When COVID-19 lockdowns came, Canada’s employment insurance system failed to cover enough of the Canadian workforce that was affected and failed administratively to keep up with the surge in applications for benefits. This new report by Jennifer Robson proposes ideas for policy reform and shows why Canadians need wage insurance in the first place.

In mid-March, as millions of Canadians were asked to stop going to work and stay home to prevent the spread of the COVID-19 virus, the employment insurance (EI) system failed.

A parallel could be drawn with the example of American Insurance Group during the 2008 financial crisis — a large, complex and opaque operation that worked well enough, but was never designed to cope with the potential of so many claims arriving at once.[1]

EI failed in at least two important ways. First, it failed to cover enough of the Canadian workforce that experienced sudden furlough, lay-offs or a significant drop in paid hours. By April, a little more than one third of all workers in Canada were unemployed, including those not looking for work during the shutdowns, and those who were employed, but had lost all or the majority of their paid hours.[2] Second, it failed administratively to keep up with a sudden surge in applications for benefits. As the minister responsible for the program later testified, best estimates on processing times for the millions of claims received ran as long as 18 months.[3]

Each of these is reviewed in more detail and directions for policy reform that could avoid the same catastrophic failure in case of a similar emergency are proposed. But, more important, these same changes might better equip the system to respond to the myriad personal and micro-economic emergencies that happen every day and are the policy reasons for why working Canadians need wage insurance in the first place.

EI is Canada’s primary social-insurance program for working-age adults. It was originally launched in 1940 as a contributory insurance program for what policy-makers seem to have conceived of as blue-collar workers[4] on and off the job with the same employer.[5] The program’s roots date even further back, to the 1919 Royal Commission on Industrial Relations (the Mathers Commission),[6] which recommended a federal program of “State Social Insurance for those who, through no fault of their own, are unable to work . . . . Such insurance would remove the spectre of fear which now haunts the wage earner.”

That report[7] also offered a recognition that some workers, in non-standard jobs, were especially at risk of unpredictable losses to their income and being left “without the means of subsistence.”

Social insurance is meant to provide protection against social risks, that is risks whose full costs ripple beyond those directly affected. There is no real incentive for private markets to offer insurance products for such risks, but there is a strong efficiency case for governments to create programs that pool resources for management and mitigation strategies.

In normal times, roughly 80 percent of the Canadian workforce is notionally covered by the program — meaning they are paying insurance premiums and accumulating insurable hours of employment that could be used to qualify for regular benefits in cases of involuntary job loss, or special benefits in cases of temporary unpaid leaves from paid work for illness or caregiving.[8] Before COVID hit, there were approximately 1.1 million Canadians who were unemployed,[9] but among those, only 460,000 (42 percent) received regular EI benefits.[10]

The fact that only a minority of unemployed Canadians usually receives EI is neither new nor newsworthy. Policymakers have known for a long time that an important share of the workforce is paying into an insurance pool from which it cannot ever draw a benefit.[11]

Previous research looking at special EI benefits paid to new parents finds a very similar pattern of a share of workers contributing into the common social fund and unable to qualify for help when they need it.[12] In 2014, one in five new mothers had been working, but were not able to collect maternity or parental benefits and one in 10 had been paying EI premiums, but didn’t get maternity benefits. This was still true in 2018 when nine percent of new mothers had paid premiums, but didn’t get benefits.[13] What’s worse is that access to maternity benefits rises with family income so that low-income families who have paid into the system are less likely to see any benefits when they need them.

A basic principle of social insurance is a wide pooling of resources to fund assistance (either directly through contributions or indirectly through general revenues) tempered by specific criteria for drawing benefits. But it is difficult to imagine that massive economic shutdowns and layoffs, as we’ve seen with this once-in-a-century pandemic, would not meet the criteria for benefits in any well-functioning social-insurance system. And yet, of the 8.2 million Canadians who have received assistance from the Canada Emergency Response Benefit (CERB), the cash benefit put in place in lieu of employment insurance, half appear, on preliminary screening, to not have been covered by EI.[14] In the best of times, with record low unemployment, the EI system covered 80 percent of the workforce, but paid benefits to less than half of workers experiencing unemployment. At the worst of times, during COVID, a policy response based on EI coverage alone would have left out a significant share of Canadians needing help in a crisis.

