AGO 2016 No. 1 - Jan 28 2016
ELECTIONS—CITIES AND TOWNS—FEDERAL PREEMPTION—Authority Of Code Cites To Modify System For Electing City Council Members To Comply With Federal Voting Rights Act
1.State law allows a code city to divide into wards for use at the primary for nominating candidates for the city council but requires that all city council members be elected at large at the general election.
2.A code city may choose to use wards for both nominating candidates in the primary and for electing city council members at the general election if, but only if, the city has a strong basis in evidence for concluding that Section 2 of the federal Voting Rights Act compels the city to do so.
January 28, 2016
The Honorable Pam Roach
Dear Senator Roach:
By letter previously acknowledged, you have requested our opinion on the following question:
May a non-charter code city subject to RCW 35A.12.180, having a large minority population, adopt a district-based general election procedure to avoid a potential violation of section 2 of the federal Voting Rights Act?
Yes. RCW 35A.12.180 allows certain cities to use districts for conducting city council primary elections but requires a citywide vote at the general election. We conclude that a city subject to RCW 35A.12.180 may both nominate and elect positions on its city council by district, but only if the city has a strong basis in evidence to conclude that its proposed change is necessary to comply with the federal Voting Rights Act (VRA). 52 U.S.C. §§ 10301-10314.
[original page 2]
Your question arises out of a situation in the City of Pasco. Pasco has opted pursuant to state law to divide itself into five wards (or districts), which are used in electing five of the city’s seven positions on its city council. Candidates for the five positions nominated by ward must be residents of that ward, and only the voters in that ward may vote for the position at the primary. The top two candidates at the primary then advance to the general election, which is conducted city-wide with participation from all voters throughout the city. RCW 35A.12.180. Candidates for those five positions are thus nominated by ward, but elected at large. The other two positions are both nominated and elected by the entire city at large. RCW 35A.12.040.
Pasco is considering changing this voting system because of concern that its current system may violate the VRA. This concern arises for a number of reasons, but in large part because Pasco is home to a substantial Hispanic population and the nearby city of Yakima recently saw its electoral system rejected as violating the VRA. Montes v. City of Yakima, 40 F. Supp. 3d 1377 (E.D. Wash. 2014), appeal docketed, No. 15-35309 (9th Cir. Apr. 21, 2015). In order to avoid similar litigation, Pasco proposed to amend its municipal code to provide for both the primary and general elections to be conducted by ward. This proposal would entail both nominating and electing five positions on the city council by ward, rather than the positions being nominated by ward and elected at large. The other two positions would continue to be both nominated and elected at large.
The city submitted this proposal to the county auditor, who serves as the ex officio election officer for the county and for the cities within the county. RCW 29A.04.216, .330; RCW 36.22.220. The auditor responded that state law did not authorize the auditor to conduct the election in this changed manner on behalf of the city. Letter from Matt Beaton, Franklin County Auditor, to Dave Zabell, Pasco City Manager (Apr. 17, 2015). The city council then adopted an ordinance that expresses its desire to change its electoral system, but retains its current system due to the question of its authority to make the change. Pasco Ordinance No. 4218, https://egov-pasco.com/weblink8/0/doc/541388/Page1.aspx (last visited Jan. 6, 2016). You pose your question to clarify the city’s authority to determine the manner in which its city council elections are conducted.
[original page 3]
Our analysis consists of three parts. First we consider the scope of your question, noting several topics we cannot explore in an Attorney General Opinion. Second, we consider how your question would be analyzed under state law if state law were the only relevant consideration. We conclude that considering state law in isolation, a code city is generally required to conduct general elections for city council at large throughout the city, and therefore a code city is generally prohibited from limiting general elections to voters within each ward. Third, we turn to the effect of federal law on our state law analysis. The third section is the most complex and gives rise to more than one possible interpretation of the law. Although there is no case law directly on point, we conclude that a code city would be justified in conducting its city council elections in a way that is otherwise inconsistent with state law if, but only if, the code city can establish a strong basis in evidence for concluding that federal law compels its action. If its electoral process were challenged in court, we believe the code city would bear the burden of proving such a strong basis in evidence. We believe that a party challenging the city’s decision could then prevail only by showing that federal law did not, in fact, require the city to deviate from state election law.
A. Scope Of Analysis
Before beginning our analysis, we pause to explain what we are and are not able to address in this opinion.
