RDP 1999-03: Householders' Inflation Expectations 3. Differences Between Householders

Since 1995, respondents have also been asked a range of questions about particular personal characteristics, including gender, location,^[3] income, age, occupation, education, political tendency and home ownership (Appendix A). In general, it would be expected that people with the most accurate inflation expectations would be those with better access to information, those with more developed information-processing skills, or those for whom the benefits of forming an accurate assessment of the prospects for inflation are relatively high. This suggests that occupation characteristics are likely to be important. When people work in areas which directly involve exposure to prices – because they sell goods or services, because they ‘set’ prices, or because they need information about prices to do their job – they would be expected to have relatively more accurate inflation expectations. Occupation categories would also tend to be correlated with education, income and location. It is more difficult to see, however, how other characteristics, such as gender, political preference or home ownership, impact on inflation expectations.

Using a probit model, Harris and Harding (1998b) have examined how inflation expectations in October 1996 differed according to personal characteristics. They found that people were less likely to give a quantitative answer if they were older or had a lower education, and more likely to give a quantitative answer if they were male or had a higher income. Of the people who answered quantitatively, they found that the young, females, the poor and those in occupations other than managers/administrators and professionals had higher inflation expectations. They also found that people are more likely to answer in relatively large round numbers when they were older, were professionals or labourers, but were less likely to answer this way if they were highly educated or trade union members.

The analysis by Harris and Harding (1998b) is restricted to one period and so maybe influenced by that sample draw. In this section, we regress individuals' inflation expectations on personal characteristics to test whether average expected inflation from January 1995 to May 1998 varies with any of these characteristics. The results are reported in Table 2. The analysis is conducted using the sample for all respondents, and then for the two subgroups of respondents whose expectations lie between 0 and 10 per cent inclusive and those whose expectations lie outside this band. The base case is (somewhat arbitrarily) selected to be a group with low average expected inflation in this period.^[4] This happens to be males aged 25 to 34 years with a postgraduate degree working as a manager/administrator, earning $81,000–90,000 a year, and who have a home loan, vote Liberal and live in the ACT.

Average expected inflation varies substantially across different categories of respondent and does so largely in the way anticipated. Both average underlying and headline inflation were 2.3 per cent over the sample period. Since almost all groups' average inflation expectations lie above actual inflation, the lower is a particular group's expected inflation, the more accurate are their expectations on average. Consider first, the analysis based on all respondents' inflation expectations. Inflation expectations are higher for respondents who have jobs which involve less exposure to price setting, such as plant and machine operators and drivers, the unemployed or those who perform home duties.^[5] People with less education or a lower income also tend to have higher inflation expectations. People with graduate or postgraduate degrees have lower (and thus in this sample, more accurate) inflation expectations, presumably because they have had the opportunity to further develop information-processing skills.

Location also matters, with people who live in the country tending to have inflation expectations higher than those living in urban areas. People who are older, however, tend to have lower inflation expectations, which is consistent with the view that experience matters in developing information-processing skills and understanding how the economy works.

When the analysis is confined to subgroups, these outcomes largely remain intact, with only a slight modification. For those people with inflation expectations between 0 and 10 per cent, location is far less important as a distinguishing characteristic than it is for those with more extreme inflation expectations. People who live in areas which are relatively isolated from urban centres, tend to report more extreme inflation expectations. Also, people who are outside the workforce tend to report higher, less accurate, inflation expectations than most people in the workforce. It is also notable that people who have more extreme inflation expectations tend to have fewer distinguishing characteristics than those who expect inflation to lie between 0 and 10 per cent: a few people will report extreme numbers regardless of who they are.

But some of the results are difficult to explain. Even after controlling for income, occupation and education, for example, average inflation expectations vary with respondents' political preferences. It is difficult to meaningfully interpret this result. Similarly, for all three groupings of respondent, average inflation expectations are higher, and thus less accurate, for females than for males. Given that the regression controls for occupation (and implicitly, different participation rates between males and females) and education, this result is odd. Females' inflation expectations are 1.45 per cent higher, on average, than males' expectations in the full sample, 0.46 per cent higher in the 0 to 10 per cent subsample and 9.2 per cent higher in the extreme subsample. Thus, it is females with inflation expectations exceeding 10 per cent whose expectations substantially exceed those of males.

Batchelor and Jonung (1986) explain differences between male and female inflation expectations in Sweden with a survey that finds that women spend up to twice as much time food shopping as men do. They argue that women's inflation expectations are likely to be particularly influenced by food price inflation. Similar tendencies in Australia could help explain females' higher average inflation expectations. The food component of the CPI has risen by 0.8 per cent more than the total CPI, on average, over the past four years. This difference is more than sufficient to explain the gender difference for our 0 to 10 per cent sample, though it clearly cannot explain the results for the ‘outlier’ sample. In principle, the inflation expectations of other groups may also differ from the average because these groups also focus on the price changes of particular components of the CPI.

The analysis above provides a perspective on people's average inflation expectations over the past four years. Further insight can be gained by looking at the distribution of expectations more generally and how it has evolved in the past four years. While there are many characteristics which have a bearing on people's average inflation expectations, we focus here on the responses of different occupation groupings. Table 3 sets out measures of central tendency and dispersion for 10 occupation categories from January 1995 to December 1998. Detailed occupation groupings are not available before this period.

Managers/administrators and professionals have the lowest average means, medians, modes and standard deviations over this period, whereas people who are unemployed or engaged in home duties have the highest averages of these statistics. Not only are managers/administrators' and professionals' inflation expectations lower, and hence, closer to actual inflation than for other groups, but their lower within-group variability suggests that this group may have more homogenous views about how the economy works.

Figure 4 provides a view on how these statistics have evolved over the past four years for professionals, labourers and the unemployed. These three categories capture three broad types of respondent. The top panel of Figure 4 shows median inflation expectations. The expectations of professionals have been consistently lower than those of labourers and the unemployed and they responded more quickly to the easing in inflationary pressure in 1996.

The middle panel of Figure 4 shows the modal inflation expectation for these three groups. The mode for the unemployed has remained stable at 5 per cent, with the exception of October and November 1998 when it rose to 10 per cent. In contrast, since September 1996, professionals' modal expectations have been much less stable, oscillating between 0 and 5 per cent, reflecting the fall in inflation over 1996 and 1997. For professionals at least, the distribution of inflation expectations is fluid and reacts to the current inflation environment. The bottom panel of Figure 4 shows that there is generally greater similarity in the inflation expectations of professionals than in other groups.

In August, October and November 1997, and in February and May 1998, respondents were also asked about how their rates of pay are determined, whether by individual contract, enterprise agreement, award, or some other mechanism. It might be expected that people whose pay is determined according to an individual contract have more of an incentive to be aware of inflationary circumstances than those who have no direct input into how their wage is set. While the results are not reported, respondents' inflation expectations do not differ according to the way their wage is determined (in a regression which controls for occupation). This is the case for all respondents and for the sub-set of respondents whose inflation expectations lie between 0 and 10 per cent.

Footnotes

By state or territory, excluding the Northern Territory, and whether urban or rural (although this distinction is not made for the ACT). [3]

There are sufficiently large numbers of respondents in each category to give reliable regression results. [4]

The question asked does not distinguish between the unemployed and those who perform home duties. From here on, we use the term ‘unemployed’ to refer to both these groups. [5]