Matthews v. Barr

16‐3145
Matthews v. Barr

UNITED STATES COURT OF APPEALS
FOR THE SECOND CIRCUIT

August Term, 2017

(Argued: May 9, 2018 Decided: June 18, 2019)

Docket No. 16‐3145
________________________________________________________________________

GERARD PATRICK MATTHEWS,

Petitioner,

‐ v. ‐

WILLIAM P. BARR, United States Attorney General,

Respondent.
________________________________________________________________________

Before:
HALL and CARNEY, Circuit Judges, KOELTL, District Judge.1

Petitioner seeks review of a decision of the Board of Immigration Appeals,
affirming a decision of an Immigration Judge that found Petitioner removable and
denied Petitioner’s applications for cancellation of removal and adjustment of
status. The agency determined that Petitioner’s New York convictions for
endangering the welfare of a child were crimes of child abuse, child neglect, or
child abandonment under the Immigration and Nationality Act, and denied relief
from removal as a matter of discretion. Because we have previously upheld the
determination of the Board of Immigration Appeals that child endangerment
offenses fall within the Immigration and Nationality Act’s crime of child abuse
provision, and we agree that New York law—which criminalizes conduct that

1Judge John G. Koeltl, of the United States District Court for the Southern District of New York,
sitting by designation.
poses a likelihood of physical, mental, or moral harm to a child—falls within the
BIA’s definition, we DENY the petition.

Petition for review DENIED.

Judge Carney dissents in a separate opinion.

DAVID J. ZIMMER, Goodwin Procter LLP, Boston,
MA (William M. Jay, Goodwin Procter LLP,
Washington, D.C.; Seymour W. James, Jr.,
Attorney‐in‐Chief, Hasan Shafiqullah, Attorney‐in
Charge, Ward J. Oliver, Supervising Attorney,
Immigration Law Unit, The Legal Aid Society,
New York, NY, on the brief), for Petitioner.

SONG PARK, Senior Litigation Counsel, Office of
Immigration Litigation, Civil Division (Chad A.
Readler, Acting Assistant Attorney General, Cindy
S. Ferrier, Assistant Director, on the brief), United
States Department of Justice, Washington, D.C., for
Respondent.

ANDREW WACHTENHEIM, Immigrant Defense
Project, New York, NY, for Amici Curiae Brooklyn
Defender Services, Queens Law Associates,
Neighborhood Defender Service of Harlem, The
Bronx Defenders, Essex County Public Defender’s
Office, Monroe County Public Defender’s Office,
Immigrant Defense Project.

2
HALL, CIRCUIT JUDGE:

Petitioner Gerard Patrick Matthews seeks review of an August 2016 decision

of the Board of Immigration Appeals (“BIA”), affirming an April 2016 decision of

an Immigration Judge (“IJ”) that ordered Matthews removed to Ireland on the

ground that his convictions for endangering the welfare of a child under New York

Penal Law (“NYPL”) § 260.10(1) made him removable under Immigration and

Nationality Act (“INA”) § 237(a)(2)(E)(i), 8 U.S.C. § 1227(a)(2)(E)(i), which

provides that “[a]ny alien who at any time after admission is convicted of . . . a

crime of child abuse, child neglect, or child abandonment is deportable.”

Matthews argues that he is not removable because the New York statute of

conviction is not categorically a crime of child abuse under the INA. First, he

argues that the INA’s definition of crime of child abuse requires that there be

actual harm to a child, or at least a high risk of serious injury. He challenges the

BIA’s definition, which encompasses state child endangerment offenses that

involve a sufficiently high risk of physical, mental, or moral harm to a child.

Because we have deferred to the BIA’s definition as a reasonable interpretation of

the crime of child abuse provision in a precedential decision, Florez v. Holder, 779

F.3d 207 (2d Cir. 2015), and we find no basis for departing from that decision, we

3
reject Matthews’s first argument and continue to apply the BIA’s definition of

crime of child abuse. Second, Matthews argues that the New York child

endangerment law, NYPL § 260.10(1), stretches even further than the BIA’s broad

definition because the statute and New York courts’ interpretation of it allow for

convictions based on conduct that poses only a minimal risk of nonserious harm

to a child. We reject this argument as well: NYPL § 260.10(1)’s text and New York

court decisions applying the statute require proof of a knowing mental state and a

likelihood of harm, and Matthews has not shown a realistic probability of

conviction for conduct that does not pose a likelihood of harm.

We DENY the petition for review.

BACKGROUND

Petitioner Gerard Patrick Matthews is a native and citizen of Ireland who

has lived in the United States as a lawful permanent resident since 1989.

Matthews, who was physically and sexually abused as a child, has a long history

of alcoholism and has repeatedly exposed himself in public while intoxicated.

