Back to Overview


 

Consumption and Investment

Shortly after Keynes published The General Theory, economists had a chance to test one of its central assumptions, that consumption depended on income. As a result of efforts to mobilize the economy for World War II, the U. S. government developed and made available in 1942 comprehensive national accounts data for the previous decade. Economists used the data to plot a consumption function, and found that the points almost formed a straight line. Since income could explain almost all changes in consumption, this line seemed to imply that any other influences must be minor.

However, this procedure of finding a consumption function overstates the influence of income on consumption. It does not allow for any effect of consumption on income, but assumes that the only source of the relationship comes from the effect of changes in income on changes in consumption. But this clearly cannot be because consumption is the largest part of income. Thus even if income had no influence on consumption, there would be a relationship between income and consumption.

To see this, consider the procedure shown in the table below, which contains thirty random digits (taken from a table of random digits) split into two groups of fifteen each and labeled A and B. Column C is obtained by multiplying each number in group A by three. Column Y is obtained by adding columns B and C. If we try to explain the C column by using the Y column in the same manner that the early estimates of the consumption function were done, we will find that the variation in Y explains much of the variation in C even though Y had nothing to do with the way in which C was obtained.

Building a "Consumption Function" with Random Digits
A
B
C
Y
1
2
3
5
8
1
24
25
3
6
9
15
8
2
24
26
2
1
6
7
7
4
21
25
9
0
27
27
5
5
15
20
7
9
21
30
1
4
3
7
5
9
15
24
6
2
18
20
5
7
15
22
1
5
3
8
7
0
21
21

To get a better estimate of the influence of income on consumption, we need to take out any influence that flows from consumption to income. A bit of mathematics and analysis that can wait for more advanced courses can solve this feedback problem. When it is solved, the relationship between consumption and disposable income is weakened only a little. Disposable income remains the dominant determinant of consumption spending, but it is not the only thing that can matter.

Ignoring feedback led many economists to emphasize that the correlation between consumption and income resulted solely from effects of income on consumption. As a result, their estimations of the simple consumption function gave results that looked stronger than they actually were. In contrast, ignoring feedback between investment and income led many economists to treat the correlation here as solely the result of effects of investment on income. To many of these economists, a simple income-expenditure model largely explained how the world worked. Autonomous (and thus unexplainable) changes in investment would change income, and these changes would be multiplied by the feedback between income and consumption. If, however, one can explain changes in investment by changes in income or financial conditions, then the simple income-expenditure model will be a partial and perhaps misleading way to view the world.

The IS curve is built assuming that business investment should be sensitive to both interest rates and expectations of future earnings. In addition, investment includes residential construction and changes in inventories. These forms of investment respond to financial conditions and total spending, though not in the same ways as business investment. As with consumption, changes in investment may react to changes in the economy with a time lag.

Changes in income can also affect investment through the accelerator principle. In 1939 Paul Samuelson published an article that showed how the accelerator could be combined with the income expenditure model to give a dynamic model. In this model investment causes changes in income and in turn changes in income (or consumption, depending on how the model is structured) cause changes in investment.


Back to Overview
Copyright Robert Schenk