First, although your question arises in the context of a specific city, our analysis will necessarily be general. Attorney General Opinions do not resolve factual issues, and as we explain below, any conclusion that a particular city might be violating the VRA would be deeply fact dependent. Thus, we offer no opinion as to whether Pasco or any other city is, or is not, currently violating the VRA.
Second, our opinion expresses no view as to whether Pasco or any other city should elect its city council in any particular way. Instead, it simply analyzes what options are available to code cities. Our formal opinions provide analysis of legal issues, but are not a forum for advocating public policy.
Finally, while this opinion highlights a very difficult question created by current state law as to code cities facing potential VRA claims, it is worth noting that this difficulty could easily be fixed by the legislature. If the legislature allowed code cities to conduct general elections by district, cities like Pasco would not face the dilemma and uncertainty described here.
With these parameters in mind, we turn to the legal analysis of your question.
[original page 4]
B. Authority Of A Code City Under State Law To Determine The Manner By Which Its City Council Is Elected
You ask specifically about the authority of a code city, such as Pasco, to both nominate and elect its city council by ward. The term “code city” refers to a city that has chosen to organize under the Optional Municipal Code. RCW 35A.01.035. The legislative body of a code city has the “power to organize and regulate its internal affairs within the provisions of this title and its charter[.]” RCW 35A.11.020. “The legislative body of each code city shall have all powers possible for a city or town to have under the Constitution of this state, and not specifically denied to code cities by law.” RCW 35A.11.020. The question under state law therefore becomes whether adopting an electoral structure under which members of a city council are both nominated and elected by ward is specifically denied to code cities by law. RCW 35A.11.020.
The ways in which code cities may elect city council members are set out in general law. “Elections to positions on the council shall be by majority vote from the city at large, unless provision is made by charter or ordinance for election by wards.” RCW 35A.12.040. State law therefore specifies the default method for electing city council members in code cities. Such elections are held at large, unless the city has opted for a different system either by charter or through an ordinance providing for election by wards.
The option to elect city council members by ward is also described in statute:
At any time not within three months previous to a municipal general election the council of a noncharter code city organized under this chapter may divide the city into wards or change the boundaries of existing wards. No change in the boundaries of wards shall affect the term of any councilmember, and councilmembers shall serve out their terms in the wards of their residences at the
[original page 5]
time of their elections: PROVIDED, That if this results in one ward being represented by more councilmembers than the number to which it is entitled those having the shortest unexpired terms shall be assigned by the council to wards where there is a vacancy, and the councilmembers so assigned shall be deemed to be residents of the wards to which they are assigned for purposes of those positions being vacant. The representation of each ward in the city council shall be in proportion to the population as nearly as is practicable.
Wards shall be redrawn as provided in chapter 29A.76 RCW. Wards shall be used as follows: (1) Only a resident of the ward may be a candidate for, or hold office as, a councilmember of the ward; and (2) only voters of the ward may vote at a primary to nominate candidates for a councilmember of the ward. Voters of the entire city may vote at the general election to elect a councilmember of a ward, unless the city had prior to January 1, 1994, limited the voting in the general election for any or all council positions to only voters residing within the ward associated with the council positions. If a city had so limited the voting in the general election to only voters residing within the ward, then the city shall be authorized to continue to do so.
RCW 35A.12.180 (emphases added). Thus, a code city is authorized to divide the city into wards of equal population, “as nearly as is practicable.” RCW 35A.12.180. If the city chooses to do so, then candidates for each position must live in the corresponding ward, and only voters of that ward would be eligible to vote in the primary. But at the general election, the statute specifies that “[v]oters of the entire city may vote.” RCW 35A.12.180. Thus, RCW 35A.12.180 specifically denies to code cities the authority to restrict voting by ward at the general election. Therefore, a local ordinance that provided for general elections by ward would conflict with RCW 35A.12.180 and be preempted by state law. See Cannabis Action Coal. v. City of Kent, 183 Wn.2d 219, 225-26, 351 P.3d 151 (2015) (ordinances that conflict with state law are preempted).
State law on its own therefore does not allow a code city to provide for the election of its city council members by ward. Wards may be used to nominate candidates at the primary, but state law requires that all voters of a code city be permitted to vote in each city council race at the general election. RCW 35A.12.180. We therefore turn to the question of whether a code city might be compelled by federal law to elect its council members differently than as provided by RCW 35A.12.180.
