I’m always amused by statements like “This post is not ‘investment advice’”, followed by exactly that. I understand that it’s necessary ass-covering, but the transparency of it is hilarious.
That said, this is interesting. I don’t plan on buying in, but I might start watching the market. I didn’t realize BTC had survived a previous crash. That actually improves my estimate of its long-term survival chances.
This is, IIRC the long aftermath of the 3rd major bubble. One in June 2011 (to ~$30), one in April 2013 (To ~$220) and one in November 2013 (To ~$1100).
Personally, I’m not expecting a major recovery until the protocol hits the next halving of the mining rate, which is July of 2016 on the current mining timetable. In the mean time, I’m dollar cost averaging my investment in Bitcoin, and stacking up whatever I can get.
What’s the point of dollar cost averaging? Why not just pick a % of your asset allocation that you want in Bitcoin and rebalance from 0% up to that ASAP? I see this as a special case of the virtue of rebalancing as often as possible.
It’s just a hedge for if you’re risk-averse or if you’re worried that market fluctuations will negatively influence your behavior about when to rebalance or invest. Here’s a Vanguard study comparing the two historically if you’re curious.
The benefit is that you get more shares during the times when it is down, the mathematics helps reduce your average cost.
For example, lets say I take $3000 and buy a stock all at once at $50. I get 60 shares.
Now instead, what if I buy $1000 each at 3 different times, once at $40, once at $50, and once at $60. I end up with 25+20+16.66 shares = 61.66 shares, even though the average price I bought at was identical.
This is generally a good idea, whether one is buying stocks or Bitcoin or anything else.
It works similar to reallocating. For example, lets say you wanted to keep 10% of your net worth in Bitcoin (or anything else). If Bitcoin doubles in price, you now are misallocated, and have close to 20% of your net worth in it (if other things stayed the same), so you would sell some. If it dropped by 50%, you would have a little over 5% of your new net worth in it, so you would need to buy some. This helps you to, on average, buy low and sell high, even without really knowing what you are doing. You don’t want to reallocate constantly, due to trading fees, but you need to do it sometimes, to gain the benefit.
When I see the concept of dollar cost averaging my math intuition module throws up a big red “This Is Clearly Wrong” sign. I never seem to have that thought when I have the time and inclination to tease out what’s wrong and write a clear explanation of why it’s BS (or find out that it’s not).
Today is no exception. But here are some pointers my math intuition module is producing which say “investigate this, it will show you what’s wrong”:
If you flip a coin and invest the lump sum $3000 at either $40, $50, or $60 with equal probability, your expected value is 61.66 shares, not 60.
The “average price” should be the harmonic mean, not the arithmetic mean, and buying at the harmonic mean gets you 61.66 shares.
If you have the option of buying $3000 worth at $50, that doesn’t mean you could switch to instead buying at a non-zero-variance-distribution with arithmetic mean $50 over time.
DCA lowers risk, while keeping the same EV. And the most common alternative, trying to time the market, has a long history of miserable failure by virtually all investors.
For a normal person who’s saving money off their paycheque, DCA is superior to saving up a lump sum and investing that. This is true for exactly the same reason that DCA is inferior to a lump sum in the case where you’re investing a lump sum—it gets you into the market faster, and stocks outperform cash.