Not only did CERB have to dramatically increase the share of workers who were covered, but it also had to take a significantly more flexible approach to defining the eligible circumstances under which a worker would be paid benefits. The COVID-related shutdowns and ripple effects have led to absences due to self-isolation or illness, temporary furloughs, permanent layoffs, reductions in paid hours of work, voluntary unpaid leaves of absence from work, reduced opportunities to seek work and reduced opportunities to earn self-employment income.

Under regular EI rules, many of these types of employment separations or reductions would be ineligible for special or regular benefits even if an applicant had contributed to the insurance fund and met the minimum insurable hours to qualify. In fact, pre-COVID, among those Canadians who contributed premiums to EI and became unemployed, 25 percent were found to have had reasons for becoming unemployed that were not valid for receiving benefits.[15]

In proving the validity of their claim, a worker may be at the mercy of a former employer and the judgment of a claims process that is difficult to scrutinize.[16] When an employee ceases working, employers are required to submit a record of employment to the EI program, a form that summarizes a worker’s recent earnings and employment history as well as the reason the employment was terminated. Employers and employees may disagree on the facts of why employment ended. In these cases, the financial risk is on the employee who no longer has employment income and now faces a significant hurdle in accessing insurance benefits for which they have contributed. There is not, in fact, always a bright line between “quit” and “fired.”

In principle, the EI system recognizes that even quitting may be out of a worker’s control. For example, EI regular benefits can be paid in cases when an employee is no longer able to work because they have lost their childcare.[17] An employer who uses code “E10 — Quit /Care for a dependent” may reduce some of the administrative obstacles facing a former employee forced to stop working when their childcare fell through, but an employer who uses the standard “E00 — Quit” may send that same former employee into the EI wilderness.[18] What’s more, claimants are expected to prove they took extraordinary steps to preserve their employment, such as seeking changes to their work schedule, considering an unpaid leave until a solution could be found, or even asking friends and family to provide childcare. Readers will immediately recognize the near impossibility of proving such efforts in the current circumstances. COVID-19 has meant mass shutdowns to child care and public-education systems, and uncertainty about capacity or sustainability of re-opening efforts to those same services. Constraints in access to childcare or interruptions in public education services are not entirely new phenomena in 2020. There’s also everything from ice storms to labour interruptions.

Either way, the EI system has only truly recognized unpaid caregiving as a valid reason for job separation in the first 12 or 18 months of a child’s life or in the awful circumstances of a life-threatening illness in a child.[19] Other systems of social insurance have been able to find far more flexible and responsive ways to provide support to parents and caregivers of infants, toddlers and even school-aged children who must take time away from paid work.[20] Even if amendments were made to cover more workers and to make it easier to meet the labour force attachment test imposed by EI, many Canadians who lost paid work would not be eligible for benefits without a more significant re-definition of the kinds of social risks that EI is legally able and intended to cover.

I’ve written elsewhere[21] that the federal response on income support during the spring of 2020 is not a story of sudden policy innovation to replace traditional EI with a new benefit. Instead, it is a story of incremental policy changes that tried first, and modestly, to make EI better fit the new circumstances. When it became clear, because of the sheer volume of claims, that the EI processing system would be paralyzed, CERB was introduced as a temporary fix — a way to get income support to the millions of Canadians in need by dramatically reducing and redistributing the administrative load on the back-end IT systems that run EI.

The majority (73 percent) of the 2.8 million new or renewed claims handled by the EI system in a normal year run through a fully or partially automated system.[22] This is made possible by having the overwhelming majority (99 percent) of EI claimants file their claims online. When the information is submitted online, and particularly, when it can be linked to other administrative data — such as an electronic record of employment, the social-insurance number registry or even tax records — government systems can be designed to more quickly complete various tasks in determining eligibility and benefit levels. When program parameters are stable and claim volume is modest, ESDC reports that the average claim takes 19 days to process.[23] But this figure includes only those workers covered by EI, who have an eligible reason to claim benefits and meet the program eligibility test of recent labour-market attachment. Any policy reform that increases the eligible population and the range of reasons that benefits can be claimed will necessarily increase overall claim volume and, accordingly, pressures on the administrative back-end system.

