Bandy NAAIM Paper
Bandy NAAIM Paper
This paper evaluates methods of using fundamental data as input for trading US equities. It
examines whether and how fundamental data can help decide what to buy, when to buy, and when
to sell.
Fundamental economic data includes gross domestic product, inflation, interest rates, commodity
prices, consumer sentiment, and similar items. Fundamental company data includes inventory
There are many economic indicators, published by many agencies. Some of those that have a
significant effect on the prices of stocks, bonds, commodities, and currencies are listed below.
Descriptions of the indicators, and of the schedule for release and revision of the data they report,
come primarily from the websites of the reporting agencies. Additional material comes from a
variety of sources accessed through the references listed in the appendix. The indicators are listed
in the order of their timeliness – beginning with those reporting with the least delay following the
Published every Thursday, the data covers the week ending the previous Saturday, with
• Consumer Confidence – Estimates of how consumers feel about jobs and the economy. The
result of surveys conducted by The Conference Board, a private agency. Published monthly
on the last Tuesday of the month being reported. Revisions are rare and tend to be minor.
Institute for Supply Management (ISM), a private agency. Published monthly on the first
business day after the reporting month. Seasonal adjustments are applied annually.
• ADP National Employment – An early report of employment and unemployment. The result
of analysis of the payroll data processed by Automatic Data Processing (ADP), a private
company. Published monthly two days before the government employment report. Minor
revisions are published the following month. Annual revisions are made in March.
• Employment Situation – The government report of jobs created or lost and the
unemployment rate. The result of surveys conducted by the Bureau of Labor Standards, an
agency of the US Department of Labor. Published monthly on the first Friday of the month
following the reporting month. The first data published is labeled “advance.” One month
later, along with that month’s regular report, a revision labeled “preliminary” is released.
Two months later, a revision labeled “final” is released. The series has its benchmark
• Retail Sales – A report of spending on goods (not including services and not adjusted for
inflation). The result of surveys of retailers conducted by the US Census Bureau, an agency
of the US Department of Commerce. Published monthly about two weeks after the reporting
month. Revisions are published for the next two months, and the series is re-benchmarked
annually in March.
activity (not taking price into account). The results of data collected from companies by the
Federal Reserve. Published monthly around the 15th of the month following the reporting
month. Revisions are published for the next three months. The series is readjusted every fall.
• Producer Price Index (PPI) – Measures the changes in prices paid by businesses. The result
of data collected from goods producers by the Bureau of Labor Statistics, an agency of the
US Department of Labor. Published two to three weeks after the reporting month. Revised
• Housing Starts and Building Permits – Measures activity related to construction of new
housing. The result of telephone and mail surveys conducted by the Census Bureau, an
agency of the US Department of Commerce. Published monthly two or three weeks after
the reporting month. Revisions are published for the next two months, and seasonal
• Consumer Price Index (CPI) – The commonly used measure of inflation at the consumer
level. The result of data collected by the Bureau of Labor Standards related to prices for
a defined basket of goods. Published monthly, usually about two days after the PPI. No
monthly revisions, but annual readjustments are computed in January. Since this series
is used as the basis for annual changes in many payments, such as Social Security, some
• Personal Income and Spending – A measure of the amount of money households receive
and the amount they spend. The result of data collected from a variety of sources by the
monthly about four weeks after the end of the reporting month. Revisions are computed
and published for the next two or more months, with annual revisions around July, and re-
• Durable Goods Orders – A measure of orders received at factories. The result of data
about three or four weeks after the end of the reporting month. Revisions, which can be
substantial, are published one week after initial release, then for the next two months.
• International Trade in Goods and Services – A measure of exports and imports of goods
and services. The result of data collected on exports by the Census Bureau and the Bureau
of Economic Analysis. Published monthly the second week of the second month following
the reporting month. Revisions are published monthly for several succeeding months. The
• Productivity and Costs – A measure of the productivity of workers who produce goods
and services. The result of data collected by the Bureau of Labor Statistics from a variety of
sources, including monthly payroll reports and gross domestic product. Reported quarterly
about five weeks following the reporting quarter. Revisions are made monthly for three
months, and as necessary as the underlying economic series are themselves adjusted.
