- Understand and use time-series graphs, tables, pie charts, and bar charts to illustrate data and relationships among variables.
You often see pictures representing numerical information. These pictures may take the form of graphs that show how a particular variable has changed over time, or charts that show values of a particular variable at a single point in time. We will close our introduction to graphs by looking at both ways of conveying information.
One of the most common types of graphs used in economics is called a time-series graph. A time-series graph shows how the value of a particular variable or variables has changed over some period of time. One of the variables in a time-series graph is time itself. Time is typically placed on the horizontal axis in time-series graphs. The other axis can represent any variable whose value changes over time.
The table in Panel (a) of Figure 21.13 “A Time-Series Graph” shows annual values of the unemployment rate, a measure of the percentage of workers who are looking for and available for work but are not working, in the United States from 1998 to 2007. The grid with which these values are plotted is given in Panel (b). Notice that the vertical axis is scaled from 3 to 8%, instead of beginning with zero. Time-series graphs are often presented with the vertical axis scaled over a certain range. The result is the same as introducing a break in the vertical axis, as we did in Figure 21.5 “Canceling Games and Reducing Shaquille O’Neal’s Earnings”.
The values for the U.S. unemployment rate are plotted in Panel (b) of Figure 21.13 “A Time-Series Graph”. The points plotted are then connected with a line in Panel (c).
Scaling the Vertical Axis in Time-Series Graphs
The scaling of the vertical axis in time-series graphs can give very different views of economic data. We can make a variable appear to change a great deal, or almost not at all, depending on how we scale the axis. For that reason, it is important to note carefully how the vertical axis in a time-series graph is scaled.
Consider, for example, the issue of whether an increase or decrease in income tax rates has a significant effect on federal government revenues. This became a big issue in 1993, when President Clinton proposed an increase in income tax rates. The measure was intended to boost federal revenues. Critics of the president’s proposal argued that changes in tax rates have little or no effect on federal revenues. Higher tax rates, they said, would cause some people to scale back their income-earning efforts and thus produce only a small gain—or even a loss—in revenues. Op-ed essays in The Wall Street Journal, for example, often showed a graph very much like that presented in Panel (a) of Figure 21.14 “Two Tales of Taxes and Income”. It shows federal revenues as a percentage of gross domestic product (GDP), a measure of total income in the economy, since 1960. Various tax reductions and increases were enacted during that period, but Panel (a) appears to show they had little effect on federal revenues relative to total income.
Laura Tyson, then President Clinton’s chief economic adviser, charged that those graphs were misleading. In a Wall Street Journal piece, she noted the scaling of the vertical axis used by the president’s critics. She argued that a more reasonable scaling of the axis shows that federal revenues tend to increase relative to total income in the economy and that cuts in taxes reduce the federal government’s share. Her alternative version of these events does, indeed, suggest that federal receipts have tended to rise and fall with changes in tax policy, as shown in Panel (b) of Figure 21.14 “Two Tales of Taxes and Income”.
Which version is correct? Both are. Both graphs show the same data. It is certainly true that federal revenues, relative to economic activity, have been remarkably stable over the past several decades, as emphasized by the scaling in Panel (a). But it is also true that the federal share has varied between about 17 and 20%. And a small change in the federal share translates into a large amount of tax revenue.
It is easy to be misled by time-series graphs. Large changes can be made to appear trivial and trivial changes to appear large through an artful scaling of the axes. The best advice for a careful consumer of graphical information is to note carefully the range of values shown and then to decide whether the changes are really significant.
Testing Hypotheses with Time-Series Graphs
John Maynard Keynes, one of the most famous economists ever, proposed in 1936 a hypothesis about total spending for consumer goods in the economy. He suggested that this spending was positively related to the income households receive. One way to test such a hypothesis is to draw a time-series graph of both variables to see whether they do, in fact, tend to move together. Figure 21.15 “A Time-Series Graph of Disposable Income and Consumption” shows the values of consumption spending and disposable income, which is after-tax income received by households. Annual values of consumption and disposable income are plotted for the period 1960–2007. Notice that both variables have tended to move quite closely together. The close relationship between consumption and disposable income is consistent with Keynes’s hypothesis that there is a positive relationship between the two variables.
The fact that two variables tend to move together in a time series does not by itself prove that there is a systematic relationship between the two. Figure 21.16 “Stock Prices and a Mystery Variable” shows a time-series graph of monthly values in 1987 of the Dow Jones Industrial Average, an index that reflects the movement of the prices of common stock. Notice the steep decline in the index beginning in October, not unlike the steep decline in October 2008.
It would be useful, and certainly profitable, to be able to predict such declines. Figure 21.16 “Stock Prices and a Mystery Variable” also shows the movement of monthly values of a “mystery variable,” X, for the same period. The mystery variable and stock prices appear to move closely together. Was the plunge in the mystery variable in October responsible for the stock crash? The answer is: Not likely. The mystery value is monthly average temperatures in San Juan, Puerto Rico. Attributing the stock crash in 1987 to the weather in San Juan would be an example of the fallacy of false cause.
Notice that Figure 21.16 “Stock Prices and a Mystery Variable” has two vertical axes. The left-hand axis shows values of temperature; the right-hand axis shows values for the Dow Jones Industrial Average. Two axes are used here because the two variables, San Juan temperature and the Dow Jones Industrial Average, are scaled in different units.