There's been enormous coverage of the new Centrelink debt letters process, whereby the Department of Human Services has automated the process of matching data from the ATO and Centrelink to try to find overpayments (but not underpayments) in welfare benefits to Australia.
The automation has involved removing human quality assurance steps, which has led to the number of debt checking letters growing from 20,000 per year to 20,000 per week. Over 260,000 of these letters have been sent out to-date.
Now automated data-matching can be a fantastic thing when used well. It can reduce duplication in identification processes, find patterns and trends that inform polices and service delivery, and even identify inaccurate payments - as was the intention with this approach.
However for a data-matching process to work well, the system rules need to be well-designed and tested, and the data needs to be comparable so as to be matchable.
The widespread issues that are being reported by former and current Centrelink payment recipients and the enormous (more than 350 articles over the last month) of media coverage suggest that the system that Centrelink put in place meet neither of these conditions.
Without going into the apparent system issues, which have been covered widely, the reason Centrelink and other agencies introduce automated systems is to allow them to achieve the same, or better, outcomes with fewer staff - something that in economic terms lifts the productivity of the agency.
Centrelink has reduced its workforce by around 5,000 staff over the last five years, and moved to have a larger casual workforce with fewer permanent staff. These types of changes are occurring across many large public sector organisations as governments tighten their belts.
However it seems that when governments today act to (hopefully) bring about productivity gains and cost cuts, they focus primarily on a subset of the economy, the public service.
They appear to overlook the potential impacts on other sectors, or the overall productivity or cost dividend to the community they serve.
Let's use the Centrelink situation as an example. By cutting a human quality assurance step, and using a purely automated approach for identifying potential overpayments, Centrelink has transferred the cost of checking that their data is accurate from internal staff to welfare recipients.
Now while it may be appropriate for the people receiving the payments to be responsible for justifying why they receive them, there is a significant productivity cost when passing the task of quality assuring claims from trained and experienced staff with strong systems supporting them to low income, sometimes low education, citizens.
This productivity cost is exacerbated when these citizens are expected to re-prove their eligibility for past welfare payments, dating back as far as six years. The citizens now must track down former employers, landlords and education providers to source the materials that Centrelink has decided it now requires (often for a second or third time) to revalidate past payments.
So not only are individual citizens required to spend significant time checking the documentation Centrelink holds on them (which is subject to errors from mistaken entry by Centrelink staff and algorithmic mistakes, such as averaging a citizen's part-time or sporadic pay over 26 fortnightly periods), but must also involve the time of a range of former employers and landlords.
Now I've been going through an employment process (you'll hear more about this shortly), which required me to source a range of documents from 4-6 years ago from former employers - both public and private sector. While not the same as the process welfare recipients are facing, it required similar information such as old payslips and employment dates.
All the organisations were very prompt in responding (thanks to you all), taking no more than two weeks to pull together what was required - however I estimate that the combined time they spent on this one matter for me exceeded ten hours work time, just for one person in one clearcut situation.
For the 260,000 welfare recipients who may have to re-source material from employers and others, this time adds up to potentially millions of hours of lost productivity for the companies involved - that's outside the time spent by the welfare recipients themselves to 'prove' they were not overpaid, or not as much as Centrelink claims they were.
The Minister for Human Services suggested before Christmas that around $300 million in debt had been recovered, this was later revised to being debt identified, with neither the Minister nor Department able to conclusively say how much had actually been paid to the Commonwealth.
Now let's look at a few estimates.
Welfare recipients are reporting that they are spending up to dozens of hours resolving this matter with Centrelink. That includes pulling together documents, dealing with former employers. waiting on the phone with Centrelink for up to four hours, with multiple call-backs for dropped lines, managing difficulties with MyGov and other associated activities.
For past recipients (who may have spent a few months or years on welfare when studying or during gaps in employment) who are currently employed, this can require non-productive time in business hours while at work (the only time Centrelink takes calls) and cutting into other job-related or educational activities outside hours.
In addition their former employers and educational institutions are spending hours pulling up old payslips from archives - noting that where employers have shut-down this becomes even more difficult and time consuming.
If we assume that the average welfare recipient is spending 6 hours on dealing with their Centrelink debt and that former employers are spending another 4 hours servicing their requests to meet DHS requirements, that's 10 hours productivity lost per debt notice.
Now if we assume that 60% of debt notices require this time investment, based on 260,000 notices issued, that's 156,000 debt notices on which people are spending 10 hours each on resolving - whether or not there is an overpayment at the end of the process.
