Optimal frequency for autoinvesting in taxable account
Optimal frequency for autoinvesting in taxable account
I'd like to set up automatic investing. For the purposes of dollarcost averaging or other "bad timing" and also taking into account transaction fees, what frequency is best? Monthly, weekly?
Investments would go to Vanguard Admiral Index funds (VTSAX, VTIAX, etc.) in a Vanguard account. I've always invested in batches.
Investments would go to Vanguard Admiral Index funds (VTSAX, VTIAX, etc.) in a Vanguard account. I've always invested in batches.
 retired@50
 Posts: 5850
 Joined: Tue Oct 01, 2019 2:36 pm
 Location: Living in the U.S.A.
Re: Optimal frequency for autoinvesting in taxable account
Twice per month.
I used the 7th and the 22nd. The 9th and 19th are also good.
Regards,
I used the 7th and the 22nd. The 9th and 19th are also good.
Regards,
This is one person's opinion. Nothing more.
Re: Optimal frequency for autoinvesting in taxable account
I just set it for every payday. Simple and easy.
Re: Optimal frequency for autoinvesting in taxable account
I think an optimal time is whenever you have the funds to purchase shares. We had monthly purchases for many years. Now quarterly, on months when dividends don't reinvest. There are no fees for purchasing mutual funds at Vanguard. Is Vanguard your broker?
Re: Optimal frequency for autoinvesting in taxable account
What's special about those dates?retired@50 wrote: ↑Fri Jan 15, 2021 11:00 am Twice per month.
I used the 7th and the 22nd. The 9th and 19th are also good.
Re: Optimal frequency for autoinvesting in taxable account
I buy weekly, but no particular reason. That creates a lot of lots, so not particularly friendly to TLH.
Other than your “savings amount” and “time in the market”, I am not aware of any optimal day or frequency.
Other than your “savings amount” and “time in the market”, I am not aware of any optimal day or frequency.
 retired@50
 Posts: 5850
 Joined: Tue Oct 01, 2019 2:36 pm
 Location: Living in the U.S.A.
Re: Optimal frequency for autoinvesting in taxable account
You'll need to read Jeremy Siegel's book called "Stocks For The Long Run" for the details. Head down to your local library and check out a copy.exodusNH wrote: ↑Fri Jan 15, 2021 12:26 pmWhat's special about those dates?retired@50 wrote: ↑Fri Jan 15, 2021 11:00 am Twice per month.
I used the 7th and the 22nd. The 9th and 19th are also good.
For more worthwhile books to read, see link: https://www.bogleheads.org/wiki/Books:_ ... nd_reviews
Regards,
This is one person's opinion. Nothing more.
Re: Optimal frequency for autoinvesting in taxable account
As soon as possible.
The thing is, if there were some sort of optimal time/date to invest, it would have been found and arbitraged away by institutions that have far more resources than you. The only thing you can control is time in the market.
The thing is, if there were some sort of optimal time/date to invest, it would have been found and arbitraged away by institutions that have far more resources than you. The only thing you can control is time in the market.
Re: Optimal frequency for autoinvesting in taxable account
Sounds complicated.retired@50 wrote: ↑Fri Jan 15, 2021 12:35 pmYou'll need to read Jeremy Siegel's book called "Stocks For The Long Run" for the details. Head down to your local library and check out a copy.exodusNH wrote: ↑Fri Jan 15, 2021 12:26 pmWhat's special about those dates?retired@50 wrote: ↑Fri Jan 15, 2021 11:00 am Twice per month.
I used the 7th and the 22nd. The 9th and 19th are also good.
For more worthwhile books to read, see link: https://www.bogleheads.org/wiki/Books:_ ... nd_reviews
Regards,
Just about anything will work fine if you automate it. Perhaps X days after each time you get paid? Or, once per month on day Y.
Re: Optimal frequency for autoinvesting in taxable account
The reason you autoinvest is to maintain discipline. There is no pattern of autoinvesting that is better or worse. No time of the day, week, or month. The strictly rational answer is the lump sum option. Invest when you can as early as you can. Total time in the market is the critical thing. But most of us are not rational so a fixed inflexible schedule is better.
