To further investigate procrastination, we sought a Natural Decision Making (NDM) setting (Klein, 2008). As reviewed by Lipshitz et al. (2001), “NDM is an attempt to understand how people make decisions in real-world contexts that are meaningful and familiar to them” (p. 332). The NDM approach is descriptive (i.e., how people actually make decisions) rather than normative (i.e., a rational or ideal decision-making model) and, by studying people in their “natural habitat,” it tests the degree to which laboratory findings generalize and are practically applicable, that is external validity. This approach also addresses the criticism that the applied psychological or organizational behavior field has a dearth of field research (Cialdini, 2009), with Baumeister et al. (2007) direct call for “a renewed commitment to including direct observation of behavior whenever possible and in at least a healthy minority of research projects” (p. 396). Ideally, in this setting individuals have autonomy over their work tasks, style, and pace. There must be a stock and flow of tasks needing attention, with incentives for completion of work. The dependent variable of time must be unobtrusively observed and accurately measured, with some consistency in task from person to person, and from product to product.

To this end, arbitrators are the ideal subjects. Labor arbitration is a widely adopted procedure to resolve disagreements about the application and interpretation of collective bargaining agreements (CBA) between unionized workers and their employers as well as the appropriateness of employe discipline (Bemmels, 1988; Bemmels and Foley, 1996; Trudeau, 2002; Ruben, 2003; Thornicroft, 2010). Under the model used in United States and Canada, employes who believe that a contractual term of their CBA has been violated can file a grievance alleging a contract violation and seeking a remedy. Most grievances are then subject to joint union-management discussion involving progressively higher levels of the respective organizations and most are resolved during such discussion (Lewin, 1999). Unresolved grievances are subject to arbitration.

Consequently, arbitrators produce written decisions that bring justice into the workplace, often juggling multiple cases while maintaining a regular pipeline of appointments. They must accept new cases while urgently trying to find time and energy to issue awards well past the dates of the actual hearings. Potentially, with arbitrators we can consider all three major sources of variability that could influence decision time: task characteristics, the work environment, and individual differences. Of particular interest is whether these individual difference variables remain relevant when taken out of the laboratory and examine in a NDM setting. We review each of these contributions in turn.

In an empirical study of labor arbitration cases using event history analysis, Ponak et al. (1996) divided the labor arbitration process into four distinct stages: (1) pre-arbitration grievance steps; (2) arbitrator selection; (3) hearing scheduling; and (4) preparation of the arbitration award (which we will refer to as “Decision Time”). This final stage accounts for approximately 20 percent of the overall elapsed time from the filing of the grievance to the arbitrator’s decision (Steiber et al., 1985; Ponak and Olson, 1992). The first three stages are subject to the input of multiple parties, often with competing interests. The final stage, however, is almost entirely controllable by arbitrators and they have considerable discretion over Decision Time. Here, their autonomy is among the highest for any professional. They customarily work independently, have a great deal of control over their work scheduling, and over how to balance their work and private commitments. There are few deadlines except ones self-imposed, that is ‘‘slippery deadlines’’ in that they can be pushed ever forward into the future¹. This makes their work environment not so strong that there isn’t room for discretionary behavior (Mischel, 1977; Withey et al., 2005).

On the other hand, unnecessary delay is strongly discouraged. A body of literature exists on the sources of arbitration delay and there is a consensus that delay is increasing and is harming the labor relations system (Berkeley, 1989; Ponak and Olson, 1992; Thornicroft, 1995; Foisy, 1998; Lewin, 1999; Trudeau, 2002). More colorfully titled articles along this line include “The Well-Aged Arbitration Case” (Ross, 1958) and “Delay: The Asp in the Bosom of Arbitration” (Seitz, 1981). The parties to the arbitration anxiously await the decision, being consequential to grievances and to union-management relations. Furthermore, arbitrators feel an obligation to dispense justice quickly, being well aware of the adage “justice delayed is justice denied.” An arbitrator’s future acceptability may be harmed by a reputation for tardiness (Berkeley, 1989). There also is a financial incentive for arbitrators to work quickly toward a decision because the issuance of an award allows final billing to the parties. The National Academy of Arbitrators (NAA), the pre-eminent professional association of North American labor arbitrators, contains a section in its Code of Professional Responsibility devoted to “Avoidance of Delay” and arbitrators have been sanctioned for undue delay. Clearly, undue delay by an arbitrator is irrational.

