Analysing the cycle

As we are seeking precise quantitative metrics for performance, risk, and forecasting we almost inevitably turn to one of the key tools used by investment professionals: the factor model. This is a brief and superficial skim of the topic, avoiding math to sketch the key concepts.

  • About factors:
    • The factors described here come from statistical techniques applied to the entire set of over 100 indexes of monthly returns

    • Factors come automatically ranked in order of decreasing importance, and they are statistically uncorrelated (independent)

    • It turns out that just 3 factors ‘explain’ about 80-90% of past performance, meaning that systematic (not random) risk and return is dominant

    • Each area is ‘decomposed’ into a mixture (linear combination) of these factors, so by weighting and adding we get very close to the index

  • About these factors
    • The first factor is ‘the market’ - we see in the barchart that all our 9 cases contribute positively to it

    • The second factor is the difference between performance of low-priced (e.g. Northern) areas and high-priced (London) areas, as in the barchart

    • The third factor is the difference between mid-priced (e.g. Midlands) areas and the average of low and high-priced areas, as in blue

    • The data tells us that the latter 2 are cyclical, about 16-17 years, have similar amplitude, and that factor 3 ‘leads’ factor 2 by a quarter cycle

    • It will be evident to STEM students that these are a bit/a lot like sin and cos waves

    • It will be clear to traders, portfolio managers, and financial analysts that these are like ‘shift’ ‘twist’ and ‘butterfly’ on the yield curve, only here along the ‘\(£/m^2\) curve’

  • Why do we care?
    • The factors so far have not helped much but if we can understand ‘what they are’ we can understand 80-90% of market behaviour

    • [spoiler alert] It will turn out that they are almost entirely a reflection of price