• Gross Domestic Product (GDP) – The single number is the sum of the value of all goods
and services produced in the United States. The report is broken down into detail by sector
of the economy and by geographic area. It is the result of data collected from a variety
of sources by the Bureau of Economic Analysis. Reported quarterly the final week of the
month following the reporting quarter (final week of April for the first quarter). Revisions
are reported each of the next two months, with annual revision in July and benchmark
• Current Account Balance – A report on the country’s trade with the remainder of the world.
Results of data collected by the Bureau of Economic Analysis. Reported quarterly about
two and one-half months after the reporting quarter. There is one monthly revision three
There are many more series than the sixteen described above. Some are followed closely and others
widely ignored. One series that has intentionally been omitted from the list given above is the
Index of Leading Economic Indicators (LEI). That index has been completely redesigned several
times in the past two decades. It currently has ten components, chosen from among hundreds of
available candidates to best fit the in-sample period used during its redesign. There is speculation
that the most recent redesign was made to avoid having the LEI forecast an economic recession.
The following figure illustrates the values reported for the US GDP. The highlighted reporting
period is the first quarter of 2006. Note the dates and reported values for the initial report and the
revisions.
The following chart shows about three years of daily price data for a stock that an active manager
might trade. The box identifies the first quarter of 2006 and the annotations identify dates the GDP
data covering that quarter are released, revised, and re-based. The data is taken from the figure
above. The chart is intended to illustrate the lag between the period being reported and the date
In order to be valuable, any indicator, including any economic indicator, must be:
• Timely
• Accurate
• Predictive
As is evident from the descriptions of the economic data and from the chart above, there are
several issues that complicate use of fundamental economic data for trading.
monthly, or weekly. Trading decisions are made monthly, weekly, daily, or intra-day. If the
stock price is reported and acted upon more frequently than the economic indicator, there
will be many time periods (data bars) where there is no data for the economic indicator. In
order to have a value to use in calculations, the latest value that does exist will be copied
forward until a new value is received. The only time that value can change is on those days
2. Timeliness related to revision. Economic indicators, and other fundamental data, are reported,
then revised at later dates. When the historical data is retrieved from the data provider, it
will usually be a series that consists of only the final revision data. In order to maintain
consistency, the data value associated with a given time period cannot be used in the
trading system until after the date its final revision is published.
An alternative is to use a series that consists solely of data initially released, with no
revisions or adjustments applied. The trading system would then be based on initial release
data rather than final release data. Some data providers maintain a record of the initial
report and all subsequent revisions, enabling the consumer of the data to create the data
series that is most useful to them. That approach may work for economic series that seldom
have significant revisions. But it will not work for series that are regularly heavily revised.
3. Accuracy related to revision. Government statistical series are regularly given annual
adjustments, are re-based, and re-benchmarked. Re-basing sets a new date for the base of
the index (the date it has a value of, say, 100) and adjusts all data in the series accordingly.
that depend on others. Any of these operations result in a revised historical data series. A
trading system based on an earlier revision may not be profitable when the newly revised
data is used.
4. Accuracy related to bias. There is a bias to any reported data. That bias is both unknown and
unknowable to users of the data. Whether the report is unintentionally biased due to an
innocent data preparation error or omission, or intentionally misleading, outsiders probably
cannot detect the bias, its amount, or its reason. They have little alternative but to accept and
use the data as reported. Bias introduces a systematic error into the reported statistic.
number representing that statistic, the reported number is only one estimate of that value. If
the same data and same procedures are used to recalculate the same statistic several times,
the results will differ slightly from each other due to random variations in measurement
and interpretation. The more subjective the question, the more random error in the answer.
Preparers of these reports must be careful to avoid confusing precision with accuracy.