For that argument for DCA to go through, we must justify that it’s the correct argument to choose from these three:
For example, lets say I take $3000 and buy a stock all at once at $50 for 1 share. I get 60 shares. Now instead, what if I buy $1000 each at 3 different times, once at $40 for 1 share, once at $50 for 1 share, and once at $60 for 1 share. I end up with 25+20+16.66 shares = 61.66 shares, even though the average price per share I bought at was identical. (original argument)
For example, lets say I take $3000 and buy a stock all at once at 0.02 shares for $1. I get 60 shares. Now instead, what if I buy $1000 each at 3 different times, once at 0.025 shares for $1, once at 0.02 shares for $1, and once at 0.0167 shares for $1. I end up with 25+20+16.66 shares = 61.66 shares, because the average shares per dollar I bought at was 0.02057 which is better than 0.02. (original argument with prices preserved, “average metric was the same” changed, and conclusion changed)
For example, lets say I take $3000 and buy a stock all at once at 0.02 shares for $1. I get 60 shares. Now instead, what if I buy $1000 each at 3 different times, once at 0.025 shares for $1, once at 0.02 shares for $1, and once at 0.015 shares for $1. I end up with 25+20+15 shares = 60 shares, because the average shares per dollar I bought at was 0.02 which is identical to 0.02. (original argument with prices changed, “average metric was the same” preserved, and conclusion changed)
Off the top of my head whether the DCA (dollar cost averaging) works depends on the persistence of the underlying time series (returns on the asset you’re buying).
If the asset returns are following a random walk, DCA is useless. It won’t hurt you and it won’t help you.
If the asset returns are persistent (momentum dominates), DCA will hurt you and decrease your performance.
If the asset returns are anti-persistent (mean-reversion dominates), DCA will help you.
Thing about the stock market is, if either momentum or mean-reversion dominated, people would use that in trading algorithms and destroy the phenomenon. Over the long run, it can be safely assumed to be neither, so DCA doesn’t hurt you(other than perhaps by delaying your investments, if you’re thinking of streaming a lump-sum in over time instead of investing it all right away, but most people invest from paycheques instead of from lump sums), but it does lower variance.
Thing about the stock market is, if either momentum or mean-reversion dominated, people would use that in trading algorithms and destroy the phenomenon.
That may be the single most colloquial academic paper I’ve ever seen. They do lay out a decent case, but remember that they’re discussing a different kind of momentum than we are—we’re discussing momentum of market returns, they’re discussing momentum in relative ranking of different securities. Also, I tend to take it as given that if a simple stock-trading strategy produced returns that were that superior, the hedge fund crowd would have jumped on it with both feet by now. Some of the hedge fund strategies I’ve seen have exploited far smaller inefficiencies to make significant returns.
Compared to what?
Compared to saving up a lump sum and investing that. And it sure returns better than just not saving at all, which is the usual default of most would-be investors.
In practice, DCA is usually used by retail advisors(of which I am one) as a psychological argument that investors should ignore short-term market turmoil in their long-term investments. Frankly, any argument that makes retail investors stop buying high and selling low is doing the lord’s work, and DCA is even mildly accurate.
It’s the same argument that the above paper is making, just in the opposite circumstances.
DCA says investing into a down market improves your returns, and the best way to systematically do so is to to just always be investing, which is accurate and a useful psychological argument besides.
I’m not speaking of trying to time investments. I’m speaking retrospectively—if you can invest at a time that is a local minimum in retrospect, that money will grow faster.
Also, I’m not a big believer in the EMH—it’s valid over a period of seconds and over a period of decades, but in between, psychological effects and the joys of market timing(i.e., knowing an investment is crap, but riding the bubble anyways) can swamp EMH easily.
I’m speaking retrospectively—if you can invest at a time that is a local minimum in retrospect, that money will grow faster.
It’s certainly true that a time machine can produce outstanding investment returns :-/ but I don’t see what does it have to do with DCA.
I’m not a big believer in the EMH
That’s perfectly fine, but if you don’t believe in EMH and believe that DCA improves your returns, then this necessarily means that you believe that markets are mean-reverting and in that case there are better than DCA ways to take advantage of it.
I’m coming at this from two angles at once, so my points are a touch scattered.
1) The basic math. This says that putting money in as you get it is the best way to invest, because the market goes up over time, and that dominates any gains from DCA or market timing. When investing a lump sum there’s a small reduction in risk from DCA, but not enough to make it worthwhile in general.