A summary of arbitrator Decision Time reported averages ranging from 37 to 101 days depending on the time period covered and region (Thornicroft, 2010). These averages undoubtedly mask considerable variation from case to case, but most studies either report averages and medians without other statistics, or do not separate Decision Time from other parts of the process. The most comprehensive study of Decision Time and its predictors analyzed 500 arbitration decisions over a 4-year time period in the province of Alberta. It found a mean Decision Time of 65 days, a standard deviation of 56 days, and a maximum Decision Time of almost 1 year (Ponak et al., 1996). Notably, all previous studies that sought to explain Decision Time were based on content analysis of written awards, and scholars were able to explain less than one-third of the variance in Decision Time.

With the assistance of two co-authors, who are themselves highly placed within the arbitration community, we were able to comprehensively assess two groups of arbitrators, one Canadian and one American. Though the American sample has the advantage of being larger in size, in Canada, the law requires that all arbitration decisions be publicly reported and are available electronically. Critically, this reporting stipulation allows the Canadian sample to provide an unobtrusive measure of delay but also allows it to almost fully meet all of Roe (2014) suggested criteria for longitudinal studies (e.g., extended time, numerous points of assessment, common task). All data and analyses are available in an Open Science archive: https://osf.io/qsfht/?view_only=3a2144fa0d9d4a388197e33caddc825b.

Decision Time should be partially accounted for by motivational individual difference factors. Even when people have the capability, opportunity and intention for action, self-imposed delays can occur. Studied under a variety of terms depending upon the field, such as an “intention-behavior gap” (Sheeran, 2002), “value-action gap” (Barr, 2004) and “attitude-behavior consistency” (Crano and Prislin, 2006), people procrastinate due to individual differences, with procrastination itself appearing to be a personality trait as well. Enduring across time and situation, self-report procrastination has an average test-retest reliability after 42 days of 0.73 (Steel, 2007) and, like other personality traits, approximately 50% of the variance in trait procrastination is inheritable (Luciano et al., 2006). Here, we consider procrastination, expectancy, value, sensitivity to time, and perfectionism.

Hoyle (2010) reviews several frameworks that integrate self-regulation strategies with personality traits. One way is to consider self-regulation as the proximal outcome of distal personality processes. From this perspective, there is a loose relationship between the two, providing room for the independent acquisition of self-regulatory strategies and for them to account for unique variance. For example, though organizational skills are related to the personality trait of conscientiousness, they can be trained and acquired (Klein and Lee, 2006). With regards to procrastination, Koch and Kleinmann (2002) argue that time management skills should help reduce hyperbolic discounting and impulsiveness. Using TMT (Steel and König, 2006), we explore three strategies that should help, hinder or reflect arbitrator delay beyond what personality can explain: Organization, Multitasking and Pacing Style.

Though all procrastination requires delay, not all delay is procrastination. People may put off finishing a task due to other more pressing obligations or a task may be more onerous and take longer. Previous arbitration studies on task delay have exclusively focused on the work or task environment (e.g., Rose, 1986; Barnacle, 1991; Ponak et al., 1996), overlooking or discounting the impact of dispositional variables and self-regulatory techniques. Still, both the type and the process of grievance did influence time delay, accounting for up to 27% of the variance (Thornicroft, 1995). Drawing on these past examinations, we re-examine several key task or environmental factors. In particular, we are interested in comparing the power of these traditional external predictors of delay with individual difference variables.

Our target population of Canadian arbitrators who are members of the NAA is relatively small, approximately 60–70 people depending on the exact year. From this group, 49 agreed to participate, a response rate of approximately 75%. This response rate is largely due, as mentioned, to two of the authors being active within the NAA. Also, these respondents represent a high proportion of the entire labor arbitration “business” as it is highly concentrated. Between 31 and 35% of all Canadian cases are decided by NAA members (Trudeau, 2002). Average age was approximately 46–50, 75% were men, and 80% were married. The average years of arbitration practice was 16–20, though most were NAA members for only 6 to 10 years.

