Structural equation modeling for objective financial forecasts
Happy New Year Paolo! :-)Early days yet, but its seems your forecasts have had a good start to the year.VIX is being "unusually" suppressed given wider recent US indice weakness. Gold is still weak and projected by mass media to weaken into Q1-2014. Similar case for silver too.The Dow may finally begin its long awaited correction, now that Fed has begun tapering. Check out this article:http://gestaltu.com/2013/11/boe-paper-signals-worrisome-outlook-equities-post-qe.htmlHave you considered forecasting the S&P 500? I feel it is more representative that the Dow 30, which is also price-weighted (arithmetic average) instead of market-cap weighted (geometric average) such as SPX and Nasdaq. The Dow hence measures "absolute" price changes regardless of component size.Maybe the results and Z-score will be better?Is it possible to forecast 10-yr bond rates? I don't trade bonds, but it can be useful for general awareness.All the best for 2014 and thanks again for choosing to share your forecasts!Mr Anon
Thanks, Mr A, and best wishes for a healthy and prosperous New Year to you as well. I chose the DJIA as a proxy for the business cycle because it allowed me to look at secular cycles, given that the EOD data goes back to 1896. As part of my end of the year review, I looked at the DJIA model changes over time, and was surprised at the behavior of the longest cycles over the past year. The longest cycle period was estimated at 140 years +/- 5 years at the beginning of 2013, and is now extended to 145 years +/- 6 years. The second longest cycle period was estimated at 42 +/- 3 years, which has now extended to 45 years +/- 4 years. Since both of these cycles are pointed down, the phase extension indicates the model estimates are still within error. I simply don't have the CPU power right now to figure out the confidence intervals for the cycle model estimates, but I strongly suspect that as deviant as the DJIA model predictions may be, they are still within the range of possibilities given that the data point to a possible secular top. The other outstanding issue as far as model internal consistency is the deviation between the gold and silver models. I agree with many readers here who have commented that it is unlikely that the gold and silver models are congruent - one of them is likely incorrect. Going forwrd, any significant deviation in actual vs predicted prices in either the gold or silver model will be taken as evidence of the weaker model. On other fronts, what is interesting is that ten year rates began to rise in April of 2013, well ahead of any serious Fed taper talk, suggesting that the bond market began to have doubts about the monetary resolution of QE. The Fed has implicitly agreed to extend temporary credit, but not permanent liquidity via monetization, but I believe the bond market is having second thoughts and the current "mini-QE taper" is an attempt by the Fed to address these doubts. Going forwards, I will think about developing a couple of cycle models for the SP and TYN for this year. Paolo
Hi Paolo,It seems that gold is tracking your model well, especially now that major analysts are (only now) predicting sizable declines this year. This in itself might create shorting and selling.The recent increased correlation between gold and SPX has encouraged me to consider gold as an inverse proxy to SPX. It certainly "fits into" impending expectations of a correction.Have you ever come across Sornette bubbles?http://gummy-stuff.org/sornette.htm?utm_source=mcverryreport.comIt seems that the major US indices might be within the midst of one, though it make take several weeks for it to start deflating. It should be noted that approx 40% of identified bubbles move sideways rather than implode.Although fundamentally and technically (TA, EW etc) many are calling for a strong correction, I suspect an upward break (Sornette blow-off) would create a bear trap before heading downwards.Kind regards,Mr Anon
Mr. A, I am familiar with Sornette's work, who like B Mandelbrot came to the same conclusion, in different words namely that volatility can be more easily predicted than price level. However, due to the fact that in a price-time series, both time and price exhibit fractal patterns, ie multifractal, if you believe Mandelbrot and others, ligating the price function to time and predicting direction remains problematic, in that kind of model. In other words, one might predict large price changes at some point, but the direction of the change can remain elusive. For cyclical models, the key assumptions are that the price changes reflect changes in the business cycle(s), and that business cycles exist at all in the first place. All that said, I think you are right in that we may see some extreme ups and down before anything resembling a trend once again re-emerges at intermediate and long term time scales.
Post a Comment