Figure 14.5 Cross-correlation: S&p and T-bonds.

Cross-Correlation Function (Lag, Cross-Correlation, Standard Error) First: T-BONDS Lagged: S&P

jhli

ibiiii.

i >i!«!ii

iiiiiiitiiiiiin

ill!"1!1

;".....;"!!

lllliiii! Silillili

.iillllll nmi

iiiiiiiiiiiii

i,.MwU I.,.

HilijiiHiijj

ffiffiffi!

, ,;!, ! i " hill , , „, S , i i,

iii !B iii ! Will !!}!!!

!!!! Wifilfi!iftiij

!i!lilpilljiii!!ii!!i! Illili!: .....iiiiiliiiiiiiiii hi

liijli!

i iililll!

ii,,!

iljii il

HliHi

lllliillllilSlJil

ihiii

iii-gifHtiUil iiUitj! Niili if.im-..L I. FtM."

.. .iiliiif lip) ii! IlllllUli..,.............., .„

Illy ; a ....... ..luirui,,

iiii! ilii ! Hi ii lilillilil 111 1

l j in! hiiiijiiiiiuei

ti!!!!!!ii!!!!!!!i!l! ! !!! !! !! !!!;!!! !!! !!5!)

!li llpl! ill! i !!1!] llll !!lil§ 11 111!

1..1,„i1i..i..i,,,..n,... j.,,.,,,.,,.,.... s;iM""t , in,,in .I..,,; !,.,..„..\,„esi.....iiH>iF>iiii lii>i>i

1 1 1 111 11111111111 1111 ¹

i!!iiii!iiii!!li!!l!i!!!iiiiiij!i!i!!i!i!iiif!!B!ili!i

"i.....i.....I......."......(Bi.......i.....»11

Iii........i!.....1..... =Bi""i" !

-0.8

-0.6

-0.4

......s......

.l-gili.

-0.2

Source: Statsoft Inc. Statistica. Permission received.

Proving That T-Bonds Lead the S&P

Once we have examined the first data stream, it is valuable to look at the second correlated data stream to determine the nature of the relationship. To determine the correlation between the S&P and T-bonds, a cross-correlation analysis is performed using STATISTICA™. As seen in Figure 14.5, the cross-correlation with 0 lag is .9135, which is highly significant.

A more difficult question is whether one is a leading indicator of the other, and by how much. It is seen that this data is highly correlated, with positive or negative lags of almost any amount showing a high correlation. With T-bonds leading the S&P by 90 days, the correlation is still .8646. With T-bonds lagging the S&P, the correlation is .7810. The slower degradation of the correlation when T-bonds led the S&P shows that the leading variable is T-bonds.

A simple visual graphical interpretation provides a rough confirmation of the correlation and an understanding of which market leads the other, and by how much. The S&P and T-bonds are plotted on the same chart, with different scales. In Figure 14.6, the S&P is shown by a scale on the left of the chart, and T-bonds are shown by a scale on the right. It is constructed by mathematically matching the slope of the trendlines of

Figure 14.6 Cross-correlation: S&p and T-bonds.

the two data streams. It is evident that T-bond prices reach peaks and bottoms before the S&P prices.

The 9/81 bottom in bonds preceded the stock market bottom in 8/82 by 11 months. The 4/87 breakdown in bonds preceded the 8/87 stock market top by 4 months. From this observation, it is easy to make a simple system to determine conclusively whether T-bonds lead the S&P We can do this with a simple trading system written in TradeStation whereby buy-and-sell signals are triggered by leading movements in T-bond prices. We can then measure the profitability of the trading strategy and use the results to confirm the hypothesis about market interaction.

The Intermarket Moving Average system, shown in Figure 14.7, generates a buy signal for the S&P when the closing price of T-bonds crosses over its own moving average. Similarly, a sell signal for the S&P is generated when the closing price of T-bonds crosses below its own moving average.

N is the number of bars to use in the T-bond moving average. In running the Optimization routine of TradeStation, it is found that every moving average between 20 and 40 days yields a substantial profit. For example, Figure 14.8 shows results for the top 10 values of N.

By contrast, a comparable one-line moving average system based on a close of the S&P above or below a moving average of S&P prices would prove to be grossly unprofitable for all values of N. This clearly shows that T-bond prices lead S&P prices.

Page 1

Type : System

Name : Intermarket Mov Avg

Notes

Last Update : 02/07/97 10:02am Printed on : 02/07/97 10:16am Verified : YES

(Datal is theS&P. Data2 is T-Bonds. This system generates trading signals for the S&P based on a close of T-Bonds above or below its moving average. Almost any value for N develops significant profits, which points to T-Bonds being a leading indicator for the S&P)

Input: N(27);

If Close Data2 crosses above Average(Close of Data2,N) then Buy {the S&P) on Open,

If Close Data2 crosses below Average(Close of Data2, N) then Sell {the S&P) on Open,

Source: Printed using TradeStation PowerEditor by Omega Research version 4.02.15-Jul < 09 1996.

FIGURE 14.8 OPTIMIZATION: INTERMARKET MOVING AVERAGE.

Optimization. Intermarket Moving Average

Optimize N = 27 to N = 36. Subtract Comrnisions = $25, Slippage = $50

Source: Printed using TradeStation by Omega Research Version 4.02.15-Jul < 09 1996.

FIGURE 14.7 SYSTEM: INTERMARKET MOVING AVERAGE.