Title | Computing the Chow statistic | |

Author | William Gould, StataCorp |

You can include the dummy variables in a regression of the full model and
then use the ** test**
command on those dummies. You could also run each of the models and then
write down the appropriate numbers and calculate the statistic by
hand—you also have access to functions to get appropriate
*p*-values.

Here is a longer answer:

Let’s start with the Chow test to which many refer. Consider the model,

y = a + b*x1 + c*x2 + u

and say we have two groups of data. We could fit that model on the two groups separately,

y = a1 + b1*x1 + c1*x2 + u for group == 1 y = a2 + b2*x1 + c2*x2 + u for group == 2

and we could fit a single, pooled regression

y = a + b*x1 + c*x2 + u for both groups

In the last regression, we are asserting that **a1==a2**, **b1==b2**,
and **c1==c2**. The formula for the “Chow test” of this
constraint is

ess_c - (ess_1+ess_2) --------------------- k --------------------------------- ess_1 + ess_2 --------------- N_1 + N_2 - 2*k

and this is the formula to which people refer. **ess_1** and **ess_2**
are the error sum of squares from the separate regressions, **ess_c** is
the error sum of squares from the pooled (constrained) regression, **k**
is the number or estimated parameters (**k**=3 in our case), and
**N_1** and **N_2** are the number of observations in the two groups.

The resulting test statistic is distributed **F(k, N_1+N_2-2*k)**.

Let’s try this. I have created small datasets:

clear set obs 100 set seed 1234 generate x1 = uniform() generate x2 = uniform() generate y = 4*x1 - 2*x2 + 2*invnormal(uniform()) generate group = 1 save one, replace clear set obs 80 generate x1 = uniform() generate x2 = uniform() generate y = -2*x1 + 3*x2 + 8*invnormal(uniform()) generate group = 2 save two, replace use one, clear append using two save combined, replace

The models are different in the two groups, the residual variances are different, and so are the number of observations. With this dataset, I can carry forth the Chow test. First, I run the separate regressions:

. regress y x1 x2 if group==1

Source | SS df MS | Number of obs = 100 | |

F(2, 97) = 19.08 | |||

Model | 156.695964 2 78.3479821 | Prob > F = 0.0000 | |

Residual | 398.206631 97 4.105223 | R-squared = 0.2824 | |

Adj R-squared = 0.2676 | |||

Total | 554.902595 99 5.60507672 | Root MSE = 2.0261 |

y | Coef. Std. Err. t P>|t| [95% Conf. Interval] | |

x1 | 3.586476 .6442618 5.57 0.000 2.307795 4.865158 | |

x2 | -1.915656 .7189693 -2.66 0.009 -3.342611 -.4887013 | |

_cons | -1.915656 .7189693 -2.66 0.009 -3.342611 -.4887013 | |

Source | SS df MS | Number of obs = 80 | |

F(2, 77) = 0.87 | |||

Model | 107.332801 2 53.6664005 | Prob > F = 0.4227 | |

Residual | 4745.17268 77 61.6256192 | R-squared = 0.0221 | |

Adj R-squared = -0.0033 | |||

Total | 4852.50548 79 61.42412 | Root MSE = 7.8502 |

y | Coef. Std. Err. t P>|t| [95% Conf. Interval] | |

x1 | -2.860412 2.840325 -1.01 0.317 -8.516223 2.795398 | |

x2 | 2.971855 3.161894 0.94 0.350 -3.324281 9.267991 | |

_cons | -1.108295 2.200774 -0.50 0.616 -5.490597 3.274006 | |

Then I run the combined regression:

. regress y x1 x2

Source | SS df MS | Number of obs = 180 | |

F(2, 177) = 0.34 | |||

Model | 21.1546997 2 10.5773499 | Prob > F = 0.7157 | |

Residual | 5587.34576 177 31.5669252 | R-squared = 0.0038 | |

Adj R-squared = -0.0075 | |||

Total | 5608.50046 179 31.3324048 | Adj R-squared = -0.0075 |

y | Coef. Std. Err. t P>|t| [95% Conf. Interval] | |

x1 | 1.081461 1.337802 0.81 0.420 -1.558633 3.721556 | |

x2 | -.2318499 1.489658 -0.16 0.876 -3.171626 2.707926 | |

_cons | -.1078511 1.056195 -0.10 0.919 -2.192207 1.976505 | |

For the Chow test,

ess_c - (ess_1+ess_2) --------------------- k --------------------------------- ess_1 + ess_2 --------------- N_1 + N_2 - 2*k

here are the relevant numbers copied from the output above:

ess_c = 5587.34576 (from combined regression) ess_1 = 398.206631 (from group==1 regression) ess_2 = 4745.17268 (from group==2 regression) k = 3 (we estimate 3 parameters) N_1 = 100 (from group==1 regression) N_2 = 80 (from group==2 regression)

So, plugging in, we get

5587.34576 - (398.206631+4745.17268) 443.96645 ------------------------------------ --------- 3 3 ----------------------------------------- = --------------- 398.206631 + 4745.17268 5143.3793 ----------------------- --------- 100+80-2*3 174 147.98882 = ---------- 29.559651 = 5.0064466

The Chow test is **F(k,N_1+N_2-2*k) = F(**3**,**174**)**, so our
test statistic is **F(**3**,**174**)** = 5.0064466.