And that system is built on a 60-year-old programming language. The fitness and ability of EI’s administrative and technological back end is an important question that seems to have been almost completely overlooked in past reports and analyses. Are they still fit for purpose and can they adequately respond to new demands in volume?

A 2016 evaluation of the program concluded that “the absence of sufficient investments in technology to support EI processing, combined with reduced budgets and an increasingly complex operating environment limited the department’s ability to fully realize the potential benefits of the automation and modernization agenda.[24] The same evaluation noted that, while 19 days might be the average wait for a first benefit payment, there has been a recent increase in the number of claims waiting 28 days or longer.

The system has seen cost savings from back-end investments to streamline or automate application review and benefit calculation. According to one estimate, Ottawa’s mean administrative cost per claim fell 10 percent (from $66.60 to $59.88) between 2008-09 and 2011-12.[25]

The scale of the challenge, and the potential opportunity for modernization, is illustrated by the fact that EI runs on COBOL, a programming language first developed in 1959.[26] It is robust, but very old. Complex changes to program parameters in response to policy changes can take considerable time and COBOL programming is specialized and increasingly rare knowledge.[27]

During the spring of 2020, the EI system faced a double imperative. It had to make payments to millions more Canadians than usually submit a claim in an entire year, and second, to deliver that money quickly. There was no existing system or database to help the government proactively identify individual Canadians and send them a payment, even if the principles of social insurance were relaxed in favour of a universal lump-sum amount.[28] Canada is not alone in lacking detailed and current information about its own population. The United States did aim to make a universal and flat payment to all American adults by using 2018 personal income tax records. But that approach resulted in other kinds of errors and gaps in coverage. According to the U.S. Government Accountability Office, nearly (US)$1.4 billion in relief cheques has been paid to deceased persons.[29] Furthermore, in early June, the House Ways and Means Committee reported that as many as 35 million living Americans were still waiting for their universal relief payments.[30] This illustrates the imperative of having valid and reliable administrative records to inform benefits programs’ design and implementation, even when the program has the broadest possible eligibility criteria.

In Canada, the government’s solution to the task of paying a lot of people money and quickly, was to drastically simplify the usual EI program rules. All claimants would receive the same amount of money and no adjustments to benefit amounts would be made within a claim period in response to a claimant’s other sources of income.[31] For most EI claimants, the calculation of payment amounts can usually be handled by an automated stage in the claims process,[32] but this presumes a modest volume of claims and limited variation in the benefit amounts to be paid. Furthermore, the government needed to distribute the volume of claims so that an important share could be handled and processed by the Canada Revenue Agency, rather than being managed solely within the outdated EI system itself. On the simple metric of capacity to pay many millions of Canadians and quickly, the EI system, as it operates during normal times, failed during COVID.

The coronavirus has created massive disruptions to nearly every aspect of Canadian life — whether and how we work for pay, the distribution of care for children and other dependent family members, the risk of illness and our collective obligation to avoid spreading infection, even if it means forgoing income. The risks of interrupted work income have become much greater for a wide range of reasons, many not well-recognized under our main social-insurance system for working-age adults. None of these social risks is actually unique to or the sole result of COVID. These were risks borne by many Canadian workers and their families before COVID, risks that will continue after it is finally gone. But these risks have become suddenly and simultaneously more visible among a larger share of the population.

I am going to avoid the thorny questions surrounding the future of the Canada labour market, of work in general and economy. I do not know whether COVID is going to lead to an increase in precarious employment, or if labour shortages will increase worker bargaining power. I would prefer not to speculate. Instead, I’ll conclude with a few ideas, rooted in this paper’s analysis earlier analysis about the EI system we might have wanted to have heading into March 2020.[33] This may inform, as we start to understand the “new normal” in our labour market and economy more generally, the debate about future directions for income support and social insurance in Canada.

