Nov

19

Chartomatic, from Kim Zussman

November 19, 2007 |

Instead of looking at charts, one way check for similarities between periods would be to regress returns of recent period against prior periods. For example, how do the 46 weekly returns of 2007 (SPY, 1/07-last week) compare against the equivalent (aligned) 1st 46 weekly returns of the prior 13 years? One would expect that years most like 2007, in terms of weekly stock returns, would be highly correlated. Here is regression of 1st 46 weeks of 2007 vs prior 13 years:

Regression Analysis: rt 07 versus rt 06, rt 05, …

The regression equation is rt 07 = 0.00101 - 0.141 rt 06 - 0.319 rt 05 - 0.200 rt 04 - 0.007 rt 03 - 0.302 rt 02 + 0.026 rt 01 + 0.042 rt 00 + 0.026 rt 99 + 0.088 rt 98 - 0.115 rt 97 +
0.015 rt 96 - 0.000 rt 95
+ 0.228 rt 94

Predictor Coef SE Coef T P VIF
Constant 0.0010 0.0039 0.26 0.796
rt 06 -0.1414 0.2369 -0.60 0.555 1.4
rt 05 -0.3189 0.2609 -1.22 0.231 1.5
rt 04 -0.1996 0.2005 -1.00 0.327 1.2
rt 03 -0.0075 0.1651 -0.05 0.964 1.4
rt 02 -0.3019 0.1229 -2.46 0.020 1.5
rt 01 0.0257 0.1169 0.22 0.828 1.6
rt 00 0.0421 0.1028 0.41 0.685 1.3
rt 99 0.0261 0.1211 0.22 0.831 1.3
rt 98 0.0884 0.1191 0.74 0.463 1.2
rt 97 -0.1148 0.1448 -0.79 0.434 1.2
rt 96 0.0146 0.1737 0.08 0.933 1.2
rt 95 -0.0001 0.3847 -0.00 1.000 1.7
rt 94 0.2276 0.2395 0.95 0.349 1.4

S = 0.0195096 R-Sq = 29.8% R-Sq(adj) = 1.3%

None of the prior 13 years is positively correlated, on a week-to-week basis. But 2002 appears to be an "anti-2007", in that equivalent weekly returns are significantly negatively correlated. Shown here plotting SPY weekly closes for 2002 and 2007 overlaid.


Comments

Name

Email

Website

Speak your mind

Archives

Resources & Links

Search