Participants responded to over 200 questions, either through an online survey or identical hardcopy paper versions handed out during the NAA Annual Meeting. This approach was necessary to ensure coverage across all participants, though the majority of arbitrators responded through the online version. There were no cases in which a participant responded to both versions of the survey. Means, standard deviations and reliabilities of the measures are all reported in Table 1. To provide ensure domain coverage, the individual difference variables were typically operationalized in several ways, all at the trait level. Procrastination was assessed through two scales, the nine item Irrational Procrastination Scale (Steel, 2010; e.g., “I delay tasks beyond what is reasonable”) and a three-item content valid procrastination scale specific for arbitration (e.g., “The quality of my arbitration decisions is impaired by my procrastination”). The measurement of expectancy was through the Work Self-Efficacy Scale (Speier and Frese, 1997; e.g., “If I am in trouble, I can think of a solution”). The measurement of value was assessed at the trait level by three scales. Energy was measured by the Lack of Energy Scale (Kuhl and Fuhrmann, 1998; e.g., “If work is distasteful or boring, I have to force myself to get going”), need for achievement was measured by the Personality Research Form (Jackson, 1984; e.g., “I enjoy difficult work”), and task aversiveness for arbitration writing was measured by a content valid three-item scale (e.g., “I enjoy writing arbitration decisions” – reverse scored). Finally, sensitivity to delay was measured by the Attentional Distractibility Scale (Kuhl and Fuhrmann, 1998; e.g., “My mind wanders when I try to concentrate”) and the Susceptibility to Temptation Scale (Steel, 2010; “It takes a lot for me to delay gratification”).

For the self-regulatory variables, we devised scales relevant for the arbitration practice based on inputs from two authors of the study who are part of the arbitration community. Organization was measured by a three-item scale (e.g., “I am very methodical in the way I approach my arbitration practice”). Multitasking was assessed with a 12-item scale (e.g., “I multitask,” “I write in short bursts between other non-writing activities”).

For busyness, we had the arbitrators retrospectively estimate work and personal commitments across separate life domains for that period. On a five-point scale, ranging from “Not at all” to “Extremely,” they indicated how involved they were in the following areas: Arbitrator selection, processes and evaluations; Mentoring other arbitrators; Conference preparation; Non-arbitration work; Physical fitness and recreation; Caring for children or grandchildren; Elder care for relatives; and Charity/community events/volunteer work. These life domains were selected based on a pilot study where five NAA practitioners were interviewed about activities that could impact the speed at which they write decisions. A self-report Workload item was devised by summing responses across all life domains.

By law, almost all arbitrators in Canada must file their decisions with a Ministry of Labour or an equivalent body and these decisions are publicly available. We acquired all cases produced by our arbitrator sample over a 3-year period (2003--2005, inclusive). Three years’ worth of cases enabled us to develop an accurate portrait of the ‘‘normal’’ workflow of any arbitrator. The comprehensive data set of cases is a significant advantage of doing research in the Canadian setting. By contrast, American arbitration decisions are not routinely filed and are considered to be private documents³.

From the original group of 49 participating arbitrators, 41 were practicing during our sample period or provided hearing dates with which to measure arbitration delay. In total, LexisNexis^® Quicklaw™ Research Service provided full-text retrieval of all 1,370 cases, for an average of 33.4 cases per arbitrator. This number reduced to 1,204 cases from 40 arbitrators after eliminating panel decisions. Notably, this figure compares favorably with the 350 cases analyzed by Thornicroft (1995), the 600 cases by Ponak et al. (1996) or the 800 cases by Barnacle (1991).

Our coding adopted the approach of Ponak et al. (1996). The dependent variable, Decision Time, is the elapsed time between the date of the final hearing and the issuance of the award. Decision Time is an unobtrusive and concrete measure, as arbitrators routinely include this information without any notion that researchers will make use of it. We found that the average Decision Time was 61 days, comparable to the average of 93 days reported in the United States (Federal Mediation and Conciliation Service, 2011). However, with a range from same day to 1639 days per case or an average delay of 13 to 718 days per arbitrator, the positive skew was substantial. Accordingly, for comparisons with level 2 variables, we used a logarithmic transformation, reducing skew from 5.92 to 1.48 but still correlating with the untransformed scale at 0.79.

