Jul 14, 2010
What is your time worth? Economists generally assume that your time is accurately valued by the market, because time can be converted into money (via wage labor) and back again (by hiring people to, e.g, do your chores for you). In this article, I argue that the economists are wrong, and that, as a result, we have some important questions to ask ourselves about the value of time:
(1) How would a rational person measure the value of an hour of her time?
(2) Can time invested now be meaningfully traded off against time available later?
(3) How much of your time should be spent assessing whether you are spending your time well?
Executive Summary (tl;dr)
The economists are wrong because, on the supply side, it's difficult to supply extra paid labor on a timescale of weeks or months; and, on the demand side, there are practical and aesthetic limits to how much extra time you can buy for yourself. Thus, a rational person will not assume that her time is worth what the market says it is worth; your time is usually more valuable (to you) than it is to the market, because there is a market failure for medium-sized chunks of your time. One way to get a sense of what your time is worth to you is to measure your ability to achieve goals or satisfy desires, but not all time is created equal -- an hour that you spend programming code may be implicitly supported by several hours of sleeping, eating, desk-cleaning, and other homeostatic tasks. Attempting to normalize the value of your time against a hypothetical "maximum productivity" level might yield more accurate information than simply checking your brain's cached value for how satisfying an experience was in hindsight, because memory focuses on events associated with intense emotions and thus often fails to adequately account for the cost of wasted time. People who spend essentially no time computing their ability to achieve their ultimate goals are probably wasting a lot of time, but the cost (in time) of evaluating your own productivity is computationally intractable; ignorance about ignorance is recursive and cannot easily be eliminated from complex chaotic systems like your life.
When Time Isn't Money
Basic models of labor economics tend to treat a person's decision about how many hours to work each week as a scalar variable; Bob could work 39 hours this week, or 40 hours, or 50.3 hours, and would be paid accordingly. There are several reasons why this assumption might not hold.
Among professionals, many people are not paid on an hourly basis at all, and must simply hope that increased work put in over a period of months or years will lead to promotions and higher annual salaries in the distant future. This might not pan out if, e.g., one's industry collapses, one's company collapses, one's immediate supervisor is vindictive, or life circumstances require a geographical move or leave of absence. To the extent that promotions are based on networking, politics, perseverance, and/or luck, professionals have no ability at all to trade off leisure for money at their primary jobs. Professionals usually have only a minimal ability to increase their income by taking on a second job, because each professional job usually takes in excess of 50 hours a week, and working 100+ hours a week is often unsustainable or undesirable. Professionals can moonlight as unskilled or semi-skilled workers, but this usually involves a severe pay cut; if you earn the equivalent of $50 an hour as a programmer, and you could earn $10 an hour as a barista, it's difficult to say that you can transform time into money efficiently enough for money to serve as an accurate valuation of your time. For example, if we really think that you value your time at $10/hour, and you usually have good self-control, we might be surprised at your decision to pay $20 for a taxi that saves you an unpleasant 1-hour walk. The solution to this trivial paradox is that the marginal price of your time differs from the marginal value of your time -- there is a market failure for small-to- medium-sized chunks of your time.
Among wage laborers, many people are specifically prohibited from working overtime, or are not assigned tasks that are expected to take more than a set number of hours per week. Working more than one full-time job at a time can be made more difficult by standard business hours; there are far more jobs available, e.g, between 9 am and 5 pm than there are between 6 pm and 2 am. Inflexible start times and long commutes can make working multiple jobs logistically impossible even where one is able to find jobs that do not literally overlap. One's ability to work may be limited by the strength and depth of her informal support system; e.g., if your parents are willing to babysit their grandkids for 10 hours a day but not for 14, then working an additional 4 hours would require paying for childcare, which could eat up much of the gains of additional wage labor. Your ability to work long hours may also be limited by your ability to spend a long time away from your kitchen; if you have a long commute and little storage space at work, then you may have to return home every so often in order to prepare your groceries, or else spend a significant fraction of your extra wages on eating out. For anyone lucky enough to be unfamiliar with these problems, Barbara Ehrenreich's Nickel and Dimed is a good place to start.