[original page 6]
C. A Code City May Adopt An Ordinance Providing For Both Nomination And Election Of City Council Members By Ward If Compelled To Do So By Federal Law
Having concluded that state law requires that all voters of code cities be permitted to vote in each city council race at the general election, we next consider whether a city may deviate from this rule to avoid a violation of Section 2 of the VRA, 52 U.S.C. § 10301(b). We conclude that a city may conduct general elections by ward when it has a strong basis in evidence to conclude that acting otherwise would violate Section 2. Determining whether Section 2 requires changes to local election processes is a highly fact-specific analysis that local jurisdictions must undertake on a case-by-case basis and that would be subject to judicial review.
Section 2 prohibits any state or political subdivision of a state from using any standards, practices, or procedures that deny or abridge the right of citizens to vote on account of their race, color, or membership in a language minority group. 52 U.S.C. § 10301 (formerly codified as 42 U.S.C. § 1973). “A Section 2 violation is any political process leading to an election that is not ‘equally open to participation’ by a minority group ‘in that its members have less opportunity than other members of the electorate to participate in the political process and to elect representatives of their choice.’” Ruiz v. City of Santa Maria, 160 F.3d 543, 549 (9th Cir. 1998) (quoting former 42 U.S.C. § 1973(b)). “Section 2 ‘requires proof only of a discriminatory result, not of discriminatory intent.’” Id. (quoting Smith v. Salt River Project Agric. Improvement & Power Dist., 109 F.3d 586, 594 (9th Cir. 1997)).
The United States Constitution establishes the supremacy of federal law within the ambit of its authority. U.S. Const. art. VI, cl. 2. “When the Federal Government acts within the authority it possesses under the Constitution, it is empowered to pre-empt state laws to the extent it is believed that such action is necessary to achieve its purposes.” City of New York v. FCC, 486 U.S. 57, 63 (1988). Thus, state laws that directly conflict with federal law yield to federal law. City of Tacoma v. Taxpayers of Tacoma, 43 Wn.2d 468, 483, 262 P.2d 214 (1953).
Where state and federal laws conflict, the Washington Supreme Court holds that a conflicting state law cannot stand as an obstacle to compliance with federal law as determined by a federal court order. Puget Sound Gillnetters Ass’n v. Moos, 92 Wn.2d 939, 951, 603 P.2d 819 (1979) (state law could not “survive the superior force of federal law” after the United States Supreme Court had construed a tribal treaty governing fishing rights). That is, once a violation of federal law is found, it is not necessary to await the enactment of a conforming state law for a state agency to comply with federal law. Id.
The circumstance you ask about differs from Puget Sound Gillnetters Association in a significant way. In that case, the courts had already reached a conclusion as to the application of federal law. Puget Sound Gillnetters Ass’n, 92 Wn.2d at 951. In contrast, you ask about a
[original page 7]
scenario in which there has been no litigation but a code city desires to change its voting system in a way precluded by state law in order to avoid Section 2 liability. We therefore consider whether federal law might compel a city’s action even without a judicial finding.
Federal cases under the VRA emphasize that the VRA will only compel a violation of an otherwise-controlling state law if that violation is necessary to avoid violating Section 2. This is because, as the Tenth Circuit explained, the remedy for a VRA violation must “adhere as closely as possible to the contours of the governing state law.” Large v. Fremont County, 670 F.3d 1133, 1135 (10th Cir. 2012). The court emphasized that in forming a remedy for a VRA violation, the county was “not free to disregard state law.” Id. at 1137. Concluding that federal courts should defer to state law (and not to local legislative bodies), the court declined to give deference to the county’s plan that did not hew as closely as possible to state law. Id. at 1142, 1146. The court reasoned that:
[T]he mere fact that some state laws may necessarily need to be displaced to permit the effectuation of a federal civil-rights remedy under Section 2 does not mean that local governmental bodies like the County may unnecessarily—as a matter of preference—disregard the dictates of state law in fashioning their plans and still claim the judicial deference for their handiwork that is traditionally accorded to legislative plans.
Id. at 1144. State law can be abrogated when it is necessary to do so, but only if the law is an “unavoidable obstacle” to compliance with the federal law. Id. at 1145.
