Australia has the world’s highest rate of rooftop solar photovoltaic (PV) adoption. The most recent data from the Australian PV Institute (APVI) reveals that Australia is on the brink of a major milestone.  At the end of Q4 2024, 39.8%, or 3.76 million households, had rooftop solar installed.

Australia’s success has not just been an accidental byproduct of abundant solar resources, but deliberate policy measures designed to accelerate adoption. 

Not only does Australia arguably have the world’s simplest and lowest-cost single-touch installation process (one phone call to an installer is enough) but the Small Scale Renewable Energy Scheme (SRES) has significantly lowered up-front costs for households.

But there has been fairly frequent criticism in the past that the SRES scheme, which is a rooftop solar subsidy, has benefited the wealthy far more than those less well off. In other words, it has been claimed that the “poor” are subsidising the “rich.” Cue the outrage!

This topic has been the subject of academic research around the world. Often the results have been inconclusive, as the drivers of PV adoption are many and varied and therefore it’s difficult to identify cause and effect. 

An article in RenewEconomy in 2021 also posed this question – is there evidence that rooftop solar is being subsidised by non-solar households?

Dr Rohan Best from the Department of Economics at Macquarie University has written a number of papers on this topic and found that depending on how the available data was analysed or aggregated, and which variables were considered causative, any of the possible three relationships could be found: a positive correlation between household wealth or income and PV adoption; a negative correlation; no correlation.

Part of the problem is that available data sets rarely link household wealth or income directly with whether that same household has rooftop solar. 

The few such data sets that do exist sample only a tiny portion of the entire rooftop solar base. The alternative of using aggregated data sets is problematic, however, because any causal links can be lost in the process of aggregation.

Undeterred, I thought it would still be useful to analyse the entire APVI data set capturing all 3.76 million rooftop solar households by postcode, and look at it through the prism of median household income provided by the Australian Bureau of Statistics (ABS) for those same postcodes. I have not previously seen analysis of this kind done.

Is there a correlation between rooftop solar adoption and household income?

Anyone can download the APVI data set by postcode, as well as the ABS Median Household Income data from the 2021 Census.  I produced the bubble chart in Figure 1 by plotting the percentage of residential dwellings with rooftop solar in each postcode against the median household income for that postcode (adjusted to 2024 dollars).

This data is not filtered – every one of Australia’s 2621 postcodes in the APVI data set is represented as a single dot in this chart, with the size of the dot being proportional to the number of households in that postcode.

Figure 1 shows a widely scattered data set, but that there is a weak negative correlation which suggests that the higher the income, the less likely a household has rooftop solar. In other words, the exact opposite outcome that the SRES critics claim.

It also shows that half of Australia’s rooftop solar installations have occurred in areas where the median household income is under $1980 per week – almost exactly on the national median. The outcome is remarkably even in that regard.

Is that case closed? Not quite. One hypothesis for this outcome is that higher income households are less sensitive to the price of electricity, hence were less motivated to reduce their grid costs by adopting rooftop solar. 

Others argued that lower income households were less likely to either have the means or appetite to pay for the high up-front capital costs of rooftop solar, so there must be other explanations for this outcome.

Best’s 2022 paper “The impact of income on household solar panel uptake: exploring diverse results using Australian data” identified that aggregated data (which the APVI data is, at a postcode level) appears to show negative correlations that could be a result of aggregation hiding causative factors associated with wealth. 

He also pointed out that income is not wealth, as wealth is the accumulation of income over time, and concluded wealth is a more important factor than income. 

For example, a low-income postcode could have a number of wealthy rooftop-solar households, which might not be apparent via the median income metric, but could result in that postcode showing a relatively high rooftop solar penetration.

That possibility however is undermined by the fact that a growing number of postcodes now have more than 50% rooftop solar penetration. In fact, 663 postcodes (or 25% of all postcodes) do, and of those, 410 postcodes have a median household income under the national average. 

That means that a large number of households earning the median income or below in each of those postcodes must have installed rooftop solar. It is mathematically impossible for that to not be the case.

For example, 73% of dwellings in postcode 2834 (Lightning Ridge and Angledool, NSW) have installed rooftop solar. Even though the median household income in that postcode is only $916 per week, a minimum of 46% of households earning this median income or less must have installed rooftop solar. 

If household wealth (or income) was a significant enabler of rooftop solar adoption, or conversely, a lack of wealth (or income) was a significant barrier, then rooftop solar adoption would rapidly decelerate as it approached 50% penetration in lower income areas.  But that has not happened.

Is size everything?

Another useful piece of data in the APVI data set is the total capacity of all the rooftop PV installed in that postcode.  This can be used to calculate the average capacity per household per postcode, and the resultant bubble chart is shown in Figure 2.

Figure 2 shows a very different outcome to Figure 1. There is a relatively strong correlation between median household income and the capacity of the installed rooftop solar. Larger systems are more expensive, so does that invalidate the earlier hypothesis?

