A new study has linked heavy air pollution from coal burning to shorter lives in northern China.

Researchers estimate the half-billion people alive there in the 1990s will live an average of 5.5 years less than their southern counterparts as they breathed dirtier air.

China itself made the comparison possible – for decades, a now-discontinued government policy provided free coal for heating, but only in the colder north.

Researchers found significant differences in both particle pollution of the air and life expectancy in the two regions, and said the results could be used to extrapolate the effects of such pollution on lifespans elsewhere in the world.

The study by researchers from China, Israel and the US was published in the Proceedings of the National Academy of Sciences.

While previous studies have found pollution affects human health, "the deeper and ultimately more important question is the impact on life expectancy," said one of the authors, Michael Greenstone, a professor of environmental economics at Massachusetts Institute of Technology, US.

"This study provides a unique setting to answer the life expectancy question because the (heating) policy dramatically alters pollution concentrations for people who appear to be of otherwise identical health.

"Further, due to the low rates of migration in China in this period, we can know people's exposure over long time periods," he said.

The policy gave free coal for fuel boilers to heat homes and offices to cities north of the Huai River, which divides China into north and south.

It was in effect for much of the 1950-1980 period of central planning, and, though ended after 1980, it has left a legacy in the north of heavy coal burning, which releases particulate pollutants into the air that can harm human health.

Researchers found no other government policies that treated China's north differently from the south.

The researchers found in the north, the concentration of particulates was 55% higher than in the south, and life expectancies were 5.5 years lower on average.