Case studies

[TBS]

the best strategy based on statistics (though none of these consider exchange rates as far as I know)

The currency issue doesn't make a big material difference tbh. E.g. lump sum investing versus DCA over 12 months into MSCI ACWI (total return, incl dividends) for all monthly starting points between 1988 and 2022:

USD-based
- LSI wins 300 out of 417 months (72% of the time)
- Average out-performance of LSI vs DCA is 3.6%

JPY-based
- LSI wins 283 out of 417 months (68% of the time)
- Average out-performance of LSI vs DCA is 3.8%

[TBS]

If anyone is interested in USA figures, MSCI gives slightly more back data to calculate with.

LSI vs 12 months DCA, MSCI USA index (total return, incl dividends) for all monthly starting points between Feb 1971 and Aug 2022:

USD-based
- LSI wins 466 out of 620 months (75% of the time)
- Average out-performance of LSI vs DCA is 4.7%

JPY-based
- LSI wins 410 out of 620 months (66% of the time)
- Average out-performance of LSI vs DCA is 3.9%

[TokyoBoglehead]

If anyone is interested in USA figures, MSCI gives slightly more back data to calculate with.

LSI vs 12 months DCA, MSCI USA index (total return, incl dividends) for all monthly starting points between Feb 1971 and Aug 2022:

USD-based
- LSI wins 466 out of 620 months (75% of the time)
- Average out-performance of LSI vs DCA is 4.7%

JPY-based
- LSI wins 410 out of 620 months (66% of the time)
- Average out-performance of LSI vs DCA is 3.9%

Lovely piece of data. Is this all in an Excel file?

Can you overlay with the strength of the yen to see generally if the trend still holds when the yen is low?

[northSaver]

Yes, this is great data. I was wondering if there was anything out there for yen-based investors. Thanks TBS
Is it possible to provide the source of these calculations? I wouldn't mind digging into it a bit deeper if it's in the public domain.

[TokyoWart]

the best strategy based on statistics (though none of these consider exchange rates as far as I know)

The currency issue doesn't make a big material difference tbh. E.g. lump sum investing versus DCA over 12 months into MSCI ACWI (total return, incl dividends) for all monthly starting points between 1988 and 2022:

USD-based
- LSI wins 300 out of 417 months (72% of the time)
- Average out-performance of LSI vs DCA is 3.6%

JPY-based
- LSI wins 283 out of 417 months (68% of the time)
- Average out-performance of LSI vs DCA is 3.8%

I completely agree with you about LSI being more likely to give a better outcome than spacing out an investment over 12 months (and I think the statistics are even more in favor of LSI when you space out the investment over more than 12 months) but I don't understand where the USD-based and JPY-based numbers are coming from here. Can you help me? The biggest pain point for me right now is converting JPY to USD so I can send the funds over to the US to buy international (i.e. non-US) index funds and doing so at current exchange rates which are depressed if you look at purchase power parity between yen and dollars. I'm still remembering the happy days when 800,000 yen converted to $10.000.

[TBS]

Is it possible to provide the source of these calculations? I wouldn't mind digging into it a bit deeper if it's in the public domain.

They are just quick calculations I did myself. The total returns index data (USD denominated) comes from MSCI. The USD-JPY FX data comes from FRED. Here are all the results: ACWI, USA

Is this all in an Excel file?

Can you overlay with the strength of the yen to see generally if the trend still holds when the yen is low?

It is not in an Excel file, but you can download the CSV results from pastebin and open them in Excel.

The first graph shows what the USD-JPY FX rate did over the period of the USA calcs. I'll leave you to judge when the yen is low, and whether that's even the case today.

The second graph shows how the outperformance of LSI swings around. Anything below the green line is DCA winning from that investment starting point.

but I don't understand where the USD-based and JPY-based numbers are coming from here. Can you help me?

The USD-based numbers are simply LSI vs DCA comparisons calculated using the USD-denominated index (i.e. buying units in USD, and measuring the outcome in USD). The JPY numbers are calculated using the index denominated in yen, that's the only difference.

The biggest pain point for me right now is converting JPY to USD so I can send the funds over to the US to buy international (i.e. non-US) index funds and doing so at current exchange rates which are depressed if you look at purchase power parity between yen and dollars. I'm still remembering the happy days when 800,000 yen converted to $10.000.