Former brokerage operations & mutual fund accountant. I hate risk, which is why I study and embrace it.
 retired@50
 Posts: 5850
 Joined: Tue Oct 01, 2019 2:36 pm
 Location: Living in the U.S.A.
Re: Optimal frequency for autoinvesting in taxable account
Since my investing by date comment has intrigued some folks, here is what I was referring to.
Credit to Jeremy Siegel, "Stocks For The Long Run". Chapter 21  Calendar Anomalies  page 335  Figure 215 shown below.
Regards,
Credit to Jeremy Siegel, "Stocks For The Long Run". Chapter 21  Calendar Anomalies  page 335  Figure 215 shown below.
Regards,
This is one person's opinion. Nothing more.
Re: Optimal frequency for autoinvesting in taxable account
That chart looks like noise. Does he present error bars / confidence intervals? (I could imagine there's some effect related to paydays... but it's not obvious apart from maybe the 1st.)retired@50 wrote: ↑Fri Jan 15, 2021 5:03 pm Since my investing by date comment has intrigued some folks, here is what I was referring to.
Credit to Jeremy Siegel, "Stocks For The Long Run". Chapter 21  Calendar Anomalies  page 335  Figure 215 shown below.
Last edited by Charon on Sat Jan 16, 2021 1:04 am, edited 2 times in total.
 retired@50
 Posts: 5850
 Joined: Tue Oct 01, 2019 2:36 pm
 Location: Living in the U.S.A.
Re: Optimal frequency for autoinvesting in taxable account
I don't imagine confidence intervals apply. It's just historical data on a graph, not some kind of prediction.Charon wrote: ↑Sat Jan 16, 2021 12:54 amThat chart looks like noise. Does he present error bars / confidence intervals? (I could imagine there's some effect related to paydays... but it's not obvious apart from maybe the 1st.)retired@50 wrote: ↑Fri Jan 15, 2021 5:03 pm Since my investing by date comment has intrigued some folks, here is what I was referring to.
Credit to Jeremy Siegel, "Stocks For The Long Run". Chapter 21  Calendar Anomalies  page 335  Figure 215 shown below.
Regards,
This is one person's opinion. Nothing more.
Re: Optimal frequency for autoinvesting in taxable account
Confidence intervals apply to any data. All experimental or observational results have confidence intervals, for which the concept is in fact more robustly defined than for predictions.retired@50 wrote: ↑Sat Jan 16, 2021 1:00 amI don't imagine confidence intervals apply. It's just historical data on a graph, not some kind of prediction.Charon wrote: ↑Sat Jan 16, 2021 12:54 amThat chart looks like noise. Does he present error bars / confidence intervals? (I could imagine there's some effect related to paydays... but it's not obvious apart from maybe the 1st.)retired@50 wrote: ↑Fri Jan 15, 2021 5:03 pm Since my investing by date comment has intrigued some folks, here is what I was referring to.
Credit to Jeremy Siegel, "Stocks For The Long Run". Chapter 21  Calendar Anomalies  page 335  Figure 215 shown below.
Regards,
To expand a bit, the fact that the effect size is nearly always larger with the smaller data set is exactly how noise behaves, unlike signal. You could try to explain that by large timedependent shifts... but that's not well motivated, and would also suggest it's a useless guide moving forward since the future will be radically different in such a model.
Re: Optimal frequency for autoinvesting in taxable account
There is a user here (and I do not mean myself) who is fond of saying:
When should you invest the money? When you have it.
When should you withdraw the money? When you need it.
Next!
When should you invest the money? When you have it.
When should you withdraw the money? When you need it.
Next!
Re: Optimal frequency for autoinvesting in taxable account
I'll second the answer of "invest it when you have it".
With respect to Jeremy Siegel, I think that:
(a) The Dow Jones Industrial Average is a terrible index to represent the market
(b) Even if the observation was actionable in 2012 for timing purchases, there's no certainty that it remains so today.
