Consistent and statistically significant reductions in mortality rates over time were observed in all socio-economic groups in New Zealand, while similar reductions in Sweden were only observed among men with low income and women with high income. Regarding absolute inequalities in mortality on average over the 1980s and 1990s, they were similar between Sweden and New Zealand by education for both men and women (with the exception of greater inequalities among men in Sweden in the 1980s). Absolute inequalities in mortality by income were greater in Sweden, although this is almost certainly due to better income measurement in Sweden. Regarding trends in absolute inequalities, there was a strong decreasing trend for men in Sweden (66% by education and 51% by income). For both men and women in New Zealand, and women in Sweden, there were approximately 12% to 21% increases in inequality by income and 8% to 19% decreases in inequality by education. Trends were clearly most favourable for men in Sweden, and possibly also more favourable for males in New Zealand.
The results presented in the present study should be interpreted with awareness of potential limitations. First the New Zealand data base was larger than that of Sweden which makes it difficult to draw conclusions in the presence of wide confidence intervals. However, many of the statistical tests of trend were significant. In addition, in spite of relatively smaller population for Sweden, ULF is a random sample representative of the Swedish population. In addition Swedish mortality rates per 100,000 by socio-economic position were comparable to the national rates (with reservations for varying age groups and standardisation methods) .
Second, sources of income data varied in the two countries. Due to the taxation system, Sweden has better income measurement than New Zealand, such that the inequalities by income in Sweden probably shift up relative to New Zealand. Thus some, if not all, of the mortality differentials by income may almost certainly be due to methodological aspects. It seems likely that mortality disparities by income in Sweden are 'too well captured' to be comparable to countries such as New Zealand at any one point in time. However, comparisons of trends over time are likely to be valid, so long as varying baselines are allowed for.
Furthermore, income was measured at the household level and didn't distinguish women's from men's individual income. This may make it difficult to draw conclusions on the observed gender differences. However, a study by Fritzell at al showed that health effects were similar regardless of whether household or individual income was used . There are also conceptual advantages with household income as opposed to individual income as a measure of one's ability to purchase items.
The advantage of this study is that it provided us with the opportunity to study trends in socioeconomic inequalities in mortality, using both education and income, among men and women. This is the first study we are aware of where inequalities are investigated from a gender perspective comparing Sweden with another non-European country, over a long period of time.
There are no strictly comparable published Swedish studies on trends in socioeconomic inequalities in mortality. Trends in total mortality by socioeconomic status have often been limited to younger population up to 64 years  and to specific causes of death. Gender differences in socioeconomic differentials have often been interpreted as being less among younger women than among younger men, but this is in part due to lower overall mortality rates among women than men. Because of this, absolute differences in mortality rates between low and high socio-economic groups are greater among men (but may be similar or greater in relative terms among women – as shown in the present paper.
In fact in a previous comparison of absolute mortality rates by occupational class between late 1980s and early 1990s demonstrated that mortality had decreased among men and women (aged 20 to 64 years) across all occupational classes with an exception of women with blue collar jobs . Reinterpreting previous comparisons in this light, trends in socioeconomic differentials in women's mortality are expected to be decreasing on a slower rate than those for men – as demonstrated in the present study.
Trends in absolute (and relative) inequality by education and income in New Zealand have been published before . These previous results adjusted for ethnicity (a confounder of the association of socio-economic position with mortality in New Zealand), but the trends over time in inequalities were similar to those published in this current paper.
We found trends in socioeconomic inequalities in mortality among women to be similar in Sweden and in New Zealand data. Results in the present study suggest that women have not benefited as much as men from the reduction in socioeconomic inequalities in mortality in the past 20 years, especially in Sweden. That said, men's inequalities by education in Sweden appeared to start at a very high level in the early 1980s, but decreased markedly to smaller inequalities than those for men in New Zealand and for women in Sweden, when measured by education. Whilst a statistical chance might explain men's trends in Sweden, (although tests of trend were statistically significant or nearly so) at least two substantive reasons might explain this divergence in trends between men and women in Sweden. Measures of SES for women, particularly income, have become better (relative to men) in recent years due to increased participation in the labour market, which may show more inequalities than in the past – or at least cause increasing income-related inequalities to be observed among women despite decreases among men.
Another potential explanation, but one that we do not think is likely, is the changing proportion of single women. In Sweden for example, during the 1990s there was an increase in the proportion of single parents (about 20% of adults, and of them 70% were women) . Single parenting has been associated with economic hardships  and increased mortality . Whilst both Sweden and New Zealand have welfare benefits specifically for solo parents, they are probably not sufficient to maintain the same level of equivalised household income as before any separation from an income-earning partner. However, single parents seem an unlikely driver of the results we see, for two reasons: mortality among 25–77 year olds is driven by adults older than those with dependent children; and whilst a greater portion of low income households may now be made up of single parent households, it does not necessarily follow that the mortality rate differences between low and high income thus will also increase.
We are not sure of the reasons for a possible profound declining trend in absolute gaps in men's mortality by socioeconomic position. However, the Swedish trends in ischemic heart disease (IHD) mortality, which is a major contributor to total mortality, may in part explain the observed declining trends. Rosengren et al have shown a larger decrease in cardiovascular morbidity among men than women between 1984 and 1999 . In addition Hallqvist et al demonstrated that a decline in mortality due to myocardial infarction (MI) among men in high socioeconomic position men started in the 1970s while that of men in low socioeconomic position started in early 1980s . Thus it is possible that the rapid decline in mortality due to MI occurred first in high socio-economic men (say in the 1970s), and later in lower socio-economic men (say in the 1980s and 1990s). If true, this would mean that our study of the 1980s and 1990s would have missed the rapid fall among higher socio-economic men (and consequent widening absolute gaps), and just observed the 'correction' as men in lower socio-economic position caught up. Such dynamic trends have been proposed by Victora as a result of the inverse equity hypothesis . Regardless, the dynamic nature of trends in inequalities over time is something that both scientists and policy makers must increasingly consider and try to understand.
Why might trends in absolute inequalities by education be decreasing, but by income increasing? First, we discuss above that increasing participation by women in the labour market may explain their increasing inequalities by income. Regarding declines in absolute educational inequalities, for both men and women, one possible reason is that there is a shift in western industrialised societies for income being a greater axis of stratification than education (other than education influencing later income), which may explain why educational gaps are tending to decrease and income gaps are tending to increase. Second, it is possible that education is becoming a weaker marker of socio-economic stratification particularly in Sweden due to the fact that that there are increasingly fewer people with no qualifications. However, the SII and RII methods used in the present paper deal with this problem. Third, and more simplistically, it may just be a mathematical consequence of absolute inequalities having to decrease at some point when average or background mortality rates are relentless falling (although relative inequalities may continue to widen, eg Table 4 of this paper).
Both New Zealand and Sweden have current national strategies to tackle health inequalities. New Zealand's strategy was established about 5 years ago , while Sweden has had a long history to tackle health inequalities. In fact, Sweden is the first country to endorse a unique national public health policy which was agreed on by a majority of political parties with the intention to promote good health for all [30, 31]. Based on the results of the present study, these strategies seem to (somehow) have been outstandingly successful for men. It is yet to be seen if these strategies will in the long-run also contribute to more successful reductions in socioeconomic inequalities among women's mortality.