“The same cannot be said where in the course of remedying an adjudged Section 2 violation a local governmental entity gratuitously disregards state laws—laws that need not be disturbed to cure the Section 2 violation.” Id. (citing Cleveland County Ass’n for Gov’t by the People v. Cleveland County Bd. of Comm’rs, 142 F.3d 468, 477 (D.C. Cir. 1998)). As the District of Columbia Circuit expressed the matter, “if a violation of federal law necessitates a remedy barred by state law, the state law must give way; if no such violation exists, principles of federalism dictate that state law governs.” Cleveland County Ass’n for Gov’t by the People, 142 F.3d at 477. As the Seventh Circuit put it, parties to litigation cannot use that litigation simply as a way of evading valid state law, and cannot simply agree to do something they would otherwise lack the authority to do. Perkins v. City of Chicago Heights, 47 F.3d 212, 216 (7th Cir. 1995). Though a local government “may chafe at [state law] restraints and seek to evade them . . . they
[original page 8]
may not do so by agreeing to do something state law forbids.” Perkins, 47 F.3d at 216 (citation omitted) (internal quotation marks omitted).
Even when a lawsuit under the VRA has been commenced, a city is not free to simply ignore state law in settling that suit except where it is necessary to do so in order to comply with federal law. League of Residential Neigh. Advocates v. City of Los Angeles, 498 F.3d 1052, 1055 (9th Cir. 2007). As the Ninth Circuit put it, “[a] federal consent decree or settlement agreement cannot be a means for state officials to evade state law.” Id. A local government may therefore violate state law if—but only if—it is necessary to do so in order to remedy a violation of federal law. Id.; Large, 670 F.3d at 1145; Cleveland County Ass’n for Gov’t by the People, 142 F.3d at 477; Perkins, 47 F.3d at 216. Adopting a remedy for violation of federal law that, in turn, violates state law “is authorized only when the federal law in question mandates the remedy[.]” St. Charles Tower, Inc. v. Kurtz, 643 F.3d 264, 271 (8th Cir. 2011) (quoting League of Residential Neigh. Advocates, 498 F.3d at 1058).
The cases discussed in the paragraphs above might be read in two ways. One way is to conclude that a state or local jurisdiction is empowered to ignore state law only if a court has found a violation of state law. In the context of reviewing a consent decree approved by a district court to settle a VRA dispute, the Seventh Circuit explained: “Once a court has found a federal constitutional or statutory violation . . . a state law cannot prevent a necessary remedy.” Perkins, 47 F.3d at 216 (emphasis added). That is, state law can be set aside “upon properly supported findings that such a remedy is necessary to rectify a violation of federal law[.]” Id. (first emphasis added). So one possible conclusion is that a code city may only disregard the prohibition in RCW 35A.12.180 against using wards at general elections (as opposed to only at primaries) if a court first rules that the city has violated Section 2 and imposes such an electoral system as a remedy.
An obvious problem with that reading of the cases is that it precludes resolving a known problem before being sued. This would both perpetuate a violation of a federal civil rights law and incur the expense that comes with litigation under the VRA. Of course, the legislature could step in with a statutory remedy that expands the authority of code cities to determine their electoral processes. But absent that, reading current law to forbid an action until a violation of federal law is judicially determined “would bring compliance efforts to a near standstill.” Ricci v. DeStefano, 557 U.S. 557, 581, 129 S. Ct. 2658, 174 L. Ed. 2d 490 (2009); see also, e.g., Lindsay v. City of Seattle, 86 Wn.2d 698, 706, 548 P.2d 320, cert. denied sub nom. Brabant v. City of Seattle, 429 U.S. 886 (1976) (holding that “voluntary compliance, rather than court ordered relief,” is the preferred method of complying with civil rights statutes).
We believe that the Ricci case just cited provides helpful guidance as to how to proceed. Ricci arose under federal civil rights laws governing employment. A city discarded the results of a promotional test for captains and lieutenants in its fire department based on allegations that the test results showed that the test was discriminatory. White and Hispanic firefighters, who likely would have been promoted based on the examination results, sued alleging disparate treatment based on race. The city defended based on the argument that using the exam results would have
[original page 9]
resulted in disparate-impact liability (employers may not use metrics that have racially disparate impacts unless those metrics are job-related and consistent with business necessity). Ricci, 557 U.S. at 562-63. The Court noted that without some justification, the city’s action in discarding the test results for expressly race-based reasons would violate the federal prohibition against adverse employment actions based on race. Id. at 579. The Court therefore faced the question of “whether the purpose to avoid disparate-impact liability excuses what otherwise would be prohibited disparate-treatment discrimination.” Id. at 580. The Court concluded that the city’s action in discarding its examination was justified if, but only if, the city had “a strong basis in evidence” for concluding that its action was necessary to avoid liability for disparate impact based on race. Id. 582 (emphasis added). That is, the city in Ricci could defend its disparate treatment based on race, in violation of one federal law, if it had a strong basis in evidence that its action was necessary to avoid violating another law. Id. at 582-83.
