Contents:

• Selecting and responding to data in exams
• Top ten exam faults when handling data

When you are studying data drawn from real-life sources, as opposed to material from fictitious scenarios, it's important to know the difference between time series and cross-sectional data.Time series data observe how economic variables change over time: from year to year, quarter to quarter, and month to month. So, for example, UK GDP data for the last five years would form a short time series.

But whereas time series data are often aggregated, cross-sectional data split up or dis-aggregate the data into its component parts. Cross-sectional and time series data can be combined, when they are jointly referred to as pooled data.

Time series data usually fluctuate in two ways, seasonally and cyclically, according to the peaks and troughs of the business cycle. Seasonal fluctuations are not shown in annual data, but they would be noticeable in monthly or quarterly data. That is unless they have been adjusted to get rid of the seasonal variation. Seasonally adjusted data are better for displaying the long-term trend from year to year. Unadjusted data show the variation from season to season within each year of the data series.

It is often given as a warning that you should take care when interpreting a short annual time series of less than ten years. This period is usually insufficient to account for a whole business cycle. With such data it is easy to confuse relatively short term fluctuations associated with the movements of the business cycle with the long term trend.

As a general rule, time series data should extend over more than one business cycle, in order to allow for a long term trend to be detectable. Where it is possible to detect a long term trend from the data, it may also be possible to extrapolate the trend, in order to predict or forecast future behaviour of the variable(s).

Be careful over how you use long term time series data to forecast (or extrapolate) the future. You should bear in mind that there is always the possibility that a structural change could occur. This type of event is often referred to as an 'exogenous' shock and it may have the effect of rendering the forecast unusable. Even when the long term trend is estimated, for example by measuring the changes in the economy from peak to peak, over a number of business cycles, there is always the possibility that structural changes have occurred in the economy that have caused the underlying trend to change.

An example of this could be seen in the long-term trend rate of growth in the UK economy. Has it changed so that higher output growth in the economy is possible without the threat of inflation?

Many economic variables are affected by inflation, which can seriously distort time series data. You should remember to check your data to see if it is unadjusted for inflation, in which case the data is presented in the current prices of each year in the data series, or whether the data have been converted into the constant prices of a particular year in the series. Alternatively, the data could have been converted into index numbers. Current price data that is unadjusted to account for inflation, are often called nominal data, whereas constant price and index data are called real data. Index numbers can easily be confused with data that are expressed in percentage terms.

Selecting and responding to data in exams:

Examiners often report that candidates have difficulties with exam questions that require data handling. It is always important to remember that these problems can be dealt with at an earlier stage than in the examination itself! So take some time to review the list of common faults that follows, and try to avoid the pifalls yourself:

Top ten exam faults when handling data

1. Using words like 'vast' and 'massive' to describe small changes in data.
2. Lack of awareness of the difference between time series and cross-sectional data
3. Confusing indexed data with percentage data
4. Confusing nominal and real data
5. Lack of awareness of role of the business cycle in time series data
6. Failure to distinguish the cyclical variations around the trend
7. Tendency to extrapolate the upswings and downturns within data sets
8. Inability to detect correlations
9. Inability to relate correlations to their causes and effects
10. Failure to identify key features of data sets
11. Tendency to convert numerical data simply into written statements

OK, so there are eleven faults here! It just goes to show how statistics can mis-used. Follow the rules in the list and you're less likely to upset your examiner. Trust me, that's important!