MasterDATA Composite Breadth Data Methodology
Historical Composite Breadth Data
MasterDATA compiles its historical
composite/breadth datafiles using the current
component list only. There are at least two
points of view regarding historical data on
composites (indexes and ETFs) as follows:
Historical There is logic to the
argument that historical composite datafiles should
reflect every component list change over the history
of an index or ETF. In other words, Unocal
(previously symbol UCL), a component in many major
indexes and ETFs, should be included in the composite/breadth
data of any index or ETF for which it was a component.
The fact that Unocal was bought out by Chevron in
August, 2005, makes no difference. The price
movements of Unocal "should" (according to this
point of view) be included in the S&P 500 Index,
its related ETF (SPY), the Russell 1000 Index, its
related ETF (IWB), the Energy SPDR and many, many
more indexes and ETFs permanently and for all time
even though it no longer exists.
Survivor: There is nothing sacrosanct
about any index or ETF. Indexes and ETFs are
simply a list of stocks, often with some commonality,
perhaps, but a list nevertheless. The Russell
2000 Index is a list of (approximately) 2000 mostly
unrelated stocks and the fact that the list was
different last year, from this point of view, is
irrelevant. In fact, the Russell 2000 Index
seldom has exactly 2000 components except on its
reconstitution date once each year. From this
point of view, focus should be on the historical
composite data only for the current list of components.
After all, you can't trade the historical list.
Another way to look at it is that if you, personally,
constructed a "basket" of stocks to analyze for
your own purposes and, over time, made changes to
the list, would you analyze the list including all
the changes you made or would you analyze your current
"basket" of stocks? Most likely, the latter.
What many investors call a "basket" is a composite
just like any other composite (indexes and ETFs
included) and should be, from this point of view,
treated the same way.
Only one of these two points of view is even feasible
in the real world. While it might be interesting
to apply the first point of view, historical changes
to the component lists of most indexes and ETFs is simply
not available. Although I am certain, records
exist somewhere, the index and ETF sponsors are either
unwilling or unable to produce such historical changes.
At best, sponsors provide a readily accessible list
of current components (some not even that). The
very best in our experience is Standard and Poors which
not only provides a current list but a list of changes
to the component list over the last few years.
But even the best sponsor does not provide component
list changes for the life of its indexes.
We are therefore left with the Survivor approach
as our only realistic alternative. MasterDATA's
historical composite data is therefore based only on
the current list of component stocks making up any index
or ETF. This means that when one component is
deleted and a new one added to replace it, we recalculate
the entire database of historical composite breadth
data using the current list only (just as if Unocal
I will add, however, that I leaned in the Survivor
direction in the first place. In scientific studies
it is common practice to discard the best and worst
results of any study and calculating on those results
remaining. Eliminating merger and buy out activity
as well as bankruptcies and stocks dropping to Pink
Sheets, being moved from one index or ETF to a different
index or ETF and only using the current list of components
approximates a somewhat similar practice. Whether
or not this is a valid argument or simply a rationalization,
is really not important. The bottom line is this.
With the information available, its either this approach
or nothing. I strongly believe it makes more sense,
from an investors point of view, to only work with a
list of stocks currently trading. Frankly, it
would not be particularly difficult to generate the
composite breadth datafiles the other way (if the information
was available), but I am convinced the results could
not be any more productive.
Equal Weight Composite Breadth Data
The S&P 500 Index is capital weighted.
Component weighting utilizing this method is based upon
market capitalization, so one component basically counts
more or less (carries more or less weight) than the
next when figuring the current value of the total composite.
A one point price move in a component with a larger
market capitalization affects the parent index or ETF
total value more than a component with a smaller market
The Dow Jones Industrial Average is price weighted.
In this method, component weighting is based upon the
price of the component issue. In other words,
if you started a new 30 component index today, this
method would add the current price of the 30 components
and divide by 30. Similar to capital weighting,
one component thus carries more or less weight than
the next, but in this method the higher priced components
carry more weight than lower priced components (as simple
as this method sounds, it gets very complicated, very
quickly when the divisor is changed - in other words,
in our example, you would no longer divide by 30, but
instead, another number calculated to compensate for
various component changes such as spin-offs, etc. -
the Dow's divisor is currently less than one).
composite breadth data is inherently equal weighted.