Let's take a hourly rate of $30 - low for Australia - as the dollar cost of those hours. Based on 156,000 notices at 10 hours effort (1,560,000 hours effort total), at that dollar rate the cost to the economy is $46.8 million dollars.
Now that's the direct productivity cost to citizens and businesses. On top of that there's been extensive involvement by not-for-profit legal and counselling services dealing with an upsurge in complaints and counselling needs and the mental and physical distress people facing large unexpected Centrelink debt notices are currently facing, harming their ongoing productivity and effectiveness.
There's also time spent by citizens on social media engagement, the creation and management of the NotMyDebt website and, finally, the time being spent by Centrelink's own staff sorting out debt issues which could have been easily screened out through a QA process.
I'd estimate from the above that the net productivity cost to Australia of saving Centrelink's QA step is already approaching about $80 million, without considering the longer-term cost of the loss of credibility and potential impact that will have on future productivity.
While the Commonwealth may be able to claw back this amount of money via the debt notices, I think that the situation is already well past the point where the situation has a net productivity loss to Australia as a society.
In other words, the cost of reclaiming this debt in this manner, is significantly outweighed by the overall productivity cost to the country. Sure it might make a good political statement for a 'no-nonsense' approach to welfare (though this appears challenged by poll results), but the economic cost makes the approach very hard to justify from a pragmatic perspective.
Government is likely to face more and more of these types of situations as it attempts to lift productivity and/or cut costs by transferring the work done by staff back onto citizens.
While I understand the importance of cost management in government, and the ongoing desire to lift productivity, looking at these metrics based on public sector inputs rather than society-wide outputs does risk governments making decisions that harm economy-wide productivity in the long-term when chasing short-term productivity gains for a specific government agency.
Governments who wish to see long-term economic gains need to carefully consider how they shift effort from experienced staff to inexperienced citizens in order to not increase burdens that reduce overall productivity and wipe out the public sector savings through lower tax receipts or large pushback costs.
Digital transformation is a key tool in this process, but must be used wisely, not simply to automate steps to remove humans, but to simultaneously cut errors and improve success-rates.
The current Centrelink debt issue is a clear example of what happens when a good automation idea is executed poorly, becoming an overall loss to government rather than a win.
The automation has involved removing human quality assurance steps, which has led to the number of debt checking letters growing from 20,000 per year to 20,000 per week. Over 260,000 of these letters have been sent out to-date.
Now automated data-matching can be a fantastic thing when used well. It can reduce duplication in identification processes, find patterns and trends that inform polices and service delivery, and even identify inaccurate payments - as was the intention with this approach.
However for a data-matching process to work well, the system rules need to be well-designed and tested, and the data needs to be comparable so as to be matchable.
The widespread issues that are being reported by former and current Centrelink payment recipients and the enormous (more than 350 articles over the last month) of media coverage suggest that the system that Centrelink put in place meet neither of these conditions.
Without going into the apparent system issues, which have been covered widely, the reason Centrelink and other agencies introduce automated systems is to allow them to achieve the same, or better, outcomes with fewer staff - something that in economic terms lifts the productivity of the agency.
Centrelink has reduced its workforce by around 5,000 staff over the last five years, and moved to have a larger casual workforce with fewer permanent staff. These types of changes are occurring across many large public sector organisations as governments tighten their belts.
However it seems that when governments today act to (hopefully) bring about productivity gains and cost cuts, they focus primarily on a subset of the economy, the public service.
They appear to overlook the potential impacts on other sectors, or the overall productivity or cost dividend to the community they serve.
Let's use the Centrelink situation as an example. By cutting a human quality assurance step, and using a purely automated approach for identifying potential overpayments, Centrelink has transferred the cost of checking that their data is accurate from internal staff to welfare recipients.
Now while it may be appropriate for the people receiving the payments to be responsible for justifying why they receive them, there is a significant productivity cost when passing the task of quality assuring claims from trained and experienced staff with strong systems supporting them to low income, sometimes low education, citizens.
This productivity cost is exacerbated when these citizens are expected to re-prove their eligibility for past welfare payments, dating back as far as six years. The citizens now must track down former employers, landlords and education providers to source the materials that Centrelink has decided it now requires (often for a second or third time) to revalidate past payments.