10
6. Predictive. Whether the fundamental data is predictive depends on the strength of the
relationship, the efficiency with which the market assimilates the information, and the
insight and skill of the developer of the trading system. Remember to follow good modeling
and validation practices. Keep enough data reserved for out-of-sample testing. In-sample
results are always good and have no value in predicting the profitability of a system when
For all of these reasons, it is difficult to incorporate the fundamental data series with the daily
or weekly price series representing the trading prices. In order to use fundamental data as a
The sixteen economic indicators described above fall into just a few categories:
• Inflation
• Debt
Fundamental data for individual companies includes sales, earnings, dividends, book value,
inventory turnovers, and similar items. Similar difficulties arise with company data as with
When dealing with company data, changes to the organization of the company result in changes to
the data series. Explanations and footnotes may be added to annual reports, but historical data is
11
seldom adjusted, or even adjustable. For example, in January 2009, pharmaceutical Pfizer acquired
Wyeth and gained their over-the-counter drug lines. Pfizer sold its consumer products division
to Johnson and Johnson in 2006, and sold its medical device division in the 1990s. Since Wyeth
no long exists as an independent company, its data is no longer maintained, but belongs in the
category of inactive companies that have been discontinued, delisted, or absorbed. The Pfizer data
might be adjusted by a pharmaceutical industry analyst at the request of a major client, but that
data series will probably never be available for wide distribution at low cost.
The problem can become even more complex when a company contains very different divisions,
each of which contributes revenue. For example, major manufacturers who operate their own
captive finance divisions may have seen the majority of their earnings move from manufacturing
to the financing operations. Since the company had been primarily a manufacturing company, they
are probably grouped with and compared to other manufacturers. If such a company divests its
financial operation, the data series will show a dramatic change, which is seldom adjusted.
The question of accuracy and trust is illustrated by the list of companies suspected or convicted of
misleading investors by misstating the company’s fundamental data. It includes Adelphia, Enron,
• Equities
• Commodities
• Currencies
A search of the databases results in a list of fourteen indexes that might be related to the economic
indicators and serve as surrogates for them and for the asset classes. In addition, most of them
trade as futures contracts. The indexes are all sponsored by reputable companies, all have history
dating from the early 1990s or earlier, all are widely followed, and all fit neatly into one of the four
12
The correlation, measured over a 252 day period, of each index with every other index is displayed
in the following table. Each value is the average of twelve years. The cells colored green have
an R-squared value of 0.50 or more. Each column is labeled as a commodity, dollar, interest rate,
or equity. There is little difference among the members of each asset class, relative to the other
indexes. For example, the correlations of all the equity indexes are very similar. One index of each
The four indexes are plotted together in a single chart, shown below. If they were to show a
consistent relationship of peaks and valleys, with one leading, another following, the third
following later, and the fourth last, then those four indexes would make an ideal set to be the basis
of a rotational trading system. Or, perhaps be useful in creating filters for broad market timing.
Even though there is no consistent sequence of peaks and valleys, there is almost always at least
one class scoring high. That is promising, but is outside the scope of this paper.
Another kind of series that can be used as a surrogate for an economic indicator is the continuous
Since the trading hours for both the indexes described above and futures contracts are
approximately the same as the trading hours for stocks, the problems with timeliness and accuracy
are removed.
14
Futures trading began with agricultural products — forward contracts in the 1850s, which evolved
into futures contracts traded on exchanges by the 1870s. Other physical commodities were added
over time. The big changes came in 1972 when currency futures were added, interest rates in 1975,
and equities in 1982. By 1985, trading in agricultural contracts accounted for less than 25 percent
of the total. Consequently, there is complete and accurate data for the prices of equities, interest
rates, currencies, and commodities from 1982 on. Individual futures contract data and continuous
contract data are available at low cost from data vendors such as Yahoo, Quotes Plus, Norgate
Premium Data, and Pinnacle. Examples of futures contracts, some of which might be useful:
• Equities — S&P 500, NASDAQ 100, Dow Jones Industrials, Russell 2000
• Materials and Energy — Crude Oil, Copper, Lumber, Gold, CRB Index
Commodities have their own fundamentals, which are closely related to fundamental economic
indicators. Data on supply, production, and consumption is available, and traders of futures
contracts may be interested in that. But when the commodity data is being used as a surrogate for
Investment advisors and publications describe the sequence that the business cycle follows.
Conventional wisdom views the business cycle as repeatedly progressing through phases.
Depending on their business operations, stocks tend to perform better in some phases than others.