Broadly if you care about the outcome of your investment in yen (i.e. how much yen capital gain you get) there shouldn't be any big difference between i) investing in yen in a Japanese mutual fund/ETF, or ii) doing the conversion to another currency yourself (whether that's USD, EUR (e.g. Irish Vanguard ETFs), or GBP for something UK listed), then converting back to yen again yourself when you sell. Sorry I'm a little confused by what you are asking?

[TokyoWart]

Broadly if you care about the outcome of your investment in yen (i.e. how much yen capital gain you get) there shouldn't be any big difference between i) investing in yen in a Japanese mutual fund/ETF, or ii) doing the conversion to another currency yourself (whether that's USD, EUR (e.g. Irish Vanguard ETFs), or GBP for something UK listed), then converting back to yen again yourself when you sell. Sorry I'm a little confused by what you are asking?

My question probably comes down to wanting to time the currency exchange market but I think it applies whether buying a fund in Japan which includes securities that were denominated in currencies other than yen (since the yen is down against most currencies, not just USD) or doing it by buying the fund in the US with USD after exchanging into yen. My understanding of the studies that show lump sum investing (LSI) is better 2/3rds of the time than averaging in is that they are usually same currency studies in that same currency market (e.g. USD buying US stocks, not JPY buying US stocks).

For instance:

NWM study is using USD for US indexes:
https://www.northwesternmutual.com/life ... investing/

Vanguard study used CRISPS US market data:
https://www.forbes.com/sites/kristinmck ... 4c4d87fa5b

This later summary from Vanguard includes a Figure & where some cross-currency (e.g. Global in euros or GBP) investing examples but nothing for yen based investors:

https://corporate.vanguard.com/content/ ... r_cash.pdf

Are you aware of any studies specific to the cross-currency LSI problem?

[TBS]

My question probably comes down to wanting to time the currency exchange market but I think it applies whether buying a fund in Japan which includes securities that were denominated in currencies other than yen (since the yen is down against most currencies, not just USD) or doing it by buying the fund in the US with USD after exchanging into yen. My understanding of the studies that show lump sum investing (LSI) is better 2/3rds of the time than averaging in is that they are usually same currency studies in that same currency market (e.g. USD buying US stocks, not JPY buying US stocks).

For instance:

NWM study is using USD for US indexes:
https://www.northwesternmutual.com/life ... investing/

Vanguard study used CRISPS US market data:
https://www.forbes.com/sites/kristinmck ... 4c4d87fa5b

This later summary from Vanguard includes a Figure & where some cross-currency (e.g. Global in euros or GBP) investing examples but nothing for yen based investors:

https://corporate.vanguard.com/content/ ... r_cash.pdf

Are you aware of any studies specific to the cross-currency LSI problem?

Ahh ok, got you now. This was basically the same question NorthSaver posed. There are likely some cross-currency studies, but today I thought it would take more time to look for them (and find something yen-relevant specifically) than to do some quick calculations like I posted.

My layman, highly uninformed opinion is the following. Stocks are value generating assets and long-term stock market indices grow exponentially. Anyone holding money out of the market to invest at a later date is betting against an exponential curve.

Currency movements - at least between the major economies which make up the majority of world stock market cap - are not the same. Any drifts between major currencies over the long term tend to be much smaller than the growth of markets.

Therefore I fully expected before I did the calcs that adding currency into the mix would not significantly alter the 2/3rds figure, and for the limited study I did, that turned out to be the case.

[northSaver]

Thank you TBS for providing the raw data. I will dig into it this week when I have more time. I guess you processed it with code such as Python and not in Excel?

A quick glance at the charts seems to indicate that yen-based DCA outperforms LSI more often than not when USD.JPY is falling (yen getting stronger), as I would expect. But I need to confirm it with the data. I'm also particularly interested in the results when USD.JPY is greater than 140, especially after 1986 when it dropped through 150 and never really got above it since then. This would reduce the dataset considerably though.

Thanks again , lots to play with

[TBS]

I guess you processed it with code such as Python and not in Excel?

Yes I used Python.

...This would reduce the dataset considerably though.

Yep. I think a common trap people fall into is wringing the data for correlations, misattributing them as causal, then being disappointed when they don't repeat in future

Or if they do repeat in future, they happen for slightly different values meaning you sell/buy too early or too late, and end up losing out.