As with age, mobility also varies according to the geographic, socioeconomic and demographic characteristics of individuals and households. One limiting factor of the NATSISS in regard to establishing the geography of mobility and its demographic impact on sending and receiving regions is the lack of any origin/destination data. Instead, we have measures of the propensity to move (or not) at fairly large geographic units, notably States and the Northern Territory and the broad remoteness categories of the ASGC. For the most part, no difference is evident between jurisdictions in the propensity to move, given the spread of the upper and lower bounds of most estimates (see Fig. 5.2). However, stand-out exceptions include Tasmania and the Northern Territory, with the rate of movement in the latter ranging from as low as 15.7 per cent to no more than 20.9 per cent. In effect, the propensity to move among Indigenous people in the Northern Territory is only half the rate in Queensland. As for movement propensity by remoteness category, this reveals that inner regional areas display significantly higher mobility than all other regions except for major cities, while at the other extreme of the remoteness classification, the population in very remote areas displays significantly lower movement than all other regions.

These observations regarding the spatial pattern of movement are interesting for their similarity to consistent findings from census data that display the lowest rates of Indigenous mobility in remote areas of the north and west of the continent, and especially across the Northern Territory (Taylor & Bell 1996b). This impression of relative immobility among Indigenous people in very remote areas and across the Northern Territory is grossly misleading. Numerous case studies attest to the importance of frequent mobility in the daily, periodic and seasonal round of activities associated with Indigenous social and economic life in such places (see Taylor & Bell 2004 for a summary). The gap between these observations and NATSISS results is all the more surprising given the more inclusive nature of the NATSISS mobility question allowing for all moves to be recorded. This differs from the census question which is ill-suited to capturing the frequent short-term circular movements that are common in remote Australia.

While explanation for this anomaly is most likely to be found in respondent error associated with the nature of the NATSISS question and the manner in which it was interpreted, even this is perplexing because previous critiques of census data have pointed to the inappropriateness of applying ‘usual residence’ concepts for measuring movement among populations that live as much in an area as a single place (Morphy 2002: 44–55). However, the NATSISS does not refer to usual residence, preferring instead to talk of the number of dwelling/houses/places ‘lived in’ which would seem to be a less ambiguous construct—although it does raise the issue of what ‘lived in’ actually means.

Along with differentials in spatial mobility, the reported data also show clear differences in the propensity to move across a range of social, economic and cultural characteristics that are selected here for their likely association with movement propensity. As indicated by the age-standardised point estimates for each of these variables in Table 5.1, and by the associated 95 per cent confidence interval, these differences tend to reduce for some characteristics and increase for others once the differences in age composition between the categories are controlled for. In some instances, the standardisation procedure even alters the relative balance of variables. For example, the unadjusted migration rate among married people was lower than for those not married, but higher for the standardised result. Overall, the results reveal that movement propensities vary widely around the national average of 30.9 per cent according to the different variables.

Table 5.1. (continued)

Source: Customised cross-tabulations from the 2002 NATSISS RADL

Although this bi-variate analysis is highly suggestive, the net effects of each of these independent variables can only be assessed by a multi-variate analysis. For this purpose, we fit a logistic regression with the dependent variable taking the form of 1 if the respondent lived in any other dwellings (or places) in the 12-month period before the survey, and 0 otherwise (see Table 5.2). In this way, the results indicate the effects of all the selected factors simultaneously on the chances of moving or not. The simplest way of summarising this relationship is to examine what happens to these chances for groups of Indigenous people with different characteristics. These changes in movement probability are best measured relative to a hypothetical reference person and the characteristics of the reference person chosen for this purpose are indicated in the note for Table 5.2.

a. Reference person is defined as: male, married, employed in the public sector, has Year 10 education, lives in public rental housing, does not live on homeland, neighbourhood has problems, and resident in an outer regional location. The results are not sensitive to the choice of the characteristics of the reference person.

b. Statistically significant at the 5% level.

Source: 2002 NATSISS RADL

These results indicate that marginal labour force status is the biggest predictor of mobility. After removing the effects of other variables, there is a significant underlying pattern of higher mobility among those unemployed or not in the labour force compared to those employed. When this is controlled for by sector of employment, the chances of movement are seen to be greater among private sector and CDEP workers compared to public sector workers. As for educational attainment, a gradient of increased mobility with higher educational attainment is evident. This is strengthened by the related result that non-attendance in a vocational training course has a negative effect on mobility.

Housing tenure is another key variable with much higher chances of movement among those in private rental dwellings, while home ownership and community rental serve to depress mobility. Contrary to expectation, respondents resident in neighbourhoods with perceived social problems are far less likely to move, thereby negating what might have been viewed as a push factor and probably reflecting limitations on residential choice. Other findings related to location appear contradictory. For example, residence on homeland is negatively associated with mobility, while remote and very remote location increases the chances of movement. One possibility is that the latter observation reflects the amalgamation of responses from both non-remote and remote non-community sample forms, whereas the former is based on remote community sample data. Finally, self-assessed health status is interesting, as this deflates mobility at extremes of the range, though no doubt for quite different reasons—one due to choice, and the other due to incapacity.