There are two time frames for this analysis. The first time frame measures cycles relative to
recession lows and the following recovery. In the 94 years since World War I, there have been 18
recessions, as defined by the National Bureau of Economic Research. The average time between
recession lows is 5.2 years, with a range of 1 year to 11 years. There are too few of these events
to draw general conclusions, and the holding periods are longer than fit the definition of active
15
management. Until the current recession, it appeared that recessions were becoming less frequent
and shorter, and expansions longer. But there is too little data to be conclusive. There may be
some indicators related to recessions that could be used as filters for more frequent timing and for
stock selection, but those are not discussed in this paper. The following figure shows the S&P 500,
plotted on a logarithmic scale, with the periods the US was officially in a recession highlighted.
Martin Pring and Sam Stovall have both written books outlining the sequence of the business cycle
as it relates to recession and recovery. Both books are listed in the References. Pring’s book is in
print, and his website has an article describing his ideas. Stovall’s book is out of print, but there are
several internet sites that outline his ideas — see Investopedia, for example.
The following figure plots the S&P 500 (Yahoo Financial symbol ^GSPC) from the beginning of
each of several recent recessions. Note that, even as the broad market is falling, there are significant
The second time frame assumes several cycles per year, and will be discussed further here.
A simple experiment will test whether two data series act in predictable phases. Define a period of
time to test, say 10 years. Working with the first data series — the one that will be traded — apply
a zigzag indicator and adjust the percentage parameter to correspond to the desired magnitude
of the trades — say, 5 percent. This means hold a long position for any move that has at least a
five percent gain but does not have a five percent drawdown. The zigzag will show the peaks and
valleys. Buy every valley and sell every peak. Counting valleys, 20 valleys would correspond to
the series having two complete cycles each year, on average, which corresponds to two holding
periods, each an average of three months long. Working with the second data series, apply a zigzag
indicator and adjust the parameter so that there are the same number of peaks and valleys as the
first series. If the two series follow each other in phase, the sequence of 40 valleys will alternate
— each valley from the first series is both preceded and followed by a valley from the second
series. Each sequence of successive valleys by one of the series is counted as a “run.” If the two
series acted perfectly in phase, the combined results of the two series would result in 40 runs, each
of length 1. Statistical tests, such as the runs test, can be used to measure the amount by which the
two series differ from being perfectly in phase. But the goal is not to perform statistical tests, rather
to develop profitable trading systems. So, if there appears to be regular alternation between valleys,
write and test a trading system that identifies valleys in the second data series. Recognizing that a
17
valley in the second series has passed alerts the manager to be ready to take a long position in the
This technique was applied to the S&P 500 and the 10 year Treasury Note Yield (symbol ^TNX) for
the period 1/1/1988 through 1/1/2008. The following figure illustrates the sequence of the peaks and
valleys.
The percentage change for the S&P 500 was set to 5%. That resulted in 58 valleys over the 20 years.
The percentage change necessary for the 10 year Treasury Note Yield to give the same number
of valleys was 5.56%, determined by trial and error. Note that the two series do not alternate, but
show runs of one series followed by runs of the other. Although this technique did not work well
using the 10 Year Treasury Note, there may be other series that would work better.
Sectors as surrogates
Sector indexes or sector exchange traded funds are created by selecting individual issues that
have similar characteristics. The price of the composite is a data series that reacts to economic
fundamentals in the same manner as the individual component issues react. But there is less
day-to-day variability. The price movements are smoother because a portion of the random noise
contained in the data of the individual issues cancels each other out. A search of the database
reveals a long list of exchange traded funds and indexes that might be useful.
18
Standard and Poor’s has created a set of nine select sector exchange traded funds. All 500 of the
stocks that are components of the S&P 500 index are used. Each stock is also a component of
the sector fund that its business fits best. The funds have between 29 and 83 components. The
• Energy — XLE
• Financials — XLF
• Industrials — XLI
• Technology — XLK
• Utilities — XLU
We know it is possible to use those sector ETFs to develop profitable trading models. The question
addressed in this paper is whether they can be used as surrogates for economic indicators.