2) The fact that I advise middle-class people on how to invest, and that this advice needs to be built primarily on psychology and not math. DCA is a good strategy for increasing real-world investment returns for normal people, because it convinces them that a down market is not the end of the world, and in some cases it convinces them to save at all(which far too many don’t, whether out of fear or short-sighted spending goals). Also, normal people tend to invest a portion of their paycheques, and that is a payment stream that leads naturally to DCA, even if the ostensible benefits of DCA don’t materialize.
The basic math. This says that putting money in as you get it is the best way to invest
I am sorry, “the basic math” says no such thing. It’s your strong assumptions (“the markets always go up”) which say this, assumptions I don’t consider to be self-evident.
primarily on psychology and not math
It would be helpful to clearly distinguish what you believe is the true state of the world (“math”) and what you think not-quite-rational people need to be told in order to do what you want to make them do (“psychology”).
I do not assume that “the markets always go up”. I merely claim that historically, the market has generally gone up faster than cash over the long run. This was the whole point of the paper linked above that said lump-sum beats DCA, and the same is true when investing a steady stream of income—you will generally do better in the market, so to maximize EV, invest right away when you get the money.
The post above was an attempt to draw that distinction. Are there any points I wasn’t sufficiently clear about?
That is not true. You advise your clients to act a way which implies a specific forecast: that the “markets” will, in the future, outperform other assets, e.g. cash, over the relevant holding period.
Agreed, and I’ve also been dollar cost averaging it. From the 400s on down.
The risk is that the previous bubble was the last one and it never recovers. The upside is that the previous bubble was another in the string of booms that must occur in something that is growing exponentially from 0.
Whats even funnier is how many different people that there are that you can subscribe to and pay a monthly fee to, in tons of different financial markets, who also “don’t give investment advice”. :)
You should mentally replace this statement with “you are responsible for your own actions, not me”.
Indeed, Bitcoin has survived a previous crash (severe bear market), several times in fact. Buying now is essentially a bet it will survive again. That bet will either pay off, or it wont.
I’m always amused by statements like “This post is not ‘investment advice’”, followed by exactly that. I understand that it’s necessary ass-covering, but the transparency of it is hilarious.
That said, this is interesting. I don’t plan on buying in, but I might start watching the market. I didn’t realize BTC had survived a previous crash. That actually improves my estimate of its long-term survival chances.
This is, IIRC the long aftermath of the 3rd major bubble. One in June 2011 (to ~$30), one in April 2013 (To ~$220) and one in November 2013 (To ~$1100).
Personally, I’m not expecting a major recovery until the protocol hits the next halving of the mining rate, which is July of 2016 on the current mining timetable. In the mean time, I’m dollar cost averaging my investment in Bitcoin, and stacking up whatever I can get.
What’s the point of dollar cost averaging? Why not just pick a % of your asset allocation that you want in Bitcoin and rebalance from 0% up to that ASAP? I see this as a special case of the virtue of rebalancing as often as possible.
It’s just a hedge for if you’re risk-averse or if you’re worried that market fluctuations will negatively influence your behavior about when to rebalance or invest. Here’s a Vanguard study comparing the two historically if you’re curious.
Yes, it is similar to rebalancing.
The benefit is that you get more shares during the times when it is down, the mathematics helps reduce your average cost.
For example, lets say I take $3000 and buy a stock all at once at $50. I get 60 shares. Now instead, what if I buy $1000 each at 3 different times, once at $40, once at $50, and once at $60. I end up with 25+20+16.66 shares = 61.66 shares, even though the average price I bought at was identical.
This is generally a good idea, whether one is buying stocks or Bitcoin or anything else.
It works similar to reallocating. For example, lets say you wanted to keep 10% of your net worth in Bitcoin (or anything else). If Bitcoin doubles in price, you now are misallocated, and have close to 20% of your net worth in it (if other things stayed the same), so you would sell some. If it dropped by 50%, you would have a little over 5% of your new net worth in it, so you would need to buy some. This helps you to, on average, buy low and sell high, even without really knowing what you are doing. You don’t want to reallocate constantly, due to trading fees, but you need to do it sometimes, to gain the benefit.