So not only are individual citizens required to spend significant time checking the documentation Centrelink holds on them (which is subject to errors from mistaken entry by Centrelink staff and algorithmic mistakes, such as averaging a citizen's part-time or sporadic pay over 26 fortnightly periods), but must also involve the time of a range of former employers and landlords.
Now I've been going through an employment process (you'll hear more about this shortly), which required me to source a range of documents from 4-6 years ago from former employers - both public and private sector. While not the same as the process welfare recipients are facing, it required similar information such as old payslips and employment dates.
All the organisations were very prompt in responding (thanks to you all), taking no more than two weeks to pull together what was required - however I estimate that the combined time they spent on this one matter for me exceeded ten hours work time, just for one person in one clearcut situation.
For the 260,000 welfare recipients who may have to re-source material from employers and others, this time adds up to potentially millions of hours of lost productivity for the companies involved - that's outside the time spent by the welfare recipients themselves to 'prove' they were not overpaid, or not as much as Centrelink claims they were.
The Minister for Human Services suggested before Christmas that around $300 million in debt had been recovered, this was later revised to being debt identified, with neither the Minister nor Department able to conclusively say how much had actually been paid to the Commonwealth.
Now let's look at a few estimates.
Welfare recipients are reporting that they are spending up to dozens of hours resolving this matter with Centrelink. That includes pulling together documents, dealing with former employers. waiting on the phone with Centrelink for up to four hours, with multiple call-backs for dropped lines, managing difficulties with MyGov and other associated activities.
For past recipients (who may have spent a few months or years on welfare when studying or during gaps in employment) who are currently employed, this can require non-productive time in business hours while at work (the only time Centrelink takes calls) and cutting into other job-related or educational activities outside hours.
In addition their former employers and educational institutions are spending hours pulling up old payslips from archives - noting that where employers have shut-down this becomes even more difficult and time consuming.
If we assume that the average welfare recipient is spending 6 hours on dealing with their Centrelink debt and that former employers are spending another 4 hours servicing their requests to meet DHS requirements, that's 10 hours productivity lost per debt notice.
Now if we assume that 60% of debt notices require this time investment, based on 260,000 notices issued, that's 156,000 debt notices on which people are spending 10 hours each on resolving - whether or not there is an overpayment at the end of the process.
Let's take a hourly rate of $30 - low for Australia - as the dollar cost of those hours. Based on 156,000 notices at 10 hours effort (1,560,000 hours effort total), at that dollar rate the cost to the economy is $46.8 million dollars.
Now that's the direct productivity cost to citizens and businesses. On top of that there's been extensive involvement by not-for-profit legal and counselling services dealing with an upsurge in complaints and counselling needs and the mental and physical distress people facing large unexpected Centrelink debt notices are currently facing, harming their ongoing productivity and effectiveness.
There's also time spent by citizens on social media engagement, the creation and management of the NotMyDebt website and, finally, the time being spent by Centrelink's own staff sorting out debt issues which could have been easily screened out through a QA process.
I'd estimate from the above that the net productivity cost to Australia of saving Centrelink's QA step is already approaching about $80 million, without considering the longer-term cost of the loss of credibility and potential impact that will have on future productivity.
While the Commonwealth may be able to claw back this amount of money via the debt notices, I think that the situation is already well past the point where the situation has a net productivity loss to Australia as a society.
In other words, the cost of reclaiming this debt in this manner, is significantly outweighed by the overall productivity cost to the country. Sure it might make a good political statement for a 'no-nonsense' approach to welfare (though this appears challenged by poll results), but the economic cost makes the approach very hard to justify from a pragmatic perspective.
Government is likely to face more and more of these types of situations as it attempts to lift productivity and/or cut costs by transferring the work done by staff back onto citizens.
While I understand the importance of cost management in government, and the ongoing desire to lift productivity, looking at these metrics based on public sector inputs rather than society-wide outputs does risk governments making decisions that harm economy-wide productivity in the long-term when chasing short-term productivity gains for a specific government agency.
Governments who wish to see long-term economic gains need to carefully consider how they shift effort from experienced staff to inexperienced citizens in order to not increase burdens that reduce overall productivity and wipe out the public sector savings through lower tax receipts or large pushback costs.
Digital transformation is a key tool in this process, but must be used wisely, not simply to automate steps to remove humans, but to simultaneously cut errors and improve success-rates.
The current Centrelink debt issue is a clear example of what happens when a good automation idea is executed poorly, becoming an overall loss to government rather than a win.