Archive

A bit more on how the HiddenLevers core model works

I covered HiddenLevers.com in this week’s Tech Update column and wanted to share a…

I covered HiddenLevers.com in this week’s Tech Update column and wanted to share a bit more on their methodology than we had the space to go into there. So what follows is information provided by HiddenLevers. Previously the site’s founders were only making it available to those enrolled in their beta testing but have since added it to the “How it Works” pages at their site.

To give those that came to this post directly from our InvestmentNews Daily e-mail newsletter I will provide a little color and additional flavor not found in the column (you will still want to read that for a more well-rounded idea of what HiddenLevers is and does; link above or at the top of “Related stories” below).
In my first meeting with co-founder Raj Udeshi he tried to paint a picture for me in describing the site (since I had not seen it yet):
“We’re a technology company, one that helps provide RIAs with a conduit to bring wider economic views to bear on the portfolios of their clients,” he said.
“Take the front page of the New York Times and link it through to my portfolio or the front page and back page of the Wall Street Journal and do the same,” Mr. Udeshi added.
Presently available to users of the site (which requires registration) are six different tools including sets of pre-determined hypothetical scenarios created by HiddenLevers (for example “Japan Nuclear Meltdown”) that a user can run against stocks, mutual funds, ETFs or a portfolio they have entered, as well as scenario modeling and economic modeling tools. There is also a macro trend screener, a list of macro profiles grouped by industry, and a list of macro strategies that advisers can study and delve more deeply into.
Again, the following methodology explanation was provided by HiddenLevers:

HiddenLevers Model Overview

Level 1: HiddenLevers Regression Model

Asset return = A * (lever return) + B * (S&P return) + C

Since up to 70% of any given stock’s return is based on overall market returns, a simple correlation of the indicator and stock/ETF can be misleading. Instead, the regression coefficient representing a stock’s sensitivity to a particular economic indicator is calculated using both the change in the indicator and the change in the market, as shown in the example equation above. HiddenLevers thus controls for the effect of the market when measuring the correlation between a stock and a lever, leading to a more accurate measure of the relationship.
Once the regression analysis is complete, HiddenLevers captures as lever impact coefficients (correlations) only those regression results which are statistically significant within a standard 95% confidence interval. These results become the actual investment-lever relationships used to drive the HiddenLevers stock screener and scenario analysis.
HiddenLevers currently uses a decade of data to perform most statistical analysis, except when the stock/ETF has not been in existence for a decade. A decade of data is preferable since it provides enough data to observe statistically significant relationships, without extending the time-frame so far that it falls into a potentially different macro-economic era. It’s important to note that relationships between particular stocks and levers may change over time; HiddenLevers accounts for this by running its statistical engine on a nightly basis in order to ensure that recent changes are taken into account. In an upcoming release, HiddenLevers plans to calculate these relationships for multiple time periods in order to better model how correlations change over time.

For more information visit the HiddenLevers How It Works page