Now I will do the same problem by running one regression and using
**test** to test certain coefficients equal to zero. What I want to do
is fit the model

y = a3 + b3*x1 + c3*x2 + a3'*g2 + b3'*g2*x1 + c3'*g2*x2 + u

where **g2**=1 if group==2 and **g2**=0 otherwise. I can do this by
typing

. generate g2 = (group==2) . generate g2x1 = g2*x1 . generate g2x2 = g2*x2 . regress y x1 x2 g2 g2x1 g2x2

Think about the predictions from this model. The model says

y = a3 + b3*x1 + c3*x2 + u when g2==0 y = (a3+a3') + (b3+b3')*x1 + (c3+c3')*x2 + u when g2==1

Thus the model is equivalent to fitting the separate models

y = a1 + b1*x1 + c1*x2 + u for group == 1 y = a2 + b2*x1 + c2*x2 + u for group == 2

The relationship being

a1 = a3 a2 = a3 + a3' b1 = b3 b2 = b3 + b3' c1 = c3 c2 = c3 + c3'

Some of you may be concerned that in the pooled model (the one estimating
**a3**, **b3**, etc.), we are constraining the var(u) to be the same
for each group, whereas, in the separate-equation model, we estimate
different variances for group 1 and group 2. This does not matter, because
the model is fully interacted. That is probably not convincing, but what
should be convincing is that I am about to obtain the same
**F(**3**,**174**)** = 5.01 answer and, in my concocted data, I
have different variances in each group.

So, here is the result of the alternative test coefficients against 0 in a pooled specification:

. generate g2 = (group==2) . generate g2x1 = g2*x1 . generate g2x2 = g2*x2 . regress y x1 x2 g2 g2x1 g2x2

Source | SS df MS | Number of obs = 180 | |

F(5, 174) = 3.15 | |||

Model | 465.121148 5 93.0242295 | Prob > F = 0.0096 | |

Residual | 5143.37931 174 29.5596512 | R-squared = 0.0829 | |

Adj R-squared = 0.0566 | |||

Total | 5608.50046 179 31.3324048 | Root MSE = 5.4369 |

y | Coef. Std. Err. t P>|t| [95% Conf. Interval] | |

x1 | 3.586476 1.728796 2.07 0.040 .174367 6.998585 | |

x2 | -1.915656 1.929264 -0.99 0.322 -5.723428 1.892115 | |

g2 | -1.471946 2.056721 -0.72 0.475 -5.531279 2.587387 | |

g2x1 | -6.446889 2.618856 -2.46 0.015 -11.6157 -1.278075 | |

g2x2 | 4.887512 2.918483 1.67 0.096 -.8726743 10.6477 | |

_cons | .3636508 1.380901 0.26 0.793 -2.361822 3.089124 | |

Same answer.

This definition of the “Chow test” is equivalent to pooling the data, fitting the fully interacted model, and then testing the group 2 coefficients against 0.

That is why I said, “Chow Test is a term I have heard used by economists in the context of testing a set of regression coefficients being equal to 0.”

Admittedly, this leaves a lot unsaid.

The issue of the variance of u being equal in the two groups is subtle, but I do not want that to get in the way of understanding that the Chow test is equivalent to the “pool the data, interact, and test” procedure. They are equivalent.

Concerning variances, the Chow test itself is testing against a pooled, uninteracted model and so has buried in it an assumption of equal variances. It is really a test that the coefficients are equal and variance(u) in the groups are equal. It is, however, a weak test of the equality of variances because that assumption manifests itself only in how the pooled coefficient estimates are manufactured. Because the Chow test and the “pool the data, interact, and test” procedure are the same, the same is true of both procedures.

Your second concern might be that in the “pool the data, interact, and test” procedure there is an extra assumption of equality of variances because everything comes from the pooled model. As shown, this is not true. It is not true because the model is fully interacted, so the assumption of equal variances never makes a difference in the calculation of the coefficients.

In Stata 12 or more recent versions, you can also use the
**contrast**
command with factor variables to perform the same test:

. regress y c.x1##i.g2 c.x2##i.g2

Source | SS df MS | Number of obs = 180 | |

F(5, 174) = 3.15 | |||

Model | 465.121148 5 93.0242295 | Prob > F = 0.0096 | |

Residual | 5143.37931 174 29.5596512 | R-squared = 0.0829 | |

Adj R-squared = 0.0566 | |||

Total | 5608.50046 179 31.3324048 | Root MSE = 5.4369 |

y | Coef. Std. Err. t P>|t| [95% Conf. Interval] | |

x1 | 3.586476 1.728796 2.07 0.040 .174367 6.998585 | |

1.g2 | -1.471946 2.056721 -0.72 0.475 -5.531279 2.587387 | |

g2#c.x1 | ||

1 | -6.446889 2.618856 -2.46 0.015 -11.6157 -1.278075 | |

x2 | -1.915656 1.929264 -0.99 0.322 -5.723428 1.892115 | |

g2#c.x2 | ||

1 | 4.887512 2.918483 1.67 0.096 -.8726743 10.6477 | |

_cons | .3636508 1.380901 0.26 0.793 -2.361822 3.089124 | |

df F P>F | ||

g2 | 1 0.51 0.4752 | |

g2#c.x1 | 1 6.06 0.0148 | |

g2#c.x2 | 1 2.80 0.0958 | |

Overall | 3 5.01 0.0024 | |

Residual | 174 | |

An additional example can be found in the
“Chow tests” section of [R] **contrast**.