When I see the concept of dollar cost averaging my math intuition module throws up a big red “This Is Clearly Wrong” sign. I never seem to have that thought when I have the time and inclination to tease out what’s wrong and write a clear explanation of why it’s BS (or find out that it’s not).
Today is no exception. But here are some pointers my math intuition module is producing which say “investigate this, it will show you what’s wrong”:
If you flip a coin and invest the lump sum $3000 at either $40, $50, or $60 with equal probability, your expected value is 61.66 shares, not 60.
The “average price” should be the harmonic mean, not the arithmetic mean, and buying at the harmonic mean gets you 61.66 shares.
If you have the option of buying $3000 worth at $50, that doesn’t mean you could switch to instead buying at a non-zero-variance-distribution with arithmetic mean $50 over time.
DCA lowers risk, while keeping the same EV. And the most common alternative, trying to time the market, has a long history of miserable failure by virtually all investors.
It’s canonical investment advice for a reason.
Ander’s claim, which I see repeated a lot, seems to be that it is positive EV rather than neutral. That’s the bit that raises my hackles.
For a normal person who’s saving money off their paycheque, DCA is superior to saving up a lump sum and investing that. This is true for exactly the same reason that DCA is inferior to a lump sum in the case where you’re investing a lump sum—it gets you into the market faster, and stocks outperform cash.
Fleshing out my intuition.
For that argument for DCA to go through, we must justify that it’s the correct argument to choose from these three:
For example, lets say I take $3000 and buy a stock all at once at $50 for 1 share. I get 60 shares. Now instead, what if I buy $1000 each at 3 different times, once at $40 for 1 share, once at $50 for 1 share, and once at $60 for 1 share. I end up with 25+20+16.66 shares = 61.66 shares, even though the average price per share I bought at was identical. (original argument)
For example, lets say I take $3000 and buy a stock all at once at 0.02 shares for $1. I get 60 shares. Now instead, what if I buy $1000 each at 3 different times, once at 0.025 shares for $1, once at 0.02 shares for $1, and once at 0.0167 shares for $1. I end up with 25+20+16.66 shares = 61.66 shares, because the average shares per dollar I bought at was 0.02057 which is better than 0.02. (original argument with prices preserved, “average metric was the same” changed, and conclusion changed)
For example, lets say I take $3000 and buy a stock all at once at 0.02 shares for $1. I get 60 shares. Now instead, what if I buy $1000 each at 3 different times, once at 0.025 shares for $1, once at 0.02 shares for $1, and once at 0.015 shares for $1. I end up with 25+20+15 shares = 60 shares, because the average shares per dollar I bought at was 0.02 which is identical to 0.02. (original argument with prices changed, “average metric was the same” preserved, and conclusion changed)
Off the top of my head whether the DCA (dollar cost averaging) works depends on the persistence of the underlying time series (returns on the asset you’re buying).
If the asset returns are following a random walk, DCA is useless. It won’t hurt you and it won’t help you.
If the asset returns are persistent (momentum dominates), DCA will hurt you and decrease your performance.
If the asset returns are anti-persistent (mean-reversion dominates), DCA will help you.
Thing about the stock market is, if either momentum or mean-reversion dominated, people would use that in trading algorithms and destroy the phenomenon. Over the long run, it can be safely assumed to be neither, so DCA doesn’t hurt you(other than perhaps by delaying your investments, if you’re thinking of streaming a lump-sum in over time instead of investing it all right away, but most people invest from paycheques instead of from lump sums), but it does lower variance.
Some people disagree.
Compared to what?
That may be the single most colloquial academic paper I’ve ever seen. They do lay out a decent case, but remember that they’re discussing a different kind of momentum than we are—we’re discussing momentum of market returns, they’re discussing momentum in relative ranking of different securities. Also, I tend to take it as given that if a simple stock-trading strategy produced returns that were that superior, the hedge fund crowd would have jumped on it with both feet by now. Some of the hedge fund strategies I’ve seen have exploited far smaller inefficiencies to make significant returns.
