# Consumption function with negative intercept?

I am doing an assignment where I calculate the consumption function of a certain country from empirical data and I am getting a negative intercept. How can this be interpreted because usually, isn't the intercept always positive?

I assume that you ran a linear regression to find the $C(Y)$ function and by negative intercept you mean $C(0)<0$. This does not contradict anything because what you have is an estimate of the $C(Y)$ function based on empirical data that was probably centered around some non-zero value of $Y$. So it is quite possible that the function's estimate is incorrect far from the observed range of $Y$, especially if you were using a linear regression which is a very basic functional form.

If the actual $C(Y)$ consumption function, not the estimate, were such that $C(0) < 0$ then indeed we would have a contradiction. But as long as you are clear that what you have is an estimate for $C(Y)$ this problem does not arise.

• Thanks for the answer. I did use a linear regression to find the $C(Y)$ function. I was wondering what empirical conclusions can I gather about this country? Would it imply that the people in that particular prefer to save when they do not have income? Furthermore, my marginal propensity to consume is greater than 1, so it seems to violate Keynes assumptions. – Mat.S Jan 20 '16 at 23:50
• @Mat.S As I sad, $C(0)$ basically has no meaning, it is one of the worst estimated points of your function. This and the marginal propensity problem most likely arises due to the function form (linear) or some other variable that you did not include in your regression. – Giskard Jan 21 '16 at 7:51

To supplement the above answer, you can force your regression to not have a constant. In this case, economic theory will indicate that $C(0)$ is at least non zero.

• If the regression does not have a constant wouldn't that imply exactly that $C(0) = 0$? (And mess up other more relevant points of the estimate as well.) – Giskard Jan 21 '16 at 7:52
• Yes, it would imply that. But thats my point- in this sense, a constrained regression would make sense as theory tells us that in the population, $C(0)=0$ – ChinG Jan 21 '16 at 16:50
• Your answer says "theory will indicate that $C(0)$ is at least non zero" which conflicts with your comment or I misunderstand something. I would also add that I don't think theory claims $C(0) = 0$, people would take loans or use savings. – Giskard Jan 21 '16 at 16:52
• Sure, which drove me to my comment that $C(0)$ is at least non-zero. People would consume something, even with 0 income. My point is simply that if we do not precisely know the value of the constant, we know at least in the population model via econ theory that the level of autonomous consumption is non-negative. So, given that the OP has estimated a negative constant, I think a plausible fix is to fit a non-constant model, rather than arbitrarily selecting a positive constant. – ChinG Jan 21 '16 at 16:57
• Now I understand. I still disagree with proposed fix :) But I am not an expert in this area so you may well be right. – Giskard Jan 21 '16 at 17:04