the latter: does anyone have any real idea how much is spent on weddings, or is it 'couples who register with us spend an average of $25k, so we extrapolate from there, even though courthouse elopements are invisible to us?'

Will Oremus in fact says it's the latter:

One of the most extensive surveys, and perhaps the most widely cited, is the “Real Weddings Study” conducted each year by TheKnot.com and WeddingChannel.com. (It’s the sole source for the Reuters and CNN Money stories, among others.) They survey some 20,000 brides per annum, an impressive figure. But all of them are drawn from the sites’ own online membership, surely a more gung-ho group than the brides who don’t sign up for wedding websites, let alone those who lack regular Internet access. Similarly, Brides magazine’s “American Wedding Study” draws solely from that glossy Condé Nast publication’s subscribers and website visitors. So before they do a single calculation, the big wedding studies have excluded the poorest and the most low-key couples from their samples.

In 2012, when the average wedding cost was $27,427, the median was $18,086. In 2011, when the average was $27,021, the median was $16,886. In Manhattan, where the widely reported average is $76,687, the median is $55,104. And in Alaska, where the average is $15,504, the median is a mere $8,440. In all cases, the proportion of couples who spent the “average” or more was actually a minority. And remember, we’re still talking only about the subset of couples who sign up for wedding websites and respond to their online surveys. The actual median is probably even lower.

So who knows? My $10,000 per typical wedding and $20 billion total might not be so far off.
mindstalk: (Default)
Well, not visualized with pictures, whaddaya expect of me. But I had this previous post about how parking minimums capped potential density, which got into algebra as I approximated infinitely tall and narrow towers to squeeze the most use out of a lot. I realized there's a much simpler way of figuring it out: dingbats. Assuming one doesn't get into the expense of multi-level parking, the maximum parking you can get out of a lot is the entire lot, with your actual building on stilts above an open garage or stretch of carports. Or not open garage; point is, your whole first story is cars, and people go on top of that.

So then it's just a matter of what the legal ratios let you do with that parking. If you want 60 m2 2 BR apartments and the law says 1 30 m2 parking space per bedroom, then you have 1:1 apartment:parking space, and all you can get is one story of living space. Your development is two stories, kind of: a level of apartments above a ground level of parking.

If the law lets you have only 1 space, then it's 2:1, and you can have two stories of residences.

1 space per 90 m2 (about 1000 square feet), 3 stories.

2 spaces per 90, that's 90 m2 apartment : 60 m2 parking, so 1.5 stories of housing, awkward. Maybe you do a terrace with more balconies or skylights, or maybe you just build bigger apartments to use up the space.

1 space per 30, that'd be the same as 2 per 60, one story.
1 space per 15, that's 1:2, and you wouldn't even be able to use your whole second story, only half of it. One level of cars, half a level of apartments. Realistically you'd just have 1/3 of your lot be first story apartments, and the other 2/3 be parking. Even more realistically, you don't build 15 m2 apartments.

How about non-housing? Same idea: 3 spaces (330 square feet each) per 1000 square feet of office space is 1:1, so at most you could have one story of offices above a layer of parking. Office parks and malls are more likely to build multi-level parking though, as are big residential businesses; it's not worth the expense of ramps and such for small ones.

By contrast, of course, without parking the sky is almost literally the limit. 5 stories, 10, 20, 40, 80...

Also note that ubiquitous dingbats means that everything at ground level is a garage. Pretty yecchy. Avoid that, and density drops again.

Often people don't do dingbats, of course, they either do a single story next to parking, or a multi-story next to parking. But those will be even less dense than if the whole lot were used for parking.

Another way of looking at it is there's a physical limit to how many cars you can have anywhere with ground-level parking, and parking requirements tie the number of apartments or bedrooms, and thus of people, to the number of cars and parking spaces. Cars first, people second. So for developers it becomes a question not of "how many people will I build for" but "how much space will I provide the maximum number of people I can have?" If there's 1 parking space per bedroom then you can't build more bedrooms by building up, you can just choose how big the bedrooms and apartments are. One size at one story, twice as big at two stories -- but it's the same number of bedrooms.

A friend seemed skeptical of "parking requirements make housing cost more". He didn't give his argument apart from seeming to think landlords will charge as much as you can pay (as opposed to supply and demand), but I came up with a counter-argument anyway. All else being equal, bigger apartments cost more than smaller ones, right? So consider a 60 m2 apartment with attached parking space. In a sense, this is like a 90 m2 apartment. Not exactly like, since the parking lacks power or quite likely even walls, but in area it's 90 m2. In cost it will be between a 60 and 90 m2 apartment without parking space; if land is expensive, dwarfing pure building costs, it'll probably be much closer to 90.

So when you rent a 60 m2 apartment with parking space, you're kind of really renting a 90 m2 apartment, whether or not you use the parking space. If you don't use it you're stuck paying up to 50% more rent anyway for something you can't use.

If there are two spaces, then you're virtually renting a 120 m2 apartment, twice as big as the actual apartment you live in.

If there's 1 space for a 30 m2 apartment, you're renting 60 m2.