Five stocks were selected from each ETF — the four stocks with the highest weighting in the ETF,
and one stock that has about 2% weighting. Keep in mind that all the stocks in the S&P 500 are
Daily percentage price changes were computed for each stock, each sector, and the S&P 500 for
about 10 years. For each of the 45 stocks, a multiple linear regression was used to measure the
relationship between the stock, its sector, and the S&P 500. This is a descriptive model, not a
predictive model, so we will not be concerned with the problems that arise when regressing non-
where Stock, Sector and SP500 are the daily percentage prices changes.
19
The regression equation for Exxon Mobile, XOM, related to the Energy Select Sector, XLE, and the
The following chart shows the daily price change observed (the blue line that has the greater range)
and the daily price change explained by the regression (the red line) for a short, but typical, period.
The scatter plot of the entire ten years is shown in the next chart. The observed change is plotted
on the horizontal axis, the change explained by the regression on the vertical axis.
20
The next chart illustrates the magnitude of the coefficients associated with the sector and the broad
market. The whiskers show the 95% confidence levels. The regression is highly significant.
The following nine charts illustrate the relationship between the stock’s price change, the price
change of the sector it belongs to, and the price change of the broad market, for the 45 stocks and
9 select sectors chosen. The percentage of the stock’s price change that is explained by the stock,
the sector, and the broad market are shown by a group of three bars for each stock. The blue bar
to the left shows the relative importance of the stock itself. The red bar in the middle shows the
relative importance of the sector. The green bar to the right shows the relative importance of the
broad market. The sum of the three bars is 100%. Note how some sectors strongly affect the stocks,
as shown by tall red bars. Other sectors contribute less, as shown by the tall blue bars. The broad
For all 45 stocks, the average percentage contributions to the stock’s price change are:
Sector 36.3%
Examination of the relationships of these stocks to their sectors shows that the nine S&P sector
ETFs are too broad and too diverse to use as surrogates for economic indicators.
To narrow the scope of the sector, we can: create custom sectors; use sectors defined by industry
To create a custom index, a population of stocks is examined. All of those in the Russell 3000, for
example. The goal is to identify a number of issues, say 10 to 20, that behave the same, or close to
the same, in all market conditions. An index of these is created and used to generate the trading
signals. The trades taken can be a basket consisting of the stocks in the index, a few highly liquid
stocks that are components of the index, or some related issue. The mathematics and mechanics
of creating a custom index are straight-forward. But creating the index raises a serious issue
of membership bias. The composition of the index when it is created today is almost certainly
different than it would be if it had been created in the past. One way to reduce the effect of
membership bias to is select the stocks that form the index as of a date in the past. Since the model
validation process will involve in-sample development and out-of-sample testing, choose the date
of the start of the out-of-sample period as the date for selecting the components of the index.
Use of industry codes removes the effort of searching for issues that behave similarly, and it
removes a portion of the membership bias. Instead, it relies on accurate assignment of industry
The North American Industry Classification System (NAICS) is the coding system used by Federal
23
statistical agencies in classifying business establishments for the purpose of collecting, analyzing,
and publishing statistical data related to the U.S. business economy. NAICS is an agency of US
Census Bureau. The NAICS codes are assigned to businesses according to their business activities.
It is difficult to determine the NAICS code for a given company — retrieval by stock symbol is not
available. NAICS codes are not usable for sector index creation.
Dow Jones and The Financial Times and the London Stock Exchange (FTSE). It is used to segregate
markets into sectors within the macroeconomy. The ICB uses a system of 10 industries, partitioned
into 18 supersectors, which are further divided into 39 sectors, which then contain 104 subsectors.
The Global Industry Classification Systems (GICS) is a company classification system developed
by Standard and Poor’s and Morgan Stanley Capital International (MSCI) Barra. GICS defines peer
groups tightly and avoids grouping unlike companies together. The GICS structure is four levels
GICS competes directly with ICB. The ten highest level categories are almost the same. ICB uses
a “product-oriented” approach, where the category a stock is assigned is determined by what the
company produces. Companies that produce goods are generally in different categories than those
that produce services, although there are some companies that produce both and are difficult
to classify. GICS uses a “market-oriented” approach, where the category a stock is assigned is
determined by how its products are sold. Consumer staples companies, for example, include
companies that provide both consumer products and services that are considered necessities.