In addition, we distinguished regular and expedited cases and coded for grievance complexity. Complexity was assessed in terms of page length and time span of hearing. However, after standardizing these two variables, they formed a scale with a Cronbach’s Alpha of just 0.52, suggesting separate analyses for each. Finally, workload was operationalized as the publicly available written arbitration decisions the arbitrator released in the given year.

Because our data have a multilevel structure (cases nested within individuals), all analyses involving case or task characteristics were conducted using multilevel modeling using the R (version 4.1.1) package “nlme” (version 3.1–153). Like other widely used programs, nlme uses random-coefficient modeling for multilevel data and performs appropriate adjustments to the analytical procedures to account for nested data. Our complexity variables of Decision Length and Time Span of Hearing as well as Procrastination were centered prior to forming their cross-level interaction. Total observations were 1204 cases (level 1) from 40 arbitrators (level 2).

Descriptive statistics and a correlation matrix of our data are presented in Table 5. Table 6 presents the results of our multilevel analysis predicting days of delay. Initially, we tested a random intercept null model, where each arbitrator gets their own intercept, using it to calculate the Intraclass Correlation Coefficient (ICC). Approximately 66% of the variance is at level 2 or the arbitrator level, justifying taking a multilevel approach and supporting that individual-differences are a major contribution to delay.

The results reconfirm our previous analysis; Observed Workload is not related to delay. They also re-confirm that Trait Procrastination is significantly associated with the number of days it takes to render a decision (i.e., Hypothesis 1.1). The coefficient indicates that for each standard deviation increase in procrastination, elapsed time from hearing to decision increased by 28.29 days (t = 2.06, p = 0.046). Having a similar effect but in the opposite direction is Expedited cases, as per Hypothesis 10.1. Expediting a case reduces Decision Time by an average of 20.54 days (t = −2.99, p < 0.01). In other words, it would be preferable to select a low procrastination arbitrator without expediting than a high one with. Using the means from Table 5 with the coefficients from Table 6 (with the exception of Time Span, which being standardized has a mean of zero), we can calculate the impact of procrastination on delay. Keeping all other variables constant at their average, arbitrators one standard deviation above on procrastination take 82.94 days while those one standard deviation below take 26.38 days.

Finally, grievance complexity as measured by Decision Length increased Decision Time, with each written page adding 1.19 days (t = 5.25, p < 0.001). Similarly, our other complexity measure, the Time Span of Hearing Days, also significantly contributed to delay (t = 11.25, p < 0.001), which is somewhat harder to interpret as it follows a Pareto distribution with approximately half the cases (643 or 53.4%) at the lowest value or 1. Still, as the time span increases, so does decision time, supporting Hypothesis 11.1. For both these complexity measures, we detected an almost identical interaction with procrastination: for Decision Length (coefficient = 6.948, t = 3.67, p < 0.001) and for Time Span of Hearing Days (coefficient = 11.41, t = 4.19, p < 0.001). Given their redundancy, we incorporate just the interaction effect for Time Span of Hearing Days in Table 6 (though both analysis scripts are available in our OSF folder). To illustrate the interaction, we use reghelper’s (version 1.0.2) “nlme” option to run a Simple Slopes analysis in R. The simple slope for time span of hearing was positive and strong (coefficient = 34.07, t = 10.41, p < 0.001) for those who were one standard deviation above the mean in procrastination, but the simple slope for time span of hearing was a third the size (coefficient = 11.25, t = 3.22, p = 0.001) for those who were one standard deviation below the mean in procrastination, supporting Hypothesis 11.2. In other words, those low in procrastination pursued tasks at nearly the same pace regardless of the underlying complexity. Those high in procrastination increasingly dillydallied the more complicated tasks became.

Of the approximately 600 United States members of the NAA, we obtained responses from 195, or 35% of the total population. Average age reported was 70 to 79 by 51% of the respondents, 72.3% were men and 27.7% were women, 84.6% were married with 2% reporting never married, 95.4% reported having an advanced graduate and/or law degree, and 71.8% hearing arbitration cases for 26 years or more. This represents a well-established and experienced group of arbitrators.