In March 2020, we should have had a social-insurance system that has maximum automation and limited administrative burden on governments, applicants and employers.

  • Rather than requiring complex applications to be verified before benefits are claimed (a “verify-then-trust” approach), we should have a system that can issue a payment for a wide range of risks to employment income without requiring cumbersome and time-consuming evidence of personal circumstances such as daycare closures or COVID symptoms pending diagnosis. CERB has, so far, demonstrated that a “trust-but-verify” approach can be used to provide benefits rapidly and flexibly.
  • Using personal income-tax data, the federal government has an unparalleled ability to verify most sources of work income (including an increasing share of self-employment income). That tax data collection could and should be improved to allow for within-year information akin to what is already reported directly to the agency on year-end tax slips such as T4s.[34]
  • The “trust-but-verify” approach used in CERB is similar in some respects to the repayable assistance proposed by Mendelson and Banting (2011),[35] who suggested that a temporary benefit be made available to those unemployed Canadians otherwise ineligible for EI, with repayment conditional on future earnings. The “trust-but-verify” version here goes a little further and suggests, for example, that parents of critically ill children or children whose daycare has closed shouldn’t have to jump through hoops in a time of personal crisis just to get a modest level of income support. Similarly, establishing that someone has become unemployed or suffered a loss of a substantial share of one’s income shouldn’t be done in a way that increases the potential for conflict between workers and employers who may offer different explanations of the same events.
  • The conditions for making “trust-but-verify” possible include a serious rethinking of the kinds of social risk that we collectively agree to cover through social insurance. There is a wide range of reasons that someone might be, as the 1919 Royal Commission wrote, unable to work through “no fault of one’s own.” We might also ask whether quitting a job, as discussed earlier, is really always a voluntary act. Instead of an opaque and complex system where benefit entitlement is very limited, would Canadian workers make better-informed choices — about bargaining with their employer, juggling demands for unpaid care, or taking time off when sick — if their access to income insurance was more certain? Income insurance might be structured more like a time bank, where balances grow to some maximum in response to labour-market participation. New entrants could be granted an asset of a reasonable minimum to build from, and individual Canadians would be able, through portals not unlike the current online application and MyAccount services, to have more agency and dignity over when and how much of their “time” benefits they use up. Credit for “time” might be individual in the case of insurances against lost work, but it could be managed differently for caregiving, allowing families to pool time and granting “credit” to a dependent (including a newborn child) instead of only to workers with specific forms of labour-market attachment. Different kinds of need could be met through a system that bundles, but is more responsive to various forms of social risks.

Automation has to be balanced with empowerment for individual citizens.

  • Another condition for making “trust-but-verify” possible is making a real investment in the back-end systems that power EI claims administration. The 2016 ESDC evaluation cited earlier should have been a warning to us all, pre-COVID, that antiquated government systems are at risk of failing.[36] That evaluation report also warned that under-investment in IT systems had limited the degree to which the EI program had moved further towards automation and reduced costs of administration. Daron Acemoglou has argued that COVID has shown that governments need to shoulder more responsibility relative to private citizens and markets, but, he cautions, governments also need to become more efficient in doing so.[37] Otherwise, he warns, we risk an expanded state that, like a super-sized U.S. Department of Motor Vehicles, is the worst combination of a powerful, inefficient, inscrutable and immovable bureaucracy. One can conclude that there is a bi-directional relationship between changes to EI that might make it a more responsive and empowering program that allows users real agency, and the necessity of investments in the government information and IT-processing capacity to make it run. We can’t move to “trust-but-verify” without better government information and IT. On the other hand, investing in better government information and IT risks becoming Acemoglou’s “DMV state” if it isn’t used to trust citizens and allow them agency in their interactions with the state.

Finally, going into March 2020, we would have wanted a more robust social-insurance system for working-age adults.