Each component of an index or ETF carries exactly the
same weight as the next. If 316 components of
the S&P 500 Index increase in price for the day, the
composite breadth statistic, advancing issues (or simply
"advances"), is 316. Because most indexes and
ETFs are not similarly equal weighted, a possible
conflict presents itself. For example, using a
capital weighted index like the S&P 500 Index with equal
weighted composite breadth data could potentially be
akin to comparing "apples and oranges".
To address this issue, in addition to providing historical
market values, MasterDATA recalculates all followed
indexes and ETFs as equal weight. Both
sets of historical data are included in downloads from
this site, both actual values and recalculated values.
The process of recalculating each index and ETF as
equal weight developed into a much larger project than
one might anticipate from the idea's initial conception.
For one thing, each of as many as 3400 component issues
had to be filtered and manually corrected for historic
price errors (our data vendor is one of the biggest
and "best", but an error here and there can quickly
result in a major impact on total values). Additionally,
numerous methods were implemented before arriving at
one that displayed absolutely no "drift", but instead
provides meaningful values comparable to the indexes'
and ETFs' actual market valuation.
The result of this work is very intriguing.
While the overall chart patterns remain basically the
same, price moves are generally smoother. A large
price move in a heavily weighted component does not
overly impact the index or ETF value unless other components
experience similar movement. From a technical
analysis point of view, equal weighting might be considered
the ideal weighting methodology for composites.
In any event, the data is provided at no additional
charge. It is your decision and your decision
alone to use it or not.
For a chart comparison between index and ETF market
values and MasterDATA's recalculated equal weight values,
click here. For more about the primary weighting
click here for S&P's brief presentation.
MasterDATA will only generate historical composite
datafiles on indexes and ETFs with a readily accessible
and accurate list of current component issues.
Some ETFs in particular, are hesitant or unable to provide
an up to date list. Indexes and ETFs with components
not trading on a US exchange are also not currently
included in MasterDATA's followed list (this will change
Addition / Deletion of Components
We make every effort to make changes to our component
list on the day of the change or, at worst, shortly
thereafter. Buy outs and mergers often cause a
component to be dropped from a component list.
Many indexes and ETFs replace such components systematically
(for instance, the S&P 500 virtually always has 500
components). Others refill their component lists
at given time intervals (i.e. the Russell indexes).
Still others, such as those with a unit investment trust
structure, never replace component issues (i.e. HOLDRS).
Deletions due to bankruptcy, delisting and dropping
to the Pink Sheets, are handled similarly.
MasterDATA's practice is to immediately delete any
security that is no longer trading or that moves to
the Pink Sheets. If the index or ETF lists a replacement
component, that component issue is added simultaneous
to the deletion or as soon thereafter as the sponsor
makes the information available.
The entire MasterDATA database of historical composite
data is rebuilt from scratch nightly and reflects the
currently published component list of each index or
ETF followed. During the trading session, updates
to the full historical database are generated hourly
(prices as of the "top" of each hour) and published
to the web site by about 20 minutes after the hour.
These "snapshots" are less for the purpose of intraday
trading and more specifically to facilitate entry or
exit of positions on the close of the trading session.
After the close of trading, numerous updates are generated
to reflect adjustments by the exchanges and data vendors
to price and volume statistics for the day. Such
adjustments often occur days and even weeks later.
All adjustments are reflected in MasterDATA's full nightly
rebuild of all historical composite datafiles.
Accuracy Best Efforts
MasterDATA applies it utmost best efforts to maintain
its historical composite datafiles to the highest standards
of accuracy and timeliness. Occasionally, uncontrollable
events temporarily preclude these efforts including
inaccurate information from our data vendors, late information
provided by the sponsor of a particular index or ETF,
computer malfunction, Internet disruptions, human error,
or some other unforeseen event. In any case, all
errors, for any reason will be corrected as soon as
possible. Subscribers will be notified on the
web site when an issue arises.