TABLE 2: REGRESSIONS OF U.S.  DEFENSE SPENDING ON SUPPORT FOR OTHER NATIONS, WARS, ELECTION CYCLES, UNEMPLOYMENT, REAGAN BUILDUP,

AND THE COLD WAR, FY 1946-1998.

 

MODEL

 

 

Variables

OLS

Real Defense

Budget

Authority

AUTO

Real Defense

Budget

Authority

OLSa

First

Differences

in Real

Defense BA

AUTOa

First

Differences

in Real

Defense BA

AUTOa

First

Differences

in Real

Defense BA

Prior Spendingb

.68***

(8.33)

.96***

(24.03)

0.19***

(3.23)

0.30**

(2.21)

0.22

(1.60)

Internatl Support

0.98

(1.23)

0.20

(0.43)

0.54

(1.27)

0.35

(0.80)

---

Korean War

82547***

(2.56)

128430***

(6.53)

117440***

(6.62)

109300***

(6.14)

116100***

(6.53)

Kennedy

Buildup

384

(0.02)

24909

(1.45)

22953

(1.52)

30264*

(1.88)

21952

(1.42)

Vietnam

War

47693**

(2.49)

33019*

(1.96)

32953**

(2.21)

25688

(1.62)

25462*

(1.65)

Gulf

War

77596*

(1.84)

28909*

(1.71)

32467**

(2.17)

21665

(1.55)

27442**

(1.95)

Presidential

Election

-11841

(-0.78)

-5953

(-1.00)

-3864

(-0.73)

-5259

(-1.05)

---

Unemploy-

ment Pct t-1

20272***

(3.98)

-883

(-0.28)

1365

(0.47)

-1847

(-0.63)

---

Reagan

Buildup

9710

(0.41)

26700

(1.55)

26091*

(1.72)

20454

(1.29)

25836*

(1.68)

Cold

War

18903

(1.28)

47119**

(2.48)

48487***

(2.87)

44134**

(2.49)

44501**

(2.49)

Constant

 

-76718*

(-1.95)

206630***

(3.48)

---

---

---

R Square

.82

.94

.78

.74

.72

Adj R Sqr

.77

.92

.73

.68

.69

DW Stat

1.03c

1.27c

1.91

1.88

1.94

DH Stat

2.49d

1.46

0.32

0.82

0.05

Residual

Rho e

 

0.28

 

0.19

 

0.04

 

0.03

 

0.00

Runs Normal Statistic

 

-3.08 f

 

-1.12

 

-0.68

 

-0.87

 

-0.29

 

T-ratios are in parentheses; SHAZAM OLS and AUTO estimates.

* - Statistically significant at the 0.10 level; ** - significant at the 0.05 level; and *** - significant at the 0.01 level, two-tailed tests.

a - All variables have been differenced once.

b - In OLS models prior spending is real defense budget authority or differences in real defense budget authority lagged one year and used as an independent variable; in AUTO models, this is rho, an estimate of the first-order autoregressive process in the error term using Cochrane-Orcutt procedures in SHAZAM.

c - Evidence from the Durbin-Watson statistic of residual autocorrelation.

d - Evidence from the Durbin h statistic of remaining higher order autocorrelation.

e - An estimate of the first order autocorrelation remaining in regression residuals.

f - Evidence from a normalized runs test of non-random residuals.