Both systems expose the definitions of their top three levels freely. Both require a subscription
to learn the definitions of the fourth and most narrowly defined categories. Both GICS and ICB
require a subscription to allow retrieval of industry code through lookup of the stock ticker.
24
The stocks in the S&P indexes have been assigned one of the two-digit GICS sector codes. Those
codes are:
10 Energy
15 Materials
20 Industrials
25 Consumer Discretionary
30 Consumer Staples
35 Health Care
40 Financials
45 Information Technology
50 Telecommunications Services
55 Utilities
In order to be useful as surrogates for economic indicators, much finer resolution is required.
Use of an index created by an agency has the advantage that membership and survivorship issues
were dealt with at the time they arose. If the components of the index are similar enough, that
might work. As of March 2, 2009, all 79 of the components of the S&P Sector Select Consumer
Discretionary ETF, symbol XLY, were assigned GICS code 25. But the XLY ETF included Amazon,
Auto Nation, International Game Technology, New York Times, and Pulte Homes — clearly too
dissimilar for this purpose. Another categorization system gives the following classification for
Some of the indexes and exchange traded funds are more precisely defined. For example, the
HOLDRs Semiconductor ETF, symbol SMH, consists of about 18 stocks. The same regression model
that was used with the nine S&P select sector ETFs was applied to daily percentage price changes
for SMH and its top five holdings. The next chart shows the relative importance of the stock, the
Note how strong the influence of the sector is. For all 5 stocks, the average percentage contributions
Sector 70.2%
The following chart shows the relative prices of SMH (the heavier blue line in the middle) and the
top five components for a few months in 2004. The SMH is very representative of the components.
26
Compare with the relative prices of XLY (the heavier blue line in the middle) and its top five
More precisely defined sectors are easier to model, because their components tend to move in
reaction to the same fundamental information. And after an accurate model has been created,
trading the sector index itself or the liquid components of it are more profitable because there is
The goal of this paper was to evaluate methods of using fundamental economic indicators and
fundamental company data as components of trading systems used by active managers. The
• Economic indicators, such as those published by governmental and other agencies, are not
appropriate for direct use in active trading systems for reasons involving their timeliness
and accuracy.
• Company fundamental data are not appropriate for use in active trading systems for
currencies — to form a trading system shows profit potential. There appears to be more
potential for use in a rotational trading system than for use as surrogates for fundamental
data.
27
a trading system also shows profit potential. Complications arise because of problems in
defining the sector indexes. Again, there is potential for use in a rotational trading system,
• If sectors are carefully and narrowly defined, most of the price movement of individual
• Surrogate data, primarily in the form of sector indexes defined and maintained by financial
• Intermarket analysis, such as examination of interest rates, can be used for broad market
• Exchange traded funds can be used both to generate the trading signals and to take the
position.
• Indexes and futures contracts can be used in trading systems of their own or as surrogates
of economic indicators.
28
References
AAII — American Association of Individual Investors. Stock Screener Pro provides historical
fundamental company data and a screening program. Subscription. http://www.aaii.com/store/
sipro/index.cfm
American Economics Association. Extensive information of interest to economists. In particular,
the Resources For Economists page, http://rfe.org/ with links to journals, articles, and data. Much of
the content is free. http://www.aeaweb.org/index.php
Automatic Data Processing. Publishes the monthly ADP National Employment Report. http://
adpemploymentreport.com/
Bandy, Howard. Introduction to AmiBroker. Blue Owl Press. 2008.
Bandy, Howard. Quantitative Trading Systems. Blue Owl Press. 2007.
Baumohl, Bernard, The Secrets of Economic Indicators, Second Edition, Wharton School Publishing,
2008.
BEA — Bureau of Economic Analysis. An agency of the US Department of Commerce. Publishes
the monthly personal income report. Also provides other historical economic fundamental data.
Free. http://www.bea.gov/
Capital IQ, a division of Standard & Poor’s. Detailed data on public and private companies.
