The highest rated answer to the question on What concepts are the most dangerous ones in quantitative finance work? is this one:


Correlations are notoriously unstable in financial time series [...]

My question
My question is a little bit broader than just about linear dependence, it is:
What is the most stable, non-trivial dependence structure in financial data?

With non-trivial I mean that I don't want answers that are about direct connections, e.g. between derivative and underlying.

The dependence structure can be either cross-sectional or through time with univariate time series, it can also be non-linear.

The context of my question is that I am preparing the documentation for a new machine learning R package I wrote and I am looking for a good showcase in the financial sphere. Now this is not a trivial feat given that correlations are notoriously... see above ;-)

  • 1
    $\begingroup$ This is an interesting question. Have you a specific asset class in mind? $\endgroup$ – Quantuple Jun 17 '16 at 19:30
  • $\begingroup$ @Quantuple: Thank you. Well, not necessarily although I am more of an equities guy. $\endgroup$ – vonjd Jun 17 '16 at 19:35
  • 1
    $\begingroup$ I see. Personally I have no good answer to your question... more specifically I have never been able to identify a "stable" dependence structure. It's essentially regimes (patterns) IMHO. Plus, the more complicated the dependence structure model, the greater the possibility of (over)fitting the data. I'm definitely looking forward to reading the answers your receive. $\endgroup$ – Quantuple Jun 17 '16 at 19:43
  • $\begingroup$ I've not looked at it, but i'd imagine things like convexity and skew, taken across the whole market (on a per asset class basis) - maybe convexity / skew as a function of credit spread or something? would be pretty stable - within the same regimes. Would be intrigued to see what those relationships look like. $\endgroup$ – will Aug 17 '16 at 14:04
  • 1
    $\begingroup$ I can only agree with the other commentators. This a good question, simple but with deep implications. Indeed this is the essence of what we look for when we design trading systems. I too cannot clearly define any relationship that I would consider stable. I guess the relationship between risk and return is the closest to a universal law I can think of. Perhaps look for something from behavioural finance, i.e. about human biases that we can't easily avoid like asymmetric attitudes to profit and loss. E.g. markets crash faster/more than they rise. $\endgroup$ – Paul Young Nov 1 '17 at 9:20

It is hard to find a stable non-trivial dependence structure in financial data. Usually when such is found it is hard to rationalize.

One of my favorite (although I am sure there are others) is the so called "Presidential Puzzle". This is an old finding by Santa-Clara and Valkanov (2003) They find that "

Excess return in the stock market is higher under Democratic than Republican presidencies: 9 percent for the value‐weighted and 16 percent for the equal‐weighted portfolio.

At the time the finding was very robust and did not seem to be explained by anything else. What is more impressive is that 12 years later the result still holds true. We now have a very good out-of-sample period. This is confirmed by Pastor and Veronesi (2017) recent work. More interesting, they racionalize the finding by building a continuous-time general equilibrium model and conclude that:

When risk aversion is high, agents are more likely to elect the party promising more fiscal redistribution. The model predicts higher average stock market returns under Democratic than Republican presidencies, explaining the well-known “presidential puzzle.”

  • 1
    $\begingroup$ @vonjd thanks for accepting my answer as the answer. I believe I did not exactly answer your question as you ask for the "most non-trivial stable dependency". I guess my example shows a dependency that is stable and non-trivial ... not sure if the "most" stable and most "non-trivial". $\endgroup$ – phdstudent Mar 21 '18 at 15:44

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.