If there's 1 space for a 15 m2 microapartment, then you're renting 45 m2. Again, for such small housing, parking is 2x as big as the human space, and no one's going to do this, meaning that such limits de facto ban such housing from existing.

More bluntly, if parking is required, then it is impossible to build an 'apartment' less than 30 m2, counting the parking. And that wouldn't even be an apartment, just a parking space. 45 m2 is the real limit. If someone desperate to not be homeless wants to just pay for 15 m2, too bad. The city won't let them.

In this case multi-level parking doesn't change matters, it just allows more apartments to be built. You're still stuck paying for an apartment+parking that's bigger and more expensive than you might want.

How much more in reality? Varies a lot. Alan Dunning says land in his part of Seattle costs $38-45 per square foot, which is up to almost $15,000 for a parking space. Shoup found UCLA was building garages at $15,000-30,000 per space, not counting land costs. Beverly Hills developers were willing to pay $53,000 to get out of providing parking.

Even at the low end, an extra $15,000 on a $200,000 condo or home is an extra 7.5%. Not huge compared to 50% or 100% but not trivial. People would sure kick if the government levied a new 7.5% tax on all housing sales. $15,000 on a $100,000 condo would be 15%.

Parking requirements both limit the supply, by limiting density, and create a minimum size of practical apartment to build and rent, and a pressure to build bigger housing where the parking add-on isn't so proportionally big.
mindstalk: (thoughtful)
CDC injury data: the US in 2009 saw 12,000 gun homicides, and 550 gun accident deaths, which I'll round up to 600. Also 18,000 gun suicides, which I'll ignore. (Though I wonder: are accidents all accidental deaths, or just accidentally shooting someone else, with shooting yourself by accident being a suicide? Ah, I see indication elsewhere of "self-inflicted" accidental deaths.)

Mercy cut )

ETA: Of course, guns cause injuries, not just deaths; that should have been factored in as well. Oops. Well, say there's 10x as many injuries. If they cause medical costs of $700,000, that basically doubles the 'budgets' or total social costs If they cause medical costs of $70,000 or less then they add 1% and are kind of irrelevant. I guess there's not just up front medical costs but permanent injuries like lost eyes or fingers. Hard to find recent sources, but late 1990s sources have non-lethal gun injuries as 4x the rate of lethal ones, ignoring suicide[1]. So they'd have to be 'worth' $3.5 million each to double the total cost of guns.

[1] About 80% of gun suicide attempts were fatal. 25% of attacks, and less than 10% of accidents.
mindstalk: (Void Engineer)
There are 149 million square kilometers of land, including all the desert, mountains, jungle and tundra. There are about 7 billion humans. This comes out to 2.12 hectares per person, or in snowflake units, 5.2 acres. An American football field with endzones is 5353 square meters, or 1.33 acres; a pro football/soccer field is 7140 square meters, or 1.76 acres. So, depending on where you live, everyone gets 3-4 football fields of their own.

I strongly suspect that with fire and metal tools, let alone power ones, even a wimp like me could totally deforest all that area. Probably not keep tree shoots and saplings from springing up -- that'd be what goats are for -- but chop down all the multi-year large trees and keep new ones from growing far. Also expect that with guns I could kill all the large animals I didn't want. So basically, humans evenly distributed could make all large animals and many trees go extinct, or undergo severe selective pressure.

As for terraforming in the most literal sense, you can imagine how much dirt you could schlep around your five acres. A deep grave is 2x1 meters x 2 deep, and seems doable in at least a day, if not a few hours. (2x2 x 1 deep would probably be easier.) 28 years to dig out your land at that rate. Non-trivial, but conversely the human race could turn over the Earth's whole surface two meters deep in half a lifetime. That's pretty geological. And that's people with shovels, not backhoes.

Also you get to imagine staying alive by farming 5 acres. Or, probably half of that or less, what with the mountains and deserts and tundra and such.

Earth's atmosphere is 5e18 kg. We breathe roughly 10 liters a minutes, or 15 kg/day. So the human race breathes 3.8e13 kg a year, or about 1/100,000 of the mass of the atmosphere. Okay, that doesn't seem huge.

A human's metabolic energy is about 100 Watts. An American uses 10,000 watts via various means, as do Canadians and Scandinavians. So an American is using the energy of 100 humans -- and the 300 million Americans are using enough energy for 30 billion people, and 4x the metabolic energy -- and respired air -- of the whole human race. Humans as a whole are using 1.7e13 watts, vs. American 3e12 watts, and world metabolic 7e11 watts, so with fossil fuels we're "breathing" 1/3,000th of the atmosphere per year.

There's about 3e15 kg of CO2 in the air, vs. the roughly 1e15 kg of air that we "breathe" industrially.

Average land rainfall is 72 cm/year, leading to about 14,500 m3 of water on your 5 acres. Americans use 1880 a year, about half of which is for power plant cooling. Collecting all the water that falls on your land may be a challenge, depending on terrain and climate. So, not using all the water, but definitely making a dent.

April 2019

141516 17181920

Expand Cut Tags

No cut tags

Style Credit