Compared to saving up a lump sum and investing that. And it sure returns better than just not saving at all, which is the usual default of most would-be investors.
In practice, DCA is usually used by retail advisors(of which I am one) as a psychological argument that investors should ignore short-term market turmoil in their long-term investments. Frankly, any argument that makes retail investors stop buying high and selling low is doing the lord’s work, and DCA is even mildly accurate.
Do note that the main author of that paper runs a hedge fund.
That’s apples and oranges, isn’t it? All you’re saying is that holding cash is less volatile in nominal terms :-)
That’s an entirely different claim from saying that DCA improves your returns (or your Sharpe ratio).
Noted.
It’s the same argument that the above paper is making, just in the opposite circumstances.
DCA says investing into a down market improves your returns, and the best way to systematically do so is to to just always be investing, which is accurate and a useful psychological argument besides.
This claim is not compatible with EMH.
I’m not speaking of trying to time investments. I’m speaking retrospectively—if you can invest at a time that is a local minimum in retrospect, that money will grow faster.
Also, I’m not a big believer in the EMH—it’s valid over a period of seconds and over a period of decades, but in between, psychological effects and the joys of market timing(i.e., knowing an investment is crap, but riding the bubble anyways) can swamp EMH easily.
It’s certainly true that a time machine can produce outstanding investment returns :-/ but I don’t see what does it have to do with DCA.
That’s perfectly fine, but if you don’t believe in EMH and believe that DCA improves your returns, then this necessarily means that you believe that markets are mean-reverting and in that case there are better than DCA ways to take advantage of it.
I’m coming at this from two angles at once, so my points are a touch scattered.
1) The basic math. This says that putting money in as you get it is the best way to invest, because the market goes up over time, and that dominates any gains from DCA or market timing. When investing a lump sum there’s a small reduction in risk from DCA, but not enough to make it worthwhile in general.
2) The fact that I advise middle-class people on how to invest, and that this advice needs to be built primarily on psychology and not math. DCA is a good strategy for increasing real-world investment returns for normal people, because it convinces them that a down market is not the end of the world, and in some cases it convinces them to save at all(which far too many don’t, whether out of fear or short-sighted spending goals). Also, normal people tend to invest a portion of their paycheques, and that is a payment stream that leads naturally to DCA, even if the ostensible benefits of DCA don’t materialize.
I am sorry, “the basic math” says no such thing. It’s your strong assumptions (“the markets always go up”) which say this, assumptions I don’t consider to be self-evident.
It would be helpful to clearly distinguish what you believe is the true state of the world (“math”) and what you think not-quite-rational people need to be told in order to do what you want to make them do (“psychology”).
I do not assume that “the markets always go up”. I merely claim that historically, the market has generally gone up faster than cash over the long run. This was the whole point of the paper linked above that said lump-sum beats DCA, and the same is true when investing a steady stream of income—you will generally do better in the market, so to maximize EV, invest right away when you get the money.
The post above was an attempt to draw that distinction. Are there any points I wasn’t sufficiently clear about?
That is not true. You advise your clients to act a way which implies a specific forecast: that the “markets” will, in the future, outperform other assets, e.g. cash, over the relevant holding period.
Quite right. Any financial advisor who does not do so is being grossly irresponsible.
Agreed, and I’ve also been dollar cost averaging it. From the 400s on down.
The risk is that the previous bubble was the last one and it never recovers. The upside is that the previous bubble was another in the string of booms that must occur in something that is growing exponentially from 0.
Indeed!
Whats even funnier is how many different people that there are that you can subscribe to and pay a monthly fee to, in tons of different financial markets, who also “don’t give investment advice”. :)
You should mentally replace this statement with “you are responsible for your own actions, not me”.
Indeed, Bitcoin has survived a previous crash (severe bear market), several times in fact. Buying now is essentially a bet it will survive again. That bet will either pay off, or it wont.