Limitations
of the Proposed Study
The empirical analysis method selected for the bulk of this
research permits comprehensive, realistic simulation of the true effects
of the three techniques studied. The
results of the simulations were also verified using actual historical
returns from multiple investment types over variable periods up to
thirty years and as far back in time as the year 1900.
Thus, the applicability of this research is generally unlimited.
However, there are a few limitations identified both prior to and
during the research process. These
limitations and their significance are discussed below.
The results of this research are limited, to an extent, in terms
of the algorithms used in the spreadsheets.
First, the spreadsheets do not account for holding and investing
money that will be used to pay taxes until the fifteenth of April.
If the spreadsheets were modified to incorporate this real world
benefit, the trends produced would not change;
only the absolute numbers would change, and the performance would
get slightly better. Paying
for taxes with funds outside of the accounts is simulated by the
non-taxed graphs; non-taxed
graphs also simulate tax-deferred accounts such as IRAs and 401(k)’s.
Secondly, capital gains and losses distributed to shareholders
due to normal management of mutual funds are ignored for all three
techniques. In other words,
the gains and losses from the investment manager buying and selling
equities held by the fund are ignored.
The net effect of this omission is a slight degradation in the
assessment of taxed accounts. If
these distributions were incorporated, the total taxes would increase
and all derivative functions (e.g., cumulative investments) would adjust
accordingly for all three techniques.
The effects of the investor buying and selling shares of the fund
itself are considered in the design and interpretation of the
spreadsheets. Finally, the
return on investment is a function with the cumulative investment in its
denominator. Thus, when the
cumulative investment is less than or equal to zero, it can really only
be zero implying that the investor has recouped all of his invested
money. Therefore, the return on investment is infinite. Hence, the
equations for return on investment are only valid for the cumulative
investment greater than zero.
As a result, when the cumulative investment graph shows an
instantaneous negative or zero value, the instantaneous return on
investment shown on the graphs is not valid.
Another limitation identified for this research relates to the
data used in the spreadsheets. The
simulated markets are contrived to investigate specific (not necessarily
realistic) market conditions and activity.
Furthermore, the historical market returns used to verify the
conclusions for realistic markets do not exhaustively examine every
possible permutation of the market.
For example, in the historical verifications, the first or last
weeks close was used throughout the spreadsheet for a given market, and
the data started with the most recent data available and continued
backward in time to the beginning of the dataset or thirty years
whichever was shortest. There
was no attempt to use a sliding window through the historical returns
(i.e., use closing returns on the first week of the month, the second
week of the month, the last week of the month, etc.) or varying the
start dates to test their impacts. However
results are consistent enough between all simulations, mathematical
predictions, and historical markets, that it is reasonable to assume
there would be no significant effect on the trends if sliding windows or
start dates had been used.
The final limitation identified for this research effort is in
the defined scenarios for asset allocation.
The case selected with 80% invested in the fund and 20% invested
in a money market is selected as a benchmark for comparing with dollar
value averaging. Additional
permutations could have been investigated even though they were not the
main emphasis of this study. For
example, different ratios of equity investment versus money market could
have been used (e.g., 90% stocks and 10% money market, or 60% stocks and
40% money market). The
intuitive impact, however, is when asset allocation performed better
than dollar value averaging or dollar cost averaging due to negative
markets, asset allocation would have been better yet if one were more
heavily invested in money markets. In
the reverse case, where asset allocation performed worse than the other
techniques, it would have performed better if more heavily invested in
stocks. However, as the
percent investment in equities approaches 100% (and the percentage in
money markets approaches 0%), it becomes equivalent to dollar cost
averaging by definition. Thus,
this limitation is deemed insignificant.
Another variation of asset allocation that could be more
significant, however, is to consider the case of investing in more than
one type of mutual fund as well as money market accounts (e.g., 40% in
an aggressive growth stock fund, 30% in a bond fund, 20% in an
international growth fund, and 10% in a money market fund).
This type of scenario would probably produce dramatically
different results for asset allocation.
However, to analyze this scenario, a much more complex
spreadsheet would be required which would have to contain periodic
results of multiple types of markets simultaneously.
Furthermore, to perform consistent comparisons, the same
investment scenario should be assumed for dollar cost averaging as well
as dollar value averaging. For
example, one would have to dollar value average into the same funds as
the asset allocation scenario which would probably lessen the dramatic
differences one would expect. In
fact, one of the powerful features of dollar value averaging is that it
forces asset allocation when using the technique on multiple markets.
For example, if you are investing in gold and long-term
zero-coupon bonds, and gold goes up, the plan would force you to redeem
shares in gold which could fund required investments in the bonds.
|