Cyberloafing: The Cost Figures Come From Companies Selling the Cure

Cyberloafing is the academic name for using the internet at work for things that are not work. It has a large research literature, a set of enormous cost figures, and a standard managerial remedy. The literature does not support the remedy. It barely supports the panic.

Two numbers carry this article. The largest synthesis of the field found that cyberloafing was unrelated to job performance. The largest synthesis of workplace electronic monitoring found that monitoring's relationship with performance was r = -0.01.

And every headline cost figure you have seen traces to a company selling monitoring software, a job board, or nothing at all.

What the evidence actually says about the behaviour

Mercado, Giordano and Dilchert (2017), in Career Development International, meta-analysed 54 independent samples contributing 609 effect sizes across 39 correlates.

The strongest predictors were boredom, engagement and self-control - self-control strongly negatively. Age effects were "negligible."

Then the finding that governs everything else. Cyberloafing was strongly correlated with overall counterproductive work behaviours but unrelated to other components of job performance.

Read that carefully, because it is not the same as saying cyberloafing is harmless. It says cyberloafing travels with a family of other rule-bending behaviours. It does not say the browsing damages the work.

That distinction matters because the entire cost argument assumes the second claim, and the evidence supports only the first.

The foundational study, and what we could not verify

The paper that named the field is Lim (2002) in the Journal of Organizational Behavior, which found that employees cyberloafed more when they perceived unfair treatment, and used neutralisation techniques - "I'm owed this" - to justify it.

We are citing it for that argument and nothing more. We could not read the paper. Four separate routes to the full text failed. We are therefore not printing its sample size, its response rate or any of its statistics, and neither should anything citing this article. The finding as summarised is real; the numbers are not in our hands.

That is worth stating plainly rather than quietly citing an N we have not seen, which is how most secondary accounts of this literature are assembled.

The remedy does not work

If cyberloafing is the disease, monitoring is the prescription. It has been tested.

Siegel, König and Lazar (2022), in Computers in Human Behavior Reports, meta-analysed 70 independent samples and 233 effect sizes on electronic monitoring:

Outcome	Association
Job satisfaction	r = -0.10
Stress	r = +0.11
Performance	r = -0.01
Counterproductive work behaviour	r = +0.09

Their own summary, verbatim: "maintaining and improving the performance of employees is an important justification of electronic monitoring. However, the current meta-analysis found no relationship between electronic monitoring and performance... These results question the benefits of electronic monitoring for organizations."

Monitoring lowers satisfaction, raises stress, buys nothing measurable in performance, and travels with more of the counterproductive behaviour it exists to suppress. Their moderator analyses suggest performance targets and feedback make the negative effects worse, and that laboratory studies underestimate the field effects.

Thiel and colleagues (2023), in the Journal of Management, propose the mechanism: monitoring increases deviance through moral disengagement and displacement of responsibility - if the employer has taken responsibility for policing you, you have less of your own. Perceptions of overall organisational justice buffered the effect. We are not printing their sample sizes, which we could not obtain.

The counter-narrative is only half true

There is a well-known argument on the other side: short non-work browsing is restorative, and people come back sharper.

Coker (2011), in New Technology, Work and Employment, is where the famous version comes from. 268 office workers, mean age 33, drawn from a commercial market-research panel. Those who engaged in workplace internet leisure browsing scored 70.30 (SD 11.90) on a work-productivity scale against 64.66 (SD 15.14) for those who did not, F(1, 258) = 12.23, p < .01 - the widely quoted "9 percent more productive."

Three things have to be said with it.

It is cross-sectional and entirely self-reported, including the outcome. It licenses no causal claim, and Coker himself calls for experimental designs.

The relationship is curvilinear, F(2, 249) = 9.00, p < .0001, with an inflection at roughly 12% of work time. Past that, the benefit reverses. That number is more useful than the 9%.

Short and frequent beat long and infrequent: 72.57 (SD 7.31) against 68.92 (SD 11.93), F(1, 249) = 3.86, p < .05.

And the broader break literature does not deliver what people expect. Albulescu and colleagues (2022), in PLOS ONE, meta-analysed micro-breaks across 22 independent samples and 2,335 participants:

• Vigor: d = 0.36, 95% CI [0.16, 0.55], p < .001
• Fatigue: d = 0.35, 95% CI [0.19, 0.50], p < .001
• Performance: d = 0.16, 95% CI [-0.04, 0.37], p = .116 - null

Breaks reliably make people feel better. They do not reliably make people measurably better. And Egger's test on the performance outcome was significant (intercept = -2.63, p = .048), indicating small-study bias in exactly the outcome that failed.

The measurement problem, stated plainly

This is the most important methodological fact in the topic and it is usually omitted.

Cyberloafing research studies workplace rule-breaking almost entirely by asking people to confess to it on a single-source questionnaire. Most designs are cross-sectional. Many samples are students or paid online panellists. Objective-log studies are rare.

The field acknowledges this. Published work in this area notes that single-source data introduces common method variance, "artificially inflating predictor intercorrelations."

So the correlations that make up this literature are, in large part, correlations between two things the same person said about themselves in the same sitting, about a behaviour they have an incentive to under-report.

Where the cost figures come from

This is the part to remember.

"$63 billion a year lost to internet distraction." Traced to Websense in 2001 - a company selling employee-internet-monitoring software - released alongside a product version. The method, as described, was the average US salary multiplied by an assumed one hour of non-work surfing per employee per week. It is an assumption multiplied by a wage, published by a firm selling the remedy.

"$54-85 billion annually." Appears throughout the secondary literature. We could not trace it to any primary measurement.

"88% of employees surf non-work sites." Attributed to a survey by a commercial job board.

"62% of employees waste around 60 minutes a day on personal phone use." From a commercial e-learning vendor's marketing survey.

"American workers waste more than two hours a day." No primary measurement located.

The pattern is not incidental. Every large prevalence and cost figure in this field originates with a company selling monitoring software, a job board, or an e-learning vendor. Not one traces to a peer-reviewed measurement.

What this leaves you with

The behaviour is real and its predictors are boring. Boredom, low engagement and low self-control. Perceived unfairness. These are conditions, not character.

The performance case is not made. The largest synthesis found cyberloafing unrelated to job performance.

Monitoring is the worst available response on the evidence. It costs satisfaction and stress, buys no performance, and correlates with more of the behaviour it targets.

If you are the one browsing: the curvilinear finding is the practical one. Short and frequent, and somewhere under roughly a tenth of the day, is where the self-reported benefit sits. Expect to feel better rather than to measurably perform better - that is what the break evidence actually supports.

Nothing here is a claim about any individual workplace, and nothing here is employment advice. This article makes no recommendation about policy, discipline or monitoring practice.

For what interruptions do and do not cost, see what notifications actually do. For the accumulation of open work, see too many tabs open. For the wider picture, see the screens, attention and focus guide.

Limitations of this article

We could not read Lim (2002) and have printed none of its numbers.

We could not obtain the total N or the specific corrected correlations from the Mercado meta-analysis; the publisher's full text was gated. We print the sample and effect-size counts and the two findings the record supports.

We could not obtain sample sizes for Thiel and colleagues (2023) and have printed none.

The Coker result is self-reported and cross-sectional throughout, including the productivity outcome.

We found no cyberloafing meta-analysis published after 2017, despite searching. If one exists, this article predates it.

We did not capture the funding statement for the micro-break meta-analysis.