  • What we should have had was a social-insurance system that could handle sudden shocks and adapt to myriad variations in need, both of which have been discussed above. But robustness here also means a system that truly meets the objectives of social-insurance systems, which are widespread insurance coverage of the population and also widespread contributions to the common insurance fund.
  • For many years, EI has had a somewhat perverse pattern of collecting contributions from many millions of workers who may be most at risk of unemployment while providing an ever-increasing set of special benefits that are most likely to be accessed by those who are least likely to face layoffs or major income losses due to unemployment. In short, we have been subsidizing the sick leave, maternity and parental leaves of better-off Canadians, in part by continuing to collect EI premiums from precarious and lower-wage workers who are less likely to ever qualify for benefits.
  • In March 2020, we should have had a system in which all workers are covered by and paying into the insurance pool. An example worthy of note is Quebec’s parental leave program, which was carved out of the larger EI system. In that province, all workers, including self-employed persons and those in non-standard employment relationships, are eligible for benefits, but also are required to pay into the program. Under EI, self-employed workers are offered an opportunity to opt-in to EI, but for special benefits alone. Take-up has been disappointingly low.[38] Defaults matter to individual behaviour and a program structured as “opt-in” will have lower participation than one that is inclusive by design.


  1. Schich, S. (2009). “Insurance Companies and the Financial Crisis,” Financial Market Trends, Organisation for Economic Co-operation and Development.
  2. Statistics Canada (2020). Labour Force Survey, June 2020, The Daily, June 10. See infographic 7.
  3. Qualtrough, Hon, C. (2020). Testimony to the Standing Committee on Government Operations and Estimates, House of Commons, Meeting 11, May 11. 43rd Parliament, 1st Session.
  4. The original program excluded professional classes, government employees and high wage earners (Lin, 1998) who made roughly double the average annual industrial wage (Author’s calculation using Statistics Canada Series E49-59, Average weekly wages and salaries, industrial composite, by province, 1939 to 1975, Historical Statistics of Canada).
  5. Workers employed on a casual basis were not covered under the 1940 legislation (Lin, 1998).
  6. Privy Council Office (undated). Report of Commission appointed under Order-in-Council (P.C. 670) to enquire into industrial relations in Canada together with a minority report and supplementary report (1919), T. Mathers Chair.
  7. While the report may have been inclusive in terms of the nature of the employment contract and even the reasons for separation from work, it also contains derogatory language about the employability of different races.
  8. Author’s estimate using Statistics Canada table: 14-10-0006-01 and Table: 14-10-0017-01.
  9. Author’s estimate of the monthly average between February 2019 and February 2020 using Statistics Canada Table: 14-10-0017-01. Here “unemployed” means a person 15 years of age or older who was not working and was looking for work at the time.
  10. Author’s estimate of the average between February 2019 and February 2020 using Statistics Canada table: 14-10-0009-01. An additional 13,000 received benefits specific to workers in the fisheries and 2,100 received benefits under a Work-Sharing Agreement with an employer where the aim is avoiding layoffs by having workers share full-time work and receive EI benefits to top-up their reduced wages.
  11. Johal, S. (2019). “Facilitating the Future of Work Through a Modernized EI System,” Public Policy Forum; Banting, K. and Medow, J. (2012). Making EI Work: Research from the Mowat Centre Employment Insurance Task Force, McGill-Queen’s University Press; Gray, D. (2004). “What Employment Insurance Reform Delivered,” C.D. Howe Institute; standing committee on human resources (2001).”Beyond Bill C-2: A review of other proposals to reform Employment Insurance.” House of Commons.
  12. See for example, Robson, J. (2017). “Parental Benefits in Canada: Which Way Forward?” IRPP Study, no. 63, Institute for Research on Public Policy, Montreal.
  13. Statistics Canada (2019). “Employment Insurance Coverage Survey,” The Daily. See Table 5.
  14. Department of Finance (2020). “Canada’s COVID-19 Emergency Response: Bi-Weekly Report on Parts 3, 8, and 18 of Bill C-13. Eighth Report, July 9, 2020”. CERB claimants are asked a series of screening questions to determine whether their claim will be processed through Service Canada or the Canada Revenue Agency. Claimants who do not appear to be covered by the EI program are directed to apply through the Canada Revenue Agency. Even assuming some margin of error in applicant understanding or compliance, a significant share of CERB claimants would not have been covered by the EI program.