Measures employed were largely a subset of Study One’s, specifically those scales dealing with: procrastination, self-efficacy, task aversiveness, lack of energy, distractibility, and multitasking. Reference Study One for details. We also made one refinement and two additions. Consistent with Study One, we assessed self-report workload or busyness but this time used a more standardized scale, the Martin and Park Environmental Demands (MPED) Questionnaire (Martin and Park, 2003). Its six-item scale asks questions such as “I am very busy during an average day.” We also assessed Pacing Style using the nine-item Pacing Action Categories of Effort Distribution (PACED). It consists of three scales: Deadlines (completing work just before the due date; e.g., “I generally do not work until there is time pressure from an approaching deadline”), U-Shaped (completing work mostly at the start and finish with a break in between; e.g., “The effort I put into projects is high at the start, low half-way through, and high again at the end”), and Steady (complete work at an even pace throughout; e.g., “I work in a slow, but steady, manner to complete tasks”). Finally, to measure perfectionism’s discrepancy dimension, we used three items from the Revised Almost Perfect Scale (Slaney et al., 2001), framed to assess other-focused rather than self-focused discrepancy, specifically: “Others are hardly ever satisfied with my performance,” “My best just never seems to be good enough for others,” and “My performance rarely measures up to other people’s standards.” As per Study One, we used Harman’s Single-Factor Test to evaluate common method variance, and found that the single factors accounted for 29, 14, and 17% of variance across item subsets.

We studied arbitrators, who are both an ideal and a stringent test of personality and task effects on workplace procrastination. Arbitrators have high degrees of autonomy, meaning that personality effects should not be erased by high situational strength. The nature of their work provides a difficult to obtain criterion: an objective and valid measure of task delay, which is a key indicator of performance for arbitrators. It is stringent because although arbitrators have opportunities to delay, they have strong motivations not to and are carefully selected by both management and union representatives. As Domke and Edmonson (2003) underscore “The arbitrator is a decisive element any arbitration. The success or failure of arbitration will largely depend on them” (p. 24). Furthermore, given that arbitrators in our sample had practiced for a number of decades on average, there was ample time for less competent arbitrators or chronic procrastinators to self-select themselves out of the profession (Schneider, 1987). Indeed, average levels of procrastination, which are over 1.5 standard deviations lower than the average population (Steel, 2010), and other related traits indicate they are a group who should procrastinate remarkably little. And yet we still detected a substantive effect.

Given that procrastination made a substantive impact on the performance of this rarefied group of highly trained and motivated arbitrators, procrastination should be a significant factor in most autonomous work situations. Epidemiological work by Steel and Ferrari (2013) indicates that procrastination is indeed a global phenomenon with significant levels in most demographic categories. Supporting this conclusion is research showing procrastination substantively occurring in settings ranging from academic publication (Lay, 1987; Ackerman and Gross, 2007) to unemployment search (Lay and Brokenshire, 1997; van Hooft et al., 2005).

Since procrastination is putting off despite expecting to be worse off, it is naturally detrimental. Only its negative impact varies, tending to intensify the higher the stakes (Steel, 2011). As mentioned, timeliness is a critical performance dimension in the justice field in general and for arbitrators specifically and, as expected, trait procrastination was a major predictor of Decision Time. Arbitrators one standard deviation above the mean in terms of procrastination took over twice as long as arbitrators one standard deviation below. The effect of procrastination is virtually identical, though opposite in direction, to that of expedited cases, which represent the combined effect of several procedural interventions designed to speed task completion (e.g., allowing arbitrators to issue non-precedent setting awards). Similarly, procrastination’s effect on Decision Time was comparable to the effect of workload, which was the combined effect of other personal and work commitments. In short, personality effects on Decision Time were as strong as environmental procedures or external obligations.

Still, it is worth stressing that the procedural interventions were successful in reducing delay (i.e., expedited cases). If organizations find timeliness a problem, it can be partially ameliorated through well-designed processes. As Heath and Anderson (2010) note while reviewing external causes of undue delays, “supportive social scaffolding enables us to keep procrastination in check” (p. 233). For example, elements that speed task completion include the establishment of social norms, environmental cues and reminders, public and specific implementation intentions, and institutionally enforced rewards and repercussions (Steel, 2011). However, given that the practice of job design has generally drawn on mechanistic rather than motivational models, this avenue appears typically unexploited (Campion et al., 2005).