I've plotted the distribution of DJIA and total market (as represented by VTSAX) dayoverday price changes over time periods shown. I don't think you'll find any compelling evidence that any one day(s) of the month is/are reliably better than others.
With respect to Jeremy Siegel, I think that:
(a) The Dow Jones Industrial Average is a terrible index to represent the market
(b) Even if the observation was actionable in 2012 for timing purchases, there's no certainty that it remains so today.
I've plotted the distribution of DJIA and total market (as represented by VTSAX) dayoverday price changes over time periods shown. I don't think you'll find any compelling evidence that any one day(s) of the month is/are reliably better than others.
 retired@50
 Posts: 5850
 Joined: Tue Oct 01, 2019 2:36 pm
 Location: Living in the U.S.A.
Re: Optimal frequency for autoinvesting in taxable account
Charon,Charon wrote: ↑Sat Jan 16, 2021 1:05 amConfidence intervals apply to any data. All experimental or observational results have confidence intervals, for which the concept is in fact more robustly defined than for predictions.retired@50 wrote: ↑Sat Jan 16, 2021 1:00 am
I don't imagine confidence intervals apply. It's just historical data on a graph, not some kind of prediction.
Regards,
To expand a bit, the fact that the effect size is nearly always larger with the smaller data set is exactly how noise behaves, unlike signal. You could try to explain that by large timedependent shifts... but that's not well motivated, and would also suggest it's a useless guide moving forward since the future will be radically different in such a model.
Why would a confidence interval be necessary (or relevant) if you're looking at the entire data set?
My (limited) understanding from my statistics class many years ago coincides with this link: https://www.investopedia.com/terms/c/co ... terval.asp . When sampling a data set, the confidence interval becomes relevant...No?
My feeling is that Mr. Siegel (the author of the book) isn't trying to make any sort of statement about whether the patterns of the past will hold in the future, and neither am I. It's nothing more than an observed pattern of price movements over time. Just like past performance of a mutual fund, it doesn't guarantee anything about the future.
To clarify my position to the OP, I didn't know about the chart posted above until after I had made nearly all of my investments. It was dumb luck that I happened to get paid on the 7th and 22nd of the month at my first job out of college. I stuck with that pattern for no real reason other than habit. The main message to take away from this thread is to invest early and often, and don't try to time the market.
Regards,
This is one person's opinion. Nothing more.
 ruralavalon
 Posts: 21390
 Joined: Sat Feb 02, 2008 10:29 am
 Location: Illinois
Re: Optimal frequency for autoinvesting in taxable account
I suggest every pay period.Miguelito wrote: ↑Fri Jan 15, 2021 10:53 am I'd like to set up automatic investing. For the purposes of dollarcost averaging or other "bad timing" and also taking into account transaction fees, what frequency is best? Monthly, weekly?
Investments would go to Vanguard Admiral Index funds (VTSAX, VTIAX, etc.) in a Vanguard account. I've always invested in batches.
"Everything should be as simple as it is, but not simpler."  Albert Einstein 
Wiki article link:Getting Started
Re: Optimal frequency for autoinvesting in taxable account
Fair questions. So let's take an example: the chart you put up claims that the 16th and 17th are different (or were in the past). But any time you're comparing two quantities, you need to know their uncertainties to know if they are significantly different. ("Significant" here is "statistically significant", which means basically "real", or more technically "very unlikely to be just due to chance". So it doesn't mean "large", the way it might in common parlance or when used in "clinically significant" in medicine.)retired@50 wrote: ↑Sat Jan 16, 2021 10:44 amCharon,Charon wrote: ↑Sat Jan 16, 2021 1:05 amConfidence intervals apply to any data. All experimental or observational results have confidence intervals, for which the concept is in fact more robustly defined than for predictions.retired@50 wrote: ↑Sat Jan 16, 2021 1:00 am
I don't imagine confidence intervals apply. It's just historical data on a graph, not some kind of prediction.
Regards,
To expand a bit, the fact that the effect size is nearly always larger with the smaller data set is exactly how noise behaves, unlike signal. You could try to explain that by large timedependent shifts... but that's not well motivated, and would also suggest it's a useless guide moving forward since the future will be radically different in such a model.