We are not aware of any Washington court that has addressed the question of what level of evidence is required before a code city may disregard state law in order to comply with federal law. We believe that a Washington court would apply a standard similar to the “strong basis in evidence” standard applied in Ricci, a standard the U.S. Supreme Court has also applied in a range of voting rights cases. See, e.g., Abrams v. Johnson, 521 U.S. 74, 91, 117 S. Ct. 1925, 138 L. Ed. 2d 285 (1997) (citing Shaw v. Reno, 509 U.S. 630, 656, 113 S. Ct. 2816, 125 L. Ed. 2d 511 (1993)). As applied here, the “strong basis in evidence” standard would leave room for a code city to voluntarily comply with Section 2 when it has a strong basis in evidence for believing that it is necessary to act contrary to state law in order to comply with federal law. See Ricci, 557 U.S. at 583. Under this approach, a code city would be justified in electing members to its city council by ward at both the primary and general election, in violation of RCW 35A.12.180, if—but only if—it had a strong basis in evidence for concluding that this action was compelled by federal law, in order to avoid violating Section 2. This would be so even if the code city was not sued first.
But to develop this strong basis in evidence it would be necessary to develop sufficient facts to provide a basis for concluding that Section 2 compels the city’s action. The question of whether any particular voting practice places a specific jurisdiction in violation of the VRA involves an inquiry into the totality of the factual circumstances and the application of federal law to those facts. See, e.g., Farrakhan v. Gregoire, 623 F.3d 990, 992-93 (9th Cir. 2010) (en banc) (discussing the application of the VRA to Washington’s laws regarding disenfranchisement of convicted felons). Satisfying the “strong basis in evidence” standard therefore necessarily involves developing a factual basis for concluding that federal law compels the course of action the city takes. See Margerum v. City of Buffalo, 24 N.Y.3d 721, 731-32, 28 N.E.3d 515, 5 N.Y.S.3d 336 (2015) (applying Ricci). A city need not be certain that its current
[original page 10]
system violates federal law to have a strong basis in evidence, but it must “have good reasons to believe” that its current system violates federal law to meet the strong basis in evidence standard. Alabama Legislative Black Caucus v. Alabama, 135 S. Ct. 1257, 1274, 191 L. Ed. 2d 314 (2015).
Of course, a code city that chose to switch to a district election system for its general election could potentially be sued for violating state law. In such a lawsuit, we believe state courts might well adopt a burden shifting approach like the one adopted by courts applying Ricci in the employment context. See, e.g., id.; Maraschiello v. City of Buffalo Police Dep’t, 709 F.3d 87, 93-94 (2d Cir.) cert. denied, 134 S. Ct. 119 (2013); United States v. Brennan, 650 F.3d 65, 93-94 (2d Cir. 2011). If the plaintiffs could show that the city’s voting system deviated from state law, then the burden would shift to the city to show that it adopted that deviation because of a strong basis in evidence to conclude that doing otherwise would violate Section 2. If the city could not make such a showing, its deviation from state law would be struck down. If it could make such a showing, the burden would shift to the plaintiff to prove that, in fact, although the city had a strong basis in evidence, there ultimately was no Section 2 violation.
In sum, code cities in Washington that believe they may be in violation of the VRA face difficult decisions and potential legal risk regardless of what course they choose. The legislature could rectify this situation by giving code cities greater authority to structure their general election processes. In the meantime, code cities do have some discretion. If they develop a strong basis in evidence to believe that they are violating Section 2, we believe that they would be justified in deviating from state law to comply with federal law. Such a conclusion would be highly dependent on the facts in any city and would require substantial research and factual support. Code cities will need to weigh their individual factual circumstances and the legal risks in making their decisions. Either course of action, whether to adhere to state law or to depart from it, may be subject to challenge in court.