The importance of procrastination should be of more surprise to those who have adopted a rational model of human behavior, often associated with neo-classical economics (Posner, 1998; Ashraf et al., 2005). According to economic theory, which is essentially Expectancy × Value formulations, procrastination should not exist, being an irrational delay, or at least not have a substantive impact. Reflecting the lack of field research on the topic, as per Berg and Gigerenzer (2010), information of the sort presented is exceedingly rare, stating that “Notably missing is investigation of whether people who deviate from axiomatic rationality face economically significant losses” (p. 133) and “the normative interpretation of deviations as mistakes does not follow from an empirical investigation linking deviations to negative outcomes” (p. 150). However, the results here are consistent with the meta-analytic research showing a reliably negative relationship between performance and procrastination (Steel, 2007) as well as research connecting procrastination to diminished health and reduced financial success (e.g., Mehrabian, 2000; Elliot, 2002; Sirois et al., 2003). When people report that they tend to put off despite expecting to be worse off, their expectations are often borne out.

There are several competing theories of procrastination, especially that procrastination is caused by perfectionism (e.g., Egan et al., 2011; Mohamadi et al., 2012). Here, we conducted an explicit test of Temporal Motivation Theory (Steel and König, 2006), which decomposes procrastination into expectancy, value and sensitivity to time related variables. Together, these variables account for approximately 57% of the variance at the trait level and their effect on observed delay is indeed mediated by procrastination. Furthermore, as predicted by Temporal Motivation Theory and confirmed by the multilevel analysis, procrastinators were especially vulnerable to difficult and less enjoyable tasks, as operationalized by grievance complexity. Those low in procrastination were far less affected by aversive task characteristics while those high in procrastination increasingly tended to delay. Given that procrastination is associated with reduced self-efficacy, these results are also consistent with Beck and Schmidt (2012) experimental research on self-efficacy, resource allocation, and goal difficulty. The more difficult a task becomes, the fewer resources are allocated to it by those lacking self-efficacy. Similarly, Louro et al. (2007) argue that resource allocation is a function of goal proximity and the valence of goal relevant emotions. Taken together, we are starting to have a firm understanding of when as well as why procrastination occurs.

The results from our investigation into self-regulatory skills and perfectionism spanned the range of outcomes: positive, negative and neutral. Notably, the propensity to procrastinate was incrementally predicted by self-regulatory skills, particularly Organization. In the other direction, Multitasking is confirmed as a largely dysfunctional work strategy, associated with procrastination, longer delays and being susceptible to temptations. In total, trait and self-regulatory variables accounted for an impressive 73% of the variance in procrastination in Study One and 58% in Study Two (which did not assess Organization). This almost exactly duplicates Steel et al. (2018) finding; a few variables accounted for 74% of the variance in procrastination, drastically limiting the amount of residual variance left to predict and other variables’ potential unique impact, such as perfectionism. As stressed here as well as by Xie et al. (2018), perfectionism’s connection to procrastination accounts for no incremental variance after these mediating processes are taken into account.

Again, the results stress the need to include time related variables and longitudinal perspectives both in motivational theory and practice. For example, Pepper and Gore (2014), also using a Temporal Motivation Theory frame, found that the excessive application of Expectancy × Value formulations has resulted in largely ineffective compensation plans, where existing “long-term incentives are not an efficient way of motivating senior executives, irrespective of national culture” (p. 26). In short, there are considerable costs in employing an overly stripped down or simplified motivational model.

This research underscores many of the same points recently made in the motivational field (Fishbach and Zhang, 2009; Mitchell et al., 2009; Lord et al., 2010), such as Beck and Schmidt (2012) who discuss how we should view “motivation as a process of decision-making over time” and that “goal difficulty is a critical determinant of motivational processes” (p. 206). Accordingly, future research should continue along these lines, looking at motivation dynamically. We consider here a potential research program that facilitates this pursuit, which also addresses inherent limitations with this study.

One of the limitations of this study comes from its strength, our realistic or NDM setting (Lipshitz et al., 2001; Klein, 2008). Our use of arbitrators enabled a rigorous, unobtrusive, temporal measure of the work performed. It also makes replication extraordinarily difficult, especially without collaboration from within the sample frame, that is, fellow arbitrators. Even with a collaborator (i.e., two of the authors are active within the arbitration community), acquiring participants while keeping the assessment battery brief enough and with sufficient face validity to ensure completion was a reoccurring issue. We struck a balance in assessment by using a combination of context specific as well as more general scales (e.g., as for procrastination). Despite this issue, we did manage to administer a large assessment battery of up to 200 items and were able to replicate many of the results in Study Two with a much larger sample. Still, for this type of multilevel design, Study One represents a large number of participants and cases (e.g., Claessens et al., 2010).