  15. Statistics Canada (2019). Supra at note 13. Table 4.
  16. Appeals are permitted through the Social Security Tribunal.
  17. Employment and Social Development Canada (undated). “Employment Insurance (EI) and voluntary leaving,” website, Government of Canada.
  18. For a discussion of record-of-employment codes and directives to employers, see Employment and Social Development Canada (undated) “ROE Secure Automated Transfer (ROE SAT) 4.0 – User guide,” website, Government of Canada.
  19. Subject to program eligibility rules, the EI program can pay benefits to mothers who have given birth, birth and adoptive parents of an infant up to 18 months of age, and family members providing care to a critically ill or injured child.
  20. Examples are discussed in Robson (2017), supra note 12.
  21. Robson, J. (2020). “Radical Incrementalism and Trust in the Citizen: Income security in Canada in the time of COVID,” Canadian Public Policy, advance online COVID-19 papers.
  22. Employment and Social Development Canada (2020). “Employment Insurance Monitoring and Assessment Report 2018-19,” Government of Canada.
  23. Ibid.
  24. Employment and Social Development Canada (2016). “Evaluation of Employment Insurance (EI) Automation and Modernization: Final Report,” Government of Canada.
  25. Ibid. The authors of that report caution that the costing method does not necessarily reflect a full-costing model. In economic terms, the efficiency gains from greater automation and simplification of the benefit application and payment systems should also measure savings to individual program users and to employers.
  26. COBOL stands for “Common Business-Oriented Language” and was developed by Grace Hopper (Vassar College) as a coding language intended to reflect English vocabulary and syntax.
  27. The same trend has also been observed in the United States. For a discussion, see Ryseff, J. (2020). “COVID-19 Highlights the Shortcomings of America’s Digital Infrastructure,” RAND Blog, May 14.
  28. For a longer discussion of this operational challenge, see J. Robson (2020). “Why can’t they just cut us all a cheque?, First Policy Response, Ryerson Leadership Lab.
  29. United States Government Accountability Office (2020). “Coronavirus Oversight,” website.
  30. Ways and Means Committee (2020). “Economic Impact Payments Issued to Date (June 5),” United States Congress.
  31. In normal times, EI recipients who continue working will, under the working-while-on-claim provision, see a portion of their benefits reduced up to a ceiling on weekly employment earnings. The system relies on participants to self-report on their employment hours and earnings, but payments will be withheld until a regular bi-weekly report is received.
  32. Internally in government this is referred to as the Unemployment Insurance Calculation Module. See ESDC (2016). Supra note 24, Annex B.
  33. This paper has dealt only with the income support side of the EI system. The employment services side forms a very important part of the program and also shows serious inequities and gaps. It is a question that merits significant discussion that goes well beyond the scope of this brief paper.
  34. Other countries have adopted systems referred to as e-payroll in which employers regularly remit payroll information to the national tax agency at the individual level, rather than the aggregate for the payroll in that period as is the case in Canada.
  35. Mendelson and Banting (2011). “Fixing the Hole in EI: Temporary income assistance for the unemployed”, Mowat Centre, University of Toronto.
  36. In fact, media reports suggested that the Government of Canada received a transition briefing in 2019, alerting it that several statutory programs were at risk of critical failure due to the aging of the IT infrastructure. See Press, J. (2020). “Federal IT systems at risk of ‘critical failure,’ Trudeau warned in memo,” The Canadian Press, February 1.
  37. Acemoglu, D. (2020). “The Post-COVID State,” Mirror-Spectator, June 11. See ESDC (2020), supra at note 22.

Private Sector Partners: Manulife & Shopify

Consulting Partner: Deloitte

Federal Government Partner: Government of Canada

Provincial Government Partners:

British Columbia, Saskatchewan, Ontario & Quebec

Research Partners: National Research Council Canada & Future Skills Centre

Foundation Partners: Metcalf Foundation 

PPF would like to acknowledge that the views and opinions expressed in this article are those of the author(s) and do not necessarily reflect those of the project’s partners.