On the demand side, there is a soft limit to the conveniences that money can buy. Although fabulously wealthy people may be able to afford private planes or helicopters, most upper-class people are still stuck commuting in relatively suboptimal conditions at the speed of a car or train. Flexibly priced toll roads are still in their infancy; it is very difficult to double your transportation budget so as to halve your transportation time. Similarly, first-class train compartments do not run on smoother tracks than the coach seats. If you like, you can pay higher rent or invest in pricier real estate to reduce your commute time, but if you don't like living in the city, you may not *want* to move downtown. Likewise, you can pay someone else to clean your bathroom, but you might not want to pay someone else to cook for you -- there's a point where delegating all of the tasks that contribute to your personal maintenance can be disempowering and dehumanizing. If you exist only as a specialist in fluid mechanics and a consumer of other people's wage labor, who are you? More prosaically, if you pay a personal assistant to organize your filing cabinet, you may be unable to quickly locate your files unless you also pay him to tell you where your files are: some tasks can't be delegated in neat little chunks.
None of this is to argue that time has no value as money, that money cannot save you time, or that, in the long-term, people are unable to shape their lifestyles so as to better approximate their demand for money and leisure. The point is that, given current conditions, money is a deeply flawed measurement of the value of time, and that, as rationalists, we may be able to significantly improve upon it.
If not money, then what should we use to measure the value of time? If you have Something to Protect, then you might try to measure changes in your expected probability that you will successfully protect [your sister, the ecosystem, the collected works of Gilbert & Sullivan, etc]. Many people around here seem to think that achieving Friendly AI is the most important instrumental goal of our generation by several orders of magnitude; if you really believe that, then why not ask, "Will action X increase or decrease the probability of fAI? By how much?"
If your values don't all converge on a single goal, you might ask a series of questions like "How much fun am I having? Will action X be fun? More fun than the alternatives? Will it help steer me into a future where I have more fun?"
This sounds both simplistic and un-workably vague, but consider the possibility that we might routinely fail to optimize our use of time even by such simple measures. If you're an SIAI-shipper, did you floss your teeth today? E-mail your aunt? Change the oil in your car? Write your Congressman? Did any of those things actually increase the likelihood that fAI gets invented in time to save us all? By an amount even remotely comparable to, say, volunteering to translate an SIAI article into Spanish? Or, suppose you're a Fun Theorist. Did you have fun today? More fun than yesterday? Do you know why or why not? What are you going to do about that tomorrow? I find that I rarely ask myself these questions. I am much more likely to ask much less relevant questions, such as "Which way is the Post Office?" or "Which element comes after tin again?" or "Do I approve of the consumer reform legislation?" The process of solving these questions sometimes either is a little bit fun or results in a marginal increase in fun, but not nearly as much fun as I could easily get by putting 5 minutes worth of thought into daily fun-optimization.
What about you? Do you ask yourself any questions like these? Do you answer yourself? Can you think of a fourth category of time-measurements (besides money, fAI-success, and fun) that might be even more useful? Of the categories, which do you think is most useful?
The Hidden Costs (in Time) of Spending Time
[Note: this section is re-posted and plagiarized from my earlier comment on an Open Thread. I feel like it might be OK to post it twice because it got 9 points and only 1 comment; the commenter basically said that he didn't know how to answer the questions in this section but that he was curious.]
Maximum Theoretical Productivity
Assuming you buy the analysis in the section above about the hidden costs of time, one could imagine a maximally productive hour in which all administrative and maintenance tasks had already been previously taken care of. When you actually get on the kayak next to your friend, or actually sit down to write the essay with a comfortably full stomach and a functioning A/C, you are close to maximum efficiency. You are efficient not in the rationalist sense of having chosen the shortest route to your terminal goal, but in the mechanical sense of moving along whatever route you have chosen at the fastest possible speed. As you deal with more and more of your homeostatic bullshit in advance of your attempt to begin your chosen task, i.e., as you reduce wasted time to zero, you asymptotically approach your Maximum Theoretical Productivity ("MTP").
Note that this may be a very stupid thing to do; there is no particular reason, usually, to want to be at peak efficiency from 4 to 5 pm, even at the cost of spending all morning and afternoon getting ready for that hour of efficiency. In fact, such a strategy would probably be a massive waste of time, and seriously lower your average productivity for the day. Final exams and job interviews are two notable exceptions, but mostly I'm just trying to get some analytical traction going here -- I'm curious what it would mean to be at maximum efficiency for an arbitrary period of time.