Why would a confidence interval be necessary (or relevant) if you're looking at the entire data set?
My (limited) understanding from my statistics class many years ago coincides with this link: https://www.investopedia.com/terms/c/co ... terval.asp . When sampling a data set, the confidence interval becomes relevant...No?
My feeling is that Mr. Siegel (the author of the book) isn't trying to make any sort of statement about whether the patterns of the past will hold in the future, and neither am I. It's nothing more than an observed pattern of price movements over time. Just like past performance of a mutual fund, it doesn't guarantee anything about the future.
Regards,
If one fossil is dated as 2 million years old, and another is dated as 2 million years plus one day old, is the second one actually older? You can't tell unless you know the uncertainty on the measurements. If the dating method has an uncertainty of 100,000 years, then no, there is no statistically significant difference between the two dates. (If the dating method somehow had an uncertainty of one minute, then the difference of one day certainly would be significant.)
So look at the chart that jsprag posted. That's a "boxandwhisker" plot, where the central values (medians or means) are horizontal lines, and the boxes contain some reasonable amount of the data. So the boxes are confidence intervals (maybe including 50% of the data  it's common to plot 25th percentile to 75th percentile, but if this were going to publication I'd ask for that to be explicitly stated ). The "whiskers" are another confidence interval of higher significance, looks like maybe 95%. Then outlier data points are individually plotted.
Consider the 16th and 17th. jsprag's plot agrees with Siegel's in that the 16th is higher than the 17th... in central value. But is the difference significant? Not at all. The boxes have substantial overlap, let alone the whiskers. The data shows no difference.
One must show a pattern isn't just noise to claim it's real. That involves getting uncertainties or confidence intervals on things. And if you're measuring a real signal, then the more data you collect the more you should home in on the correct value of the signal  that is, more data leads to about the same effect size. But noise gets relatively less important as you collect more data, so noise gets smaller when your data set gets larger. In Siegel's data, his claimed effect gets smaller with more data  a sign that it's just noise, and not signal, even without running any statistical tests.
Re: Optimal frequency for autoinvesting in taxable account
I guess one more thing to add. From a data set you can conclude that the sample mean return on the 16th is indeed different from the sample mean return on the 17th. What I'm saying is that this doesn't imply that the population mean is different on those days (i.e., that this is a real effect having anything to do with those days, rather than an artifact of your sample). If you happened to choose some short NBA players and some tall random guys off the street and compared sample means, you could conclude that the average NBA player is shorter than the average man. And it would be true of your sample... but not a useful thing to conclude.retired@50 wrote: ↑Sat Jan 16, 2021 10:44 am My (limited) understanding from my statistics class many years ago coincides with this link: https://www.investopedia.com/terms/c/co ... terval.asp . When sampling a data set, the confidence interval becomes relevant...No?
Re: Optimal frequency for autoinvesting in taxable account
To the OP's actual question, I practice "regret minimization" where I contribute to my Roth on a different day than my paydays (when I contribute to my work retirement accounts). Spreading things out a bit means I haven't put all my contributions for the month (or indeed year) in on a single day... after which the market could drop. (Or it could go up. But most people, including me, feel loss more strongly than gain.)
If you assume the market always goes up on average, then the optimal time contribute is "as early as possible". I.e., do all your Roth contributions on Jan. 1, if you can afford to. But I don't mind paying a very small overhead to minimize my regret.
If you assume the market always goes up on average, then the optimal time contribute is "as early as possible". I.e., do all your Roth contributions on Jan. 1, if you can afford to. But I don't mind paying a very small overhead to minimize my regret.
Last edited by Charon on Sun Jan 17, 2021 11:24 am, edited 1 time in total.
 retired@50
 Posts: 5850
 Joined: Tue Oct 01, 2019 2:36 pm
 Location: Living in the U.S.A.