Not necessarily. The difference is not massive – the weighted average PV capacity in the top 50% is only 0.4 kW more than the bottom 50%, which is not material in terms of total cost. 

However there are a large number of higher-income areas with PV arrays averaging over 6kW, but very few lower-income areas. So it does pay to look closer at the data, and in particular, how rooftop solar adoption has changed over time in each postcode, and how this might explain the outcomes. No previous analysis I have read has included a temporal dimension.

Time might be the key

I kept APVI postcode data sets at the end of 2018, 2021 and 2024. To simplify a comparison over time, I sorted median household incomes into $200 ranges and recalculated the weighted average rooftop solar penetration in each income range. 

To check that this second-level aggregation didn’t destroy the integrity of the underlying data, I checked that the linear trend for 2024 matched the slope of the trend in Figure 1 which captures all of the data without range binning (it did). The result and trends are shown in Figure 3.

Figure 3 shows some interesting outcomes:

– Rooftop solar adoption has increased in all household income ranges over time; and

– Lower income postcodes have consistently adopted rooftop solar in higher proportions than higher income postcodes; but

– The gap between lower and higher income postcodes is closing very slowly over time – meaning higher incomes postcodes are adopting rooftop solar at a slightly faster rate than lower income postcodes.

The average capacity of a single solar panel in 2018 was around 250W, with lower cost panels around 200W each. Given the fixed costs of installation and inverters, and average daily household electricity use, it made little economic sense for any household to install anything less than a 3-4 kW solar array.

That required up to 20 solar panels to be installed, which in turn required a reasonably large roof that is not overshadowed, and a climate with enough sunshine to make it all worth it. 

That ticked all the boxes in rural and remote areas, but less so in urban areas. When coupled with early solar incentives that were very generous, including regulated and unsustainably high feed-in tariffs (FITs), it was financially possible for lower income households in regional and remote areas to adopt rooftop solar.

In contrast, in the larger cities, and especially in the inner-city areas of the capital cities, roofs are on average much smaller, limiting the number of solar panels that can be installed. 

These areas also tend to have higher than average incomes, but perhaps that is a red herring. Arguably, wealth or the lack thereof was not the driver here, but homeowners were making a pragmatic financial decision to wait until solar technology reached the point their roof could fit a 5 kW solar array or larger.

In my view, that is the most logical explanation for the result in Figure 2, especially taking into account the change in rooftop PV penetration over time in Figure 3. 

Lower income areas have smaller than average rooftop solar capacity simply because on average they were installed earlier. In fact Figure 3 shows that it has taken about five years for the highest-income areas to reach the same penetration of rooftop solar that the lowest-income areas had achieved in 2018.

As urban areas expanded, new housing estates in fringe urban areas could accommodate larger capacity solar arrays as panels up to 500W became available. For example, postcode 2765 in NSW covers the northwest fringe of Sydney from Marsden Park to Box Hill and semi-rural areas beyond. 

In 2018, this postcode had a median income of $2010 (adjusted to 2024 dollars), 5149 dwellings, a rooftop solar penetration of 21.8% and an average size of 5.4 kW. By 2024, this had increased to a median income of $2972, 19906 dwellings, rooftop solar penetration of 54.6% and an average size of 8.0 kW.

My conclusion is that higher income areas do not have larger solar installations than lower income areas solely because they are wealthier, but mostly because those systems were installed much later, meaning more kW of solar PV could fit on constrained roof sizes.

Also note that the SRES scheme and its associated Small Technology Certificate (STC) system is being phased out, so solar subsidies are getting smaller over time. 

The subsidy for an 8 kW system installed in 2024 is about the same as the subsidy a 4 kW system would have received in 2018. So even on that measure, wealthier areas installing larger solar arrays later did not receive larger subsidies than poorer areas installing smaller solar arrays earlier.

It should also be recognised that inner city areas will hit a natural ceiling of rooftop solar penetration at some point. Many roofs suffer high levels of overshadowing due to immediately adjacent buildings, or have only one roof plane facing the wrong direction, or are in heritage areas which restrict solar PV installations. 

Also, there are a large number of apartment buildings, few of which have rooftop solar due to the complexity of determining how to fairly share the electricity it generates among unitholders.

Conclusion

Overall, I think the SRES scheme has been an incredibly successful piece of social policy as well as an incredibly successful energy policy, reducing grid costs for millions of Australians while increasing our production of clean energy. 

There appears to be little statistical evidence to support the claim that this policy has resulted in the ‘poor’ subsidising the ‘rich’. The evidence shows, with a high degree of confidence, that lower than average income areas have benefited from solar subsidies just as much as, if not slightly more than, higher than average income areas.

Dr Brendan Jones is an electrical engineer who has had a long interest in renewable energy and electric vehicles, as well as being a data nerd who enjoys statistical data analysis