If such a concept could be effectively developed, it would be very useful for certain kinds of utility-comparisons that are currently very difficult to perform. For example, suppose I'm trying to decide which hobbies or which friends to hang on to after a move. These are really difficult things to rationally compare...the availability heuristic tends to make us disproportionately remember the most startling or sensually rich features of our activities, and these may not be a good guide to what is truly rewarding. I recently drove 6 hours round-trip to visit a modern art museum for 1 hour, and now, two months later, all I can remember is my strong emotional reaction to some of the artwork, plus one or two of the more interesting views from the rural highway. What I've forgotten (at an emotional level) is the boredom and nausea of the bumpy journey. Is it wise to rely on my heavily filtered impressions to properly weight the costs and benefits? Wouldn't I do better to focus on the experience of viewing the art, and then mentally dividing it by 7 before comparing it to other experiences, which will also need to be divided by their administrative support ratios ("ASRs")?
Another application would be figuring out how much time you could gain or lose, net of hidden costs, by changing your lifestyle. For example, at 3 am, as I write this post, I am not a very efficient typist or composer. Writing 4000 interesting words takes me 5 hours. Suppose I wake up at the same time every morning, and I need to go to bed at 2 am to be maximally efficient. Had I gone to bed an hour earlier, I might have been able to type 4000 interesting words in only 3 hours. With adequate sleep, i.e., at my MTP, I can write 4000 words in 4 hours; 3 for typing and 1 for sleeping. So my MTP can be expressed as 1000 words/hour. Without adequate sleep, I need 5 hours to write 4000 words, so I write at only 800 words / hour. My decision to stay up an extra hour has created a deadweight loss of 200 words, or, equivalently, 12 minutes at my MTP. If, on average, it takes me 3 minutes of homeostasis and maintenance for each minute of MTP, then I have wasted the equivalent of 12 * 3 = 36 real minutes of my life. This kind of analysis can be both a motivating tool and a way of figuring out which anti-akrasia tools are most effective.
MTP can break down when you are trying to have fun, rather than to accomplish a goal, but sometimes you can pay attention to how long you have to do something in order to get a roughly equivalent sense of satisfaction. Sometimes I amble through an expressive but highly arbitrary series of chords on my guitar, with the intention of airing whatever emotions might be burbling under my layers of conscious thought and also of exploring a bit of music theory. If I am overtired, or thirsty, or worried, I will be less able to focus on what I am playing, I wlll hit chords that are "wrong" in the sense of not fitting into even my loosely structured pattern more often, and it will take me longer to come up with a series of chords that actually produces catharsis and/or satisfies my musical curiosity.
Rationally Researching the Cost of Research
In order to do anything purposefully, you first have to know something about your situation; it wouldn't make any sense to take action with a perfectly blank set of Bayesian priors, because, on average, you would expect your action to have no effect. Unfortunately, gathering information about your situation is costly. Worse, since you are not a perfect rationalist, gathering information about how well you are gathering information is also costly. So far as I can tell, this phenomenon is endlessly recursive. There are certain well-defined situations where the cost of information is expected to converge; i.e., finding out how much you know about how much you know about how much you know will provide you with so little in the way of an increased chance of choosing the optimal action that you can safely engage in zero research (or epsilon research) at that layer of recursion and still get pretty good results.
So far as I can tell (irony not intended), these situations are few and far-between . No matter how accurate and complete you might think your model is, you could still be totally surprised by variables that you never expected to influence your predictions. Take climate change, for interest. You might correctly model the effects of CO2, dust, ash, sunspots, ocean currents, carbon dissolved in the ocean floor, albedo lost by melting ice, and geo-engineered reflectors, only to discover (too late) that a new kind of uber-popular roofing tile that happens to be pure black or pure white will dramatically change the Earth's albedo as the developing world urbanizes. It's practically impossible to tell how many variables you need to correctly model in order to generate arbitrarily accurate predictions, because in a chaotic system, even a single rogue variable can totally throw off your results. Or, it might not -- maybe, on balance, the more variables you consider, the more accurate your results will be. Or maybe it just looks like adding variables increases accuracy, but then the eleventh and twelfth variable suddenly turn out to be useless.
Many aspects of our personal lives might turn out to be nearly as chaotic as the Earth's climate -- you can gather data all week about how your new productivity plan is working out, and conclude that it's time to cut down on data-gathering and start re-investing an extra 5 minutes a week in actually getting stuff done, only to be surprised by a huge bout of akrasia that you could have prevented if you'd been on guard. You can spend hours each day gathering data on yourself all month and be pleased with your results, only to fail some crucial real-life test by a few percentage points that you could have easily passed if you'd just taken the most obvious strategy.
Are there any shortcuts at all here? Is anyone familiar with a literature on the recursive cost of information in a complex situation?
Anyway, thanks for your time, and your comments.