Subscription. https://www.capitaliq.com/Main3/ourproducts_platform_fundamentals.asp
Compustat, a division of Standard and Poor’s. Current and historical fundamental data on current
and inactive companies. Subscription. https://www.compustatresources.com/support/index.html
Compustat. The top three levels of GICS. http://mi.compustat.com/docs-mi/help/gics_info.htm#utils
Conference Board. Publishes the monthly Consumer Confidence Index and Index of Leading
Indicators. Mostly subscription. http://www.conference-board.org/
Department of the Census, an Agency of the US Department of Commerce. Publishes the monthly
retail sales report. http://www.census.gov/
Economagic. Historical economic fundamental data. Free. http://www.economagic.com/
Edgar. Current and historical fundamental company data. Security and Exchange Commission
filings. Subscription. http://www.edgar-online.com/
Federal Reserve. http://www.federalreserve.gov/
Federal Reserve Bank of New York. Reports related to monetary policy. Free. http://www.
newyorkfed.org/
ICB. Click the “structure” link to open a pdf with the four level ICB structure defined and codes
listed. http://www.icbenchmark.com/icb_structure.html
GICS definitions in Excel file format. http://www.mscibarra.com/products/gics/structure.jsp
Global Financial Data. Historical end-of-day OHLCV price data for active and delisted US stocks,
company fundamental historical data, economic fundamental historical data. Subscription. https://
www.globalfinancialdata.com/
29
Hemscott, a division of Morningstar. Current and historical fundamental data on current and
delisted companies. Subscription. http://www.hemscottdata.com/database.html
ICB — Industry Classification Benchmark. Information about the ICB. Access to list of classification
by company is by subscription. http://www.icbenchmark.com/
ICB — Industry Classification Benchmark. Wikipedia entry giving the definitions for industry,
supersector, and sector. http://en.wikipedia.org/wiki/Industry_Classification_Benchmark
Institute for Supply Management. Publishes the monthly manufacturing survey. http://www.ism.
ws/
Investopedia. Article on sector rotation giving Sam Stovall’s industry rotation sequence. http://
www.investopedia.com/articles/trading/05/020305.asp
Investopedia. Definitions of GICS and ICB classification methods. http://www.investopedia.com/
articles/stocks/08/global-industry-classification-industrial-classification-benchmark.asp
MSCI Barra — GICS definitions in Excel file format. http://www.mscibarra.com/products/gics/
structure.jsp
NBER — National Bureau of Economic Research. Definition of recession and list of recessions and
depressions in the US. http://www.nber.org/cycles/
NAICS — North American Industry Classification System. An agency of the US Census Bureau.
Definitions of the NAICS industry codes. http://www.census.gov/eos/www/naics/index.html
Norgate Premium Data. End-of-day historical and daily update data for US and Australian equities,
futures, and foreign exchange. Subscription. http://www.premiumdata.net/
OECD – Organization for Economic Coordination and Development. Statistics, economic, and
social data for 30 global countries. Free. http://www.oecd.org/
Pinnacle Data. End-of-day historical and daily update data for indexes and commodities.
Subscription. http://www.pinnacledata.com/products.asp
Price Data. Historical intra-day OHLCV price data for US stocks and futures. Subscription. http://
www.grainmarketresearch.com/gm.cfm
Pring, Martin. The Investor’s Guide to Active Asset Allocation. McGraw-Hill. 2006.
Pring, Martin. Understanding the Business Cycle. http://www.pring.com/articles/article8.htm
Quotes Plus. End-of-day historical and daily update data for equities, mutual funds, exchange
traded funds, commodities. Subscription. http://quotes-plus.com/joomla/index.php
Rimes. Coordinator and redistributors of fundamental data on companies. Subscription. http://
www.rimes.com/Default.aspx?xml=mainmenu.xml&node=1.1
Standard and Poor’s. GICS pages. Extensive information about categorization of stocks. http://
www2.standardandpoors.com/portal/site/sp/en/us/page.topic/indices_gics/2,3,1,7,0,0,0,0,0,0,0,0,0,0,0,0
.html
Standard and Poor’s. Select Sector pages. Descriptions of the sectors and their component lists.
http://www.sectorspdr.com
30