Re: Optimal frequency for autoinvesting in taxable account
What if the sample size is the same as the population size? In other words, you sample the entire population. Then what?Charon wrote: ↑Sat Jan 16, 2021 11:57 pmI guess one more thing to add. From a data set you can conclude that the sample mean return on the 16th is indeed different from the sample mean return on the 17th. What I'm saying is that this doesn't imply that the population mean is different on those days (i.e., that this is a real effect having anything to do with those days, rather than an artifact of your sample). If you happened to choose some short NBA players and some tall random guys off the street and compared sample means, you could conclude that the average NBA player is shorter than the average man. And it would be true of your sample... but not a useful thing to conclude.retired@50 wrote: ↑Sat Jan 16, 2021 10:44 am My (limited) understanding from my statistics class many years ago coincides with this link: https://www.investopedia.com/terms/c/co ... terval.asp . When sampling a data set, the confidence interval becomes relevant...No?
Regards,
This is one person's opinion. Nothing more.
Re: Optimal frequency for autoinvesting in taxable account
Every Monday auto invest for me.
sptm 
vigi 
xbt/eth

 Posts: 695
 Joined: Fri Nov 29, 2019 11:25 am
Re: Optimal frequency for autoinvesting in taxable account
That's why I love having after tax in my 401k
It automatically invests for me. All I have to do is roll it over to roth
It automatically invests for me. All I have to do is roll it over to roth

 Posts: 7700
 Joined: Mon May 18, 2009 5:57 pm
Re: Optimal frequency for autoinvesting in taxable account
First day of the month for money from paycheck because it's easy to remember since I'm already updating spreadsheets and paying bills
Otherwise from things like RSU's and ESPP, it's however much time it takes from the time they appear in my account, are sold, settled then transferred.
For anything else, it's as soon as it appears in my checking account plus a day or so to transfer
The fact that I'm saving whenever and as much as I can is more important to me than the exact timetable/frequency/whatever...
Otherwise from things like RSU's and ESPP, it's however much time it takes from the time they appear in my account, are sold, settled then transferred.
For anything else, it's as soon as it appears in my checking account plus a day or so to transfer
The fact that I'm saving whenever and as much as I can is more important to me than the exact timetable/frequency/whatever...
Re: Optimal frequency for autoinvesting in taxable account
Deleted
Last edited by L82GAME on Sun Jan 17, 2021 7:26 am, edited 1 time in total.
"Still I am learning."  Michelangelo
Re: Optimal frequency for autoinvesting in taxable account
You're right  those details are importantCharon wrote: ↑Sat Jan 16, 2021 11:43 pm So look at the chart that jsprag posted. That's a "boxandwhisker" plot, where the central values (medians or means) are horizontal lines, and the boxes contain some reasonable amount of the data. So the boxes are confidence intervals (maybe including 50% of the data  it's common to plot 25th percentile to 75th percentile, but if this were going to publication I'd ask for that to be explicitly stated ). The "whiskers" are another confidence interval of higher significance, looks like maybe 95%. Then outlier data points are individually plotted.
 Central values are medians (Q2)
 Box is 25^{th} (Q1) to 75^{th} (Q3) percentiles
 Whiskers extend to furthest data point that is within 1.5 x the interquartile range above or below the box (where IQR = Q3Q1)
 HMSVictory
 Posts: 254
 Joined: Sun Nov 01, 2020 7:02 am
Re: Optimal frequency for autoinvesting in taxable account
Once per month. Keep it simple. I do the 15th.
Stay the course!
Re: Optimal frequency for autoinvesting in taxable account
+1
There is no other right answer.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
Re: Optimal frequency for autoinvesting in taxable account
Paid every two weeks on Friday. Auto invest into taxable the following Monday.
I think vanguard only has so many options for auto invest frequency
I think vanguard only has so many options for auto invest frequency

 Posts: 7700
 Joined: Mon May 18, 2009 5:57 pm
Re: Optimal frequency for autoinvesting in taxable account
I do the same, but not auto. I do it manually. This allows me to invest exactly the money I have to invest, not less and not more. I invest everything extra after accounting for cash obligations due before the next payday.
Last edited by Triple digit golfer on Sun Jan 17, 2021 9:24 am, edited 1 time in total.

