Saturday, March 2, 2013

Learning from the Past ~ mfim ~ ModularFinance ~ in ~ motion

BROKEN PLANS

~Hat tip - Andrew Gluck, Financial Advisor Magazine
September 2, 2010



The financial crisis has laid bare some of the folly of long-term financial planning. The notion that you can use past data on assets to create forward-looking plans that stretch out ten, 20 or 30 years is naive.

When advisors make predictions about the return on stocks, for instance, they are often looking back at data on returns, standard deviations and correlation coefficients going back to 1926. The conventional wisdom holds that the farther back in years the data goes, the better. But what the 2008 crisis taught us is that this method of considering possible bad outcomes is deficient.

Monte Carlo simulation in financial planning programs, for instance, uses historical returns of asset classes and their standard deviations to project possible future outcomes. In most planning apps, a lognormal or fat-tail distribution is used to better reflect the unlikely chance of a really big gain or loss in any given year. But it turns out that Monte Carlo simulations randomize returns and risk data too much to model the real world accurately. The math doesn't work. It fails to see that some economic conditions linger-that portfolios can perform very poorly or very well several years in a row.

Actual economic scenarios, in other words, instead of randomness, should thus become the basis of investment projections. Real-world historical data should form the building blocks of simulations.

This is not a new concept. And yet scenario planning has not caught on among advisors. The financial crisis, though, may have shown we must have it in our financial planning applications.

No comments: