The Employment Appeals Board held that claimant’s discharge was not for “misconduct connected with work,” ORS 657.176(2)(a); OAR 471-030-038(3), and that he is therefore entitled to receive unemployment benefits. Employer seeks review, and we affirm.
EAB found that claimant worked for employer from 1979 to 1989 and was last employed as a chipper operator. In 1987, while working for employer, claimant submitted to a drug test and tested positive for methamphetamines. As a condition of continued employment, he agreed to participate in a rehabilitation program at employer’s expense and to sign a “last chance” return-to-work agreement, which provided that he would face disciplinary action, including dismissal, if he again tested positive. Claimant completed the rehabilitation program and returned to work. In 1989, the day after “claimant took some methamphetamines right after work,” he was randomly selected for testing while at work. He tested positive and was discharged.
At the hearing, employer offered no evidence that claimant’s off-duty controlled substance use affected his work or the workplace. EAB applied Lyle E. Zufelt, Jr., Opinion and Order 87-S-1587-E (1987) (Zufelt), where the Employment Division articulated its policies regarding eligibility for unemployment benefits of claimants who leave work or are discharged because of drug use or other drug-related activities.1 Zufelt says, as material:
“[BJefore benefits are denied for failing a drug test it should be shown that: (1) proper testing and confirmation procedures have been followed; and (2) there is clear objective evidence of impairment (e.g., bizarre behavior, substantial loss of productivity, etc.).
“DISCHARGE. If an employe is discharged for failing a drug test and there is clear objective evidence that the employe was drug impaired, it is a discharge for misconduct.”
EAB concluded that, because there was no evidence that on-*201the-job impairment resulted from claimant’s off-duty drug use, the discharge was not for “misconduct connected with work,” within the meaning of ORS 657.176(2)(a), and claimant was therefore not disqualified from receiving unemployment compensation.
Employer first argues that EAB erred by concluding that actual on-the-job impairment must be established in order for a claimant’s off-duty drug use to constitute work-connected misconduct. Employer relies on our statement in Glide Lumber Prod. Co. v. Emp. Div. (Smith), 86 Or App 669, 674, 741 P2d 907 (1987), that “off-the-job drug use can be disqualifying only if it does in fact or is reasonably likely in a specific case to have actual on-the-job effects.” (Emphasis supplied.)2 Employer contends that its evidence demonstrated *202a reasonable likelihood of on-the-job effects.
The difficulty with employer’s argument is that the only evidence to which it points is “claimant’s admitted amphetamine use before his work-shift, and subsequent positive drug test.” That evidence is no more probative of “reasonably likely” effects than of actual impairment. Neither claimant’s drug use the previous day nor the positive drug test reveal anything about actual or likely impairment or other drug effects at the time of the test or at any time that claimant was at work. We explained in Glide Lumber Prod. Co. v. Emp. Div. (Smith), supra, that the test used there could detect marijuana for approximately 30 days after its use, but could not reveal when the drug was ingested or what amount of it remained in the employee’s system. However, the record in Glide showed that the impairing effect of marijuana lasts for only 12 hours after ingestion. We therefore noted that that test could “come no closer than a 29-day range of identifying when the marijuana was pharmacologically active in any way which could bear on performance or safety.” 86 Or App at 674. It may be that some testing methods are capable of demonstrating more of a correlation; but that is a question of fact, *203and employer did not offer evidence that there was such a correlation here.
Employer also relies on PGE v. Employment Division, 95 Or App 647, 770 P2d 940 (1989), where we held that the claimant’s off-duty ingestion of alcohol was reasonably likely to have on-the-job effects, in the light of a urine test, administered three hours after the claimant reported for work, that revealed a blood alcohol content of .138. Employer analogizes the two cases, because the claimants in both tested positive “three hours into [their] work shift[s].” The point, however, is not when the tests were administered, but what they could show. We noted in PGE:
“Unlike the drug test used in Glide, however, the legislature has decided that chemical analysis of a person’s blood or urine provides a reasonably accurate measure of the level of intoxication present at or near the time of the test.” 95 Or App at 651 n 1.
Here, employer presented no such evidence about the capability of the test that it used to show actual drug influence or residual impairment at the time the test was administered.
Employer also relies on Silverton Forest Prod. Co. v. Emp. Div. (Arrant), 86 Or App 684, 741 P2d 915 (1987), to argue that the “impairing effects from amphetamines can last up to three days after they are taken” and, therefore, that claimant’s use of the substances the day before he was tested established a reasonable likelihood of on-the-job impairment. Employer does not read Silverton correctly. We said that the drugs were detectable by testing for three days after they were taken, not that their impairing effects could last that long. Contrary to employer’s argument, we said that
“impairment from cocaine use can last up to eight hours, and amphetamines may produce impairing effects which last for a somewhat longer period.” 86 Or App at 686.
In any event, our comments in Silverton were based on the evidence in the record there. Employer offered no evidence in this case to prove the duration of possible impairment from the drugs that claimant used. There was also no evidence that claimant’s drug use occurred within the approximate eight-hour period before he reported for work on the day that he was tested; claimant worked the day shift, and EAB found that he ingested the drugs “right after” his shift ended *204at 4:30 p.m. on the previous day. Silverton does not assist employer.
Employer next argues that the safety needs of claimant’s job were such that his misconduct was connected with his work, even in the absence of a showing of impairment. Employer relies on a 1986 memorandum from the assistant director to Division managers, that states, in part:
“In cases involving significant threats to public safety, it may be permissible to deny benefits for misconduct where the individual fails a drug test even though there is no evidence of impairment.”
Again, however, employer produced no evidence about the safety needs of claimant’s job. As the Division states in its brief, the record “is completely silent as to the nature of employer’s business, or even the nature of claimant’s job, beyond the [unembellished] fact that he was a ‘chipper operator.’ ”
For the same reasons, the dissent’s emphasis on safety concerns is misplaced. We might agree that employees in certain kinds of work, such as airline pilots or nuclear submarine commanders, cannot perform safely to a satisfactory probability if any intoxicating substances are in their systems. We would also agree, and Glide and our other cases have recognized, that the characteristics of some jobs, other than their safety requirements, might make off-duty drug use intrinsically connected with their performance. See note 4, infra. However, whether safety or other requirements make that true of particular jobs is a question of fact, and employer produced no evidence to demonstrate that it is true of claimant’s job.
Employer’s most strongly urged argument is that the last chance agreement, in which claimant agreed not to use drugs on or off the job, provides an independent basis for concluding that his misconduct was connected with his work, even if it would not have been in the absence of the agreement. We said in Glide Lumber Prod. Co. v. Emp. Div. (Smith), supra:
“An employer cannot reduce its statutory unemployment insurance responsibilities simply by promulgating an in-house rule. A rule may be relevant to the proof of facts bearing on whether off-duty misconduct is connected with work, but *205it cannot give rise to different substantive tests of work-connectedness than the ones expressed in the statute, the administrative rule and the cases construing them.” 86 Or App at 674.3
Employer argues that, unlike the employer’s rule in Glide, this case involves a mutual agreement rather than a unilaterally imposed job requirement. There may be little difference in that distinction. The prohibition of off-duty drug use was a condition of employment, and employer was free to impose it whether or not claimant agreed to it.
Be that as it may, an employer and an employee acting in concert have no more authority than does an employer acting alone to change the unemployment compensation law or to alter the statutory eligibility and disqualification criteria. Moreover, ORS 657.855 provides, as relevant, that “[n]o agreement by an individual to waive the individual’s rights under this chapter is valid.” The agreement between employer and claimant does not, by its terms, purport to affect claimant’s right to unemployment compensation, as opposed to establishing a requirement for continued employment. Nevertheless, employer’s argument would make the agreement an independent basis for disqualifying claimant from receiving benefits, in circumstances where he would not otherwise be disqualified under the statute. We do not agree that either party alone or both parties together may do what employer contends their agreement does.4
*206The dissent appears to agree that the parties’ agreement cannot alter the statutory eligibility standards and also that there was no evidence in this case aside from the agreement on which a finding of work-connectedness can be based. The dissent would nevertheless hold that the last chance agreement, together with the presumed fact that employer considered the job requirement important enough to insist that there be an agreement embodying it, can give rise to a finding of fact that claimant’s misconduct was connected with his work. We do not agree that the agreement can achieve in the guise of a fact the very thing that it is not permitted to do as a matter of law, i.e., make claimant’s misconduct work-connected and disqualify him from benefits when the statute does not.5
*207The dissent would also hold that Glide Lumber Prod. Co. v. Emp. Div. (Smith), supra, and Zufelt take too restrictive a view of what misconduct can be found to be work-connected, that they are therefore contrary to the policy of the statute, that Glide should be overruled and that Zufelt should be treated as invalid. We note, first, that the Division’s policy explication of the delegative statutory term, see note 2, supra, cannot be invalidated on the strength of the analysis that the dissent provides. That analysis comes to nothing more than a statement of disagreement with the responsible policymaker’s interpretation and application of the statute, and mere disagreement is not enough for us to replace the Division’s interpretation with our own. See Springfield Education Assn. v. School Dist., supra n 2.
More fundamentally, Glide and Zufelt are clearly consistent with the statute. We said in Giese v. Employment Div., 27 Or App 929, 557 P2d 1354 (1976), rev den (1977):
“So far as we can determine the term ‘misconduct connected with his work’ was inserted into our unemployment compensation law in 1937. Oregon Laws 1937, ch 398, § 4. When first enacted in 1935, our unemployment compensation statute simply provided that a claimant could be disqualified for ‘proven misconduct.’ Oregon Laws 1935 (Spec Sess), ch 70, § 4(b)(2).” 27 Or App at 933 n 2.
We then concluded
“that the phrase ‘connected with his work’ was added to our statute by the legislature to draw a distinction between misconduct while off-duty and misconduct in the course and scope of employment.” 27 Or App at 934.
*208Glide and Zufelt recognize and implement that fundamental distinction in the statute. They also recognize that off-duty drug use sometimes does have job-related effects and, when it does, it may be work-connected and disqualifying. However, Glide and our other cases make clear that, to be disqualifying, the relationship between a claimant’s off-duty drug use and his work must be established by proof, not by mere surmise and unsubstantiated assumptions of the kind that employer offers here.6
Affirmed.
The assistant director of the Division is the rulemaking and policymaking entity. However, for convenience, we refer to the Employment Division as the responsible agency. The Zufelt opinion was a contested case decision by the assistant director. Zufelt has the force of an administrative rule. See Trebesch v. Employment Division, 300 Or 264, 710 P2d 136 (1985); Redman Homes, Inc. v. Employment Div., 97 Or App 653, 777 P2d 414 (1989).
Both employer and the Division rely on Glide and later decisions by this court, as well as on Zufelt. “Misconduct connected with work” is a delegative statutory term. See Redman Homes, Inc. v. Employment Div., supra n 1; Oliver v. Employment Division, 40 Or App 487, 595 P2d 1252 (1979). Zufelt establishes the agency’s policy for the application of that term to discharges resulting from drug use and positive drug tests. To the extent that our earlier decisions interpreted and applied the term in that context, Zufelt may supersede them. However, because those decisions and Zufelt are consistent for purposes of the issues in this case, we will discuss the parties’ arguments according to their terms, including the case law on which they rely. See note 6, infra.
The dissent would hold that the term is inexact rather than delegative. We disagree. The labels “inexact” and “delegative” were devised in Springfield Education Assn. v. School Dist., 290 Or 217, 621 P2d 547 (1980). However, the court noted there that the concept of different kinds of statutory terms, with different accompanying levels of agency and judicial authority, originated in its earlier decision in McPherson v. Employment Division, 285 Or 541, 591 P2d 1381 (1979). Springfield traced the delegative term concept to the statement in McPherson that the term “good cause” for voluntarily leaving work in ORS 657.176(2)(c) “calls for completing a value judgment that the legislature itself has only indicated: evaluating what are ‘good’ reasons for giving up one’s employment and what are not.” 285 Or at 550; see 290 Or at 228. Shortly after McPherson was decided, and shortly before Springfield was, we said in Oliver v. Employment Division, supra, 40 Or App 491-92:
“McPherson specifically held that the ‘good cause’ language in [ORS 657.176(2)(c)] delegates a ‘value-judgment’ type of policy determination to the administrative agency. Given that holding, we think it follows that the other questions that can arise under ORS 657.176(2) — ‘misconduct’ in subparagraphs (2)(a) and (b), and ‘suitable work’ in subparagraphs (2)(d) and (e) — are likewise ‘value-judgment’policy questions for the agency.” (Emphasis supplied.)
See also Redman Homes, Inc. v. Employment Div., supra, n 1.
We adhere to Oliver. The Supreme Court expressly said in McPherson and Springfield that “good cause” is a delegative term that confers policymaking authority on the Division. The term “misconduct connected with work” is part of the same subsection of ORS 657.176 as “good cause,” and it plays an identical role in the statutory and administrative scheme. Each term provides a general standard for disqualification from unemployment benefits, and the agency must apply those stan*202dards in a great variety of factual contexts. In Springfield, the court explained that “good cause” was an example of the legislature’s use of a delegative term
“because it cannot foresee all the situations to which the legislation is to be applied and deems it operationally preferable to give an agency the authority, responsibility and discretion for refining and executing generally expressed legislative policy. * * * The task of the agency administering such a statute is to complete the general policy decision by specifically applying it at retail to various individual fact situations.” 290 Or at 228.
“Misconduct connected with work” fits as squarely within that description as “good cause” does. Moreover, neither the terms themselves nor anything in their context or purpose allows the conclusion that the legislature intended for the Division to have lesser policy responsibility in deciding eligibility questions that arise from discharges than those arising from a worker’s voluntarily leaving his employment. Like “good cause,” “misconduct connected with work” is a delegative term.
The dissent states, correctly, that the “majority reads the phrase ‘misconduct connected with work’ as a whole.” 105 Or App at 209. The dissent disagrees with that approach and appears to conclude that “misconduct” is a delegative term but “connected with work” is not. We adhere to the view that the phrase must be read and applied as a whole. The entire phrase constitutes a single statutory test for disqualification from benefits. It simply defies belief that the legislature could have intended to give the Division policy discretion in applying one word in that test, but not the others. The dissent posits that “misconduct” is a matter of “value judgment,” but “connected with work” entails only “factfinding.” We disagree. Both parts of the term require the making of general policy judgments as well as the finding of facts in specific cases.
Our statement that an employer’s “rule may be relevant to the proof of facts bearing on whether off-duty misconduct is connected with work” means that a rule may be relevant to defining the job and job requirements and, therefore, may have some bearing on whether particular misconduct is connected with the job; we clearly did not say that the employer could decide what misconduct is connected with an employee’s job for purposes of the Unemployment Compensation Law, as opposed to defining the job itself. This case does not present the question of whether a rule that proscribes drug use by employees whose jobs entail exceptional safety requirements could support a finding of work-connectedness under the statute or OAR 471-30-038(3).
We do not imply that the Division cannot, by rule, make last chance agreements germane to work-connectedness issues. However, at least as of the relevant time, neither the statute nor the regulatory provisions did so.
Employer also relies on Black v. Employment Div., 301 Or 221, 721 P2d 451 (1986), vacated and remanded sub nom Employment Division v. Smith, 485 US 660, 108 S Ct 1444, 99 L Ed 2d 753 (1988), and Smith v. Employment Div., 301 Or 209, 721 P2d 445 (1986), vacated and remanded sub nom Employment Division v. Smith, 485 US 660, 108 S Ct 1444, 99 L Ed 2d 753 (1988). See also Smith v. Emp. Division, 310 Or 376, 799 P2d 148 (1990). We rejected the employer’s similar argument in Glide Lumber *206Prod. Co. v. Emp. Div. (Smith), supra, and explained:
“The claimants in those cases were discharged from their employment in drug counseling programs after they ingested certain drugs in connection with religious ceremonies. * * *
“Employer reasons that the court necessarily decided that Black’s and Smith’s off-duty drug use was work-connected, because it would not otherwise have reached the constitutional question. However, even assuming employer’s premise, we agree with the Division:
“ ‘Both cases are easily distinguishable from this case. Smith and Black involved drug counselors who, before they were hired, agreed not to use mind-altering drugs. The employer, a drug treatment center, hired recovered alcoholics and drug abusers whom it viewed as role models for its clients. Smith, 301 Or at 211; Black, 301 Or at 223. The use of drugs, even off-duty, destroyed the employe’s value as a role model for the employer’s clients.’ ” 86 Or App at 676 n 2.
See also City of Portland v. Employment Division, 94 Or App 279,765 P2d 222 (1988).
Finally, employer relies on PGE v. Employment Division, supra, where we concluded that the claimant’s violation of a last chance agreement showed that he was aware of the employer’s expectation that he stay sober, and that the misconduct was wilful. See OAR 471-30-038(3). However, we made it clear in that case that work-connectedness and wilfulness are separate questions, and that the misconduct had to be work-connected as well as wilful to be disqualifying. 95 Or App at 650. Our conclusion that the claimant’s misconduct was work-connected had nothing to do with the agreement, but was based on the facts that he was legally intoxicated and “manifested other physical symptoms of intoxication” while on the job. 95 Or App at 651.
It is not clear whether the dissent regards the importance that employer attaches to the agreement as having evidentiary significance independent of that of the agreement itself. If it does, we disagree. The only reason for concluding that employer regards the job requirement in the agreement as important is that employer demanded that there be an agreement to establish the requirement. Employer’s view of the importance of the agreement adds nothing to the fact that there is an agreement. Moreover, employer’s subjective belief in the importance of the requirement cannot demonstrate work-connectedness in itself, in the absence of evidence about safety or other job-related matters to show that there might be a factual basis for employer’s *207belief.
The dissent also suggests that claimant’s “misconduct” consisted of working with the substances in his body, rather than the off-duty act of taking the substances. That suggestion begs the question (or, more accurately, evades it). In order for the act of taking the drugs or the fact of having them in one’s body to be misconduct connected with work, within the meaning of the statute, there must be some effect on the employee’s work. The dissent attempts to define a consequence of the misconduct as being the misconduct. Even if that reasoning could otherwise succeed, it surely cannot in light of the dissent’s inability to show that the consequence has any more of a connection with work than the underlying misconduct itself does. However, the reasoning also fails for a more basic reason. The plain and inescapable fact is that the misconduct for which claimant was fired was his off-duty drug use'. Employer’s and the dissent’s attempts to recharacterize and avoid that fact are word games.
It is unnecessary for us to decide whether some or all of the facts that employer assumes could, if proven, have permitted claimant’s disqualification under Glide or Zufelt. Arguably, Zufelt is more restrictive than Glide, because it does not expressly make a reasonable likelihood of work-related effects, as well as actual effects, a basis for finding work-connectedness. That difference is of no consequence in this case, however, because employer offered no evidence to prove either of the alternatives.