 Posts: 7700
 Joined: Mon May 18, 2009 5:57 pm
 ruralavalon
 Posts: 21390
 Joined: Sat Feb 02, 2008 10:29 am
 Location: Illinois
Re: Optimal frequency for autoinvesting in taxable account
+ 3
"Everything should be as simple as it is, but not simpler."  Albert Einstein 
Wiki article link:Getting Started
Re: Optimal frequency for autoinvesting in taxable account
This is a point well taken. If anyone is interested in actually getting confidence intervals on the medians to see if they're really different, I recommend a nonparametric bootstrap.
Edit: if you're interested in estimating the confidence intervals without having the underlying data or doing the bootstrap, you can assume a normal distribution and divide jsprag's 50% boxes by 3.5. (He has 108 data points for most days, and 50% is about 0.67σ. The standard error of the mean is the sample standard deviation/sqrt(N) for N data points, and sqrt(108)*0.67 is about 7. 95% confidence is about 2σ.)
Also keep in mind the Bonferroni correction if you're doing multiple comparisons. The above applies to comparing one specific day to one other specific day. If you start comparing all days hunting for one comparison that's significant, you need to use much larger error bars (equivalent to much higher confidence levels).
Obviously the scale of the chart is set by the outliers, so it's a bit hard to see... but it really doesn't look like any of the differences are significant.
Last edited by Charon on Sun Jan 17, 2021 12:38 pm, edited 2 times in total.
Re: Optimal frequency for autoinvesting in taxable account
At the risk of digressing too much from the thread topic... this is an interesting question. In some contexts it never shows up (one can never perform all possible experiments to measure the mass of an electron, for example), but it does show up in cosmology. What if you have catalogued all possible quasars in the entire (observable) universe? Isn't the uncertainty then zero?retired@50 wrote: ↑Sun Jan 17, 2021 1:06 am What if the sample size is the same as the population size? In other words, you sample the entire population. Then what?
The answer is no, because we assume that there is some underlying true distribution from which our sample  even if it's the entire population  is drawn. In cosmology the resulting uncertainty is called "cosmic variance".
To return to the investing topic, the hypothesis being tested is "does it matter on which day of the month you invest?" This assumes that there's some true underlying distribution of daily returns  maybe they're all the same, maybe some are different. Even if you look at all the data on hand, there's uncertainty because it's a random realization of that underlying distribution. (In this case you're not even running into cosmic variance  you know you'll have more and different data in the future.)
You could argue that the past has nothing to do with the future  there is no underlying distribution, or it varies with time in completely unconstrained and unpredictable ways. But then there's absolutely no justification in taking averages or medians, or comparing one historical year to another. Means and medians are estimators of parameters in an underlying distribution.

 Posts: 662
 Joined: Wed May 09, 2018 12:26 pm
Re: Optimal frequency for autoinvesting in taxable account
I buy everyday via vanguard automated scheduled trades. But thinking whether to move to monthly once trade since daily is generating too many transactions
Re: Optimal frequency for autoinvesting in taxable account
OptimaL
Whenever You Have The Monies
To Invest
Whenever You Have The Monies
To Invest
"One does not accumulate but eliminate. It is not daily increase but daily decrease. The height of cultivation always runs to simplicity" –Bruce Lee

 Posts: 783
 Joined: Mon Mar 02, 2020 5:33 pm
Re: Optimal frequency for autoinvesting in taxable account
Nothing's wrong with investing as frequently as you have money available, but have anyone run into any practical complications due to having too many tax lots?
Re: Optimal frequency for autoinvesting in taxable account
The answer is invest as much as possible as early as possible.Miguelito wrote: ↑Fri Jan 15, 2021 10:53 am I'd like to set up automatic investing. For the purposes of dollarcost averaging or other "bad timing" and also taking into account transaction fees, what frequency is best? Monthly, weekly?
Investments would go to Vanguard Admiral Index funds (VTSAX, VTIAX, etc.) in a Vanguard account. I've always invested in batches.
The frequency doesn't matter. If it's up to you, invest everything you have on the 1st of the month, or the 31st of the previous month, or better yet last week, last month, last year.