Finally, using archival case data limits the information available from these cases to what has been recorded. For example, we are unable to examine state level manifestations of expectancy, value and sensitivity to time that the arbitrators actually experience as they encounter each individual case. Accordingly, these issues are somewhat unavoidable, inherent to constraints that most field studies with measures of actual behavior encounter (Baumeister et al., 2007). However, that the available case information exists is itself notable. The arbitration cases themselves are part of the public record within the Canadian system but are only sporadically available in other countries, such as the United States. To our knowledge, there are no other comparable groups of professionals where people are performing essentially the same or easily comparable but measurable “slippery deadline” task frequently over potentially several years. However, there is a promising non-professional option to explore.

Steel and König (2006) recommend investigating motivation dynamically with courses taught through a computerized personal system of instruction (C-PSI). As they review, a C-PSI course allows: “hundreds of students simultaneously working toward completing a university course at their own pace, allowing choice and thus motivated behavior. Furthermore, progress is assessed at a large number of data points as the course is broken down into numerous assignments, all computer administered with completion precisely recorded. Similarly, a host of other observed and self-reported measures can be easily inserted into this framework” (p. 906–907). In short, we have a meaningful task, a “slippery deadline,” and precise start and completion dates. As per the present study, this is ideal since the tight connection between observed delay and procrastination means, in the absence of an accompanied early start, observed delay over several tasks can be taken as a reasonable proxy of procrastination despite the lack of a formal assessment of irrationality. Previous use of the venue by Steel et al. (2001) established that C-PSI courses are rife with procrastination and other forms of self-regulatory failure.

Furthermore, there has been rapid growth of Massive Open Online Courses (MOOCs), such as those being offered by MIT, Harvard and Berkeley (Martin, 2012). With enrollment in the hundreds of thousands, rather than the hundreds, it allows for an ongoing and detailed investigation of self-regulation. Partly due to our choice of sample, our battery of self-regulatory measures was limited and a few measurement issues arose after contextualizing it to an arbitration setting (e.g., the reliability for the self-regulatory measure of Organization was lower than expected). Though we could touch on the importance of self-regulation, proper examination of self-regulation requires deep examination, befitting a topic that consists of “multiple processes and components that interact in complex ways” (Boekaerts, 2010, p. 71). MOOCs appear ideal for this purpose, given that their sample size and number allow for continuous and varied exploration (Diver and Martinez, 2015).

The benefit of a more detailed examination would be to isolate key proximal skills that influence goal striving but are also trainable. As Steel and Klingsieck (2016) emphasize, procrastination interventions should not focus on the neurotic procrastinator, which comprises perhaps 10% of all procrastinators, but customized to individuals’ specific set of temptations, circumstances and target tasks. Since procrastination has separate expectancy, value and impulsiveness components, treatment should first proceed by identifying specific weaknesses in these areas and then targeting appropriate skills. For example, those specifically lacking self-efficacy will not necessarily respond to value interventions (e.g., interest enhancement) or impulsiveness interventions (e.g., attentional control). One size does not fit all. Online computerized courses will enable us to assess the interaction effect between prior need and the development of self-regulatory skills, rather easily if the self-regulatory interventions can also be administered as a C-PSI. If the massive and constantly renewing sample offered by MOOCs can be drawn upon, the dynamic study of motivation can be expected to advance quite rapidly. A particularly notable though early effort along these lines is Huang et al. (2021).

Finally, arbitrators represent some of the lowest scoring professions for procrastination, but even within this elite group procrastination had an impact. In today’s deadline strewn and temptation-soaked world, completely avoiding procrastination appears to be a rarity. As research such as Rinaldi et al. (2021) stress, this ubiquity reflects that the neurobiological roots of procrastination run through the brain’s executive function, which we all have. Though this is consistent with Temporal Motivation Theory (Steel, 2011), we would benefit by moving beyond summary variables (e.g., impulsivity) and take a more detailed approach. By identifying the exact neural mechanisms that are impaired or at least operating less than optimally, new diagnostics and interventions to minimize procrastination are likely to emerge.