We trust that the foregoing will be useful to you.
ROBERT W. FERGUSON
JEFFREY T. EVEN
Deputy Solicitor General
 During the time we have been working on this opinion, the United States Court of Appeals for the Ninth Circuit issued an opinion holding that a somewhat similar voting system used by the City of Tucson, Arizona, is unconstitutional for reasons unrelated to the Voting Rights Act or to any of the issues considered in this opinion. Pub. Integrity All., Inc. v. City of Tucson, 805 F.3d 876 (9th Cir. 2015). The system at issue in that case is not entirely the same as the system currently used in Pasco, and the issues considered in that case are distinctly different than the topic of your question. That opinion thus has no direct effect on our analysis of your question.
 Pasco’s Hispanic population is estimated at 55.3 percent of the city’s population as of 2013. U.S. Census Bureau, American Fact Finder, data for City of Pasco, Washington, http://factfinder.census.gov/faces/nav/jsf/ pages/community_facts.xhtml (last visited Jan. 6, 2016).
 Your question concerns the authority of a code city, rather than the authority of a county auditor. This opinion therefore does not address the question of whether the county auditor acted within his authority.
 This is only one of several ways in which cities can organize. Other forms of city organization include first class cities, second class cities, and towns. RCW 35.01.010 (first class cities, also known as charter cities); RCW 35.01.020 (second class cities); RCW 35.01.040 (towns). The internal organization of each of these types of cities can differ from that of code cities. See RCW 35.22 (first class cities); RCW 35.23 (second class cities); RCW 35.27 (towns). This opinion does not consider how your question might be answered if it had been posed with regard to any of these other types of city.
 Code cities have the option of organizing under their own charters, thus becoming charter code cities. RCW 35A.08.010. This opinion does not address charter code cites, whose internal organization is governed by their charters.
 The state constitution grants cities broad powers to “make and enforce within its limits all such local police, sanitary and other regulations as are not in conflict with general laws.” Wash. Const. art. XI, § 11. The optional municipal code reflects this broad authority, but limits it to powers “not specifically denied to code cites by law.” RCW 35A.11.020. What might otherwise be a constitutional analysis accordingly collapses into consideration of whether a particular power is specifically denied by statute. See, e.g., Housing Auth. of the City of Pasco & Franklin County v. City of Pasco, 120 Wn. App. 839, 844-45, 86 P.3d 1217 (2004); see also Hugh Spitzer, “Home Rule” vs. “Dillon’s Rule” for Washington Cities, 38 Seattle U. L. Rev. 809, 841-42 (2015) (discussing the scope of authority of code cities).
 RCW 35A.12.180 appears in the statutory chapter relating to code cities using the mayor-council plan of government. Code cities opting for the council-manager form of government are also subject to the same provisions for electing council members. RCW 35A.13.020 (incorporating by reference RCW 35A.12.040, which in turn incorporates RCW 35A.12.180 by providing for the option of using wards).
 In two cases the United States Supreme Court has considered efforts by a state to use Section 2 as a defense against allegations that the state had violated state law in drawing legislative or congressional districts. Bartlett v. Strickland, 556 U.S. 1, 6-7, 129 S. Ct. 1231, 173 L. Ed. 2d 173 (2009) (legislative district plan divided counties in contradiction of North Carolina Constitution); Shaw v. Hunt, 517 U.S. 899, 915-16, 116 S. Ct. 1894, 135 L. Ed. 2d 207 (1996) (state relied upon a “strong basis in evidence” that consideration of race in drawing congressional district boundaries was necessary to avoid a Section 2 violation). The Court ultimately resolved both cases in ways that shed little light on your question. They provide only minimal support to our analysis by illustrating that some jurisdictions have used Section 2 in an attempt to justify an action.
 This is not a task to take lightly. The factual inquiry involved can be complex, requiring consideration of several threshold factors, followed by an inquiry into the totality of the circumstances applicable to elections in the jurisdiction. Thornburg v. Gingles, 478 U.S. 30, 50, 106 S. Ct. 2752, 92 L. Ed. 2d 25 (1986) (reciting threshold factors); Johnson v. De Grandy, 512 U.S. 997, 1011-12, 114 S. Ct. 2647, 129 L. Ed. 2d 775 (1994) (describing the totality of the circumstances analysis).