Between 1990 and 2011, these public exposure incidents resulted in at least nine

convictions for public lewdness and two convictions for endangering the welfare

4
of a child under New York Penal Law (“NYPL”) § 260.10(1).2 Section § 260.10(1)

prohibits, as relevant here, “knowingly act[ing] in a manner likely to be injurious

to the physical, mental or moral welfare of a child less than seventeen years old.”

Matthews was convicted in 2002 of violating NYPL § 260.10(1) and was sentenced

to six months’ imprisonment. In 2003, he pleaded guilty to another NYPL

§ 260.10(1) violation as well as a public lewdness charge under NYPL § 245.00 and

was sentenced to concurrent terms of imprisonment of one year and 90 days,

respectively. Matthews testified at his removal hearing that he never intentionally

targeted children when he engaged in publicly lewd behavior; rather, he targeted

adults, and the child endangerment charges were tacked onto his public lewdness

charges because children happened to be present. However, the IJ did not credit

Matthews’s accounts of the circumstances of these convictions because they were

in stark contrast to the criminal complaints, which alleged that he targeted a nine‐

year‐old boy in one case and two teenage girls in the other.

In 2011, the Department of Homeland Security placed Matthews in removal

proceedings through service of a Notice to Appear (“NTA”). The NTA charged

Matthews as removable on the ground that his New York convictions for

2Matthews also has a 1995 conviction for third‐degree assault under NYPL § 120.00, for which he
was sentenced to three years’ probation. This conviction was not charged as a basis for removal.

5
endangering the welfare of a child were crimes of child abuse, child neglect, or

child abandonment under 8 U.S.C. § 1227(a)(2)(E)(i). An IJ initially granted

Matthews discretionary relief from removal, but the BIA overturned that ruling in

2013. Matthews petitioned for review, and we remanded the case for the agency

to explain fully its denial of relief. Matthews v. Holder, 590 F. App’x 75 (2d Cir.

2015). We declined to consider Matthews’s then‐unexhausted argument that his

convictions for endangering the welfare of a child were not removable offenses

(i.e., crimes of child abuse, child neglect, or child abandonment), but observed that

Matthews could renew this challenge before the agency on remand. Id. at 77.

On remand, the Government added an additional charge that Matthews

was removable for having been convicted of two crimes involving moral turpitude

(“CIMTs”) based on the public lewdness convictions in 1990 and 1994. See INA

§ 237(a)(2)(A)(ii), 8 U.S.C. § 1227(a)(2)(A)(ii) (“Any alien who at any time after

admission is convicted of two or more crimes involving moral turpitude, not

arising out of a single scheme of criminal misconduct . . . is deportable.”).

In April 2016, the IJ denied all relief and ordered Matthews removed. No.

A042 231 142 (Immig. Ct. N.Y. City Apr. 7, 2016). The IJ found that Matthews’s

public lewdness convictions were not CIMTs, but concluded that Matthews was

6
removable for crimes of child abuse, neglect, or abandonment based on his

convictions under NYPL § 260.10(1). The IJ relied on the BIA’s precedential

decision in Matter of Mendoza Osorio, 26 I. & N. Dec. 703 (B.I.A. 2016), which held

that a conviction for endangering the welfare of a child under NYPL § 260.10(1) is

a crime of child abuse, neglect, or abandonment under the INA. The IJ denied

both adjustment of status and cancellation of removal as a matter of discretion,

reasoning that Matthews’s “several positive factors” (over twenty‐five years’

residence in the United States, a strong relationship with his U.S.‐citizen wife,

work history, and payment of taxes) were outweighed by his criminal history and

lack of rehabilitation.

In August 2016, the BIA dismissed Matthews’s appeal. In re Gerard Patrick

Matthews, No. A042 231 142 (B.I.A. Aug. 30, 2016). The BIA agreed that Matthews

was removable, rejecting his argument that Mendoza Osorio misinterpreted the

scope of New York’s law and holding that NYPL § 260.10(1) “requires a knowing

act with a likelihood of harm to a child.” Certified Administrative Record (“CAR”)

at 4. The BIA also agreed with the IJ that Matthews did not merit discretionary

relief. Id. at 5.

7
 Matthews, who is represented by pro bono counsel, timely petitioned for

review.

DISCUSSION

I. Standard of Review

Where, as here, “the BIA has adopted the IJ’s reasoning and offered

additional commentary, we review the decision of the IJ as supplemented by the

BIA.” Gertsenshteyn v. U.S. Dep’t of Justice, 544 F.3d 137, 142 (2d Cir. 2008).

“Whether a conviction qualifies as a removable offense under a stated provision

of the INA is a question of law.” Mizrahi v. Gonzales, 492 F.3d 156, 157–58 (2d Cir.

2007). We review the BIA’s construction of state criminal law de novo. Id. at 158.

“To the extent the question requires us to construe a provision of the INA,

however, because the administration of that statute is entrusted to the BIA, our

review follows the two‐step process outlined in Chevron, U.S.A, Inc. v. Natural Res.

Def. Council, Inc., 467 U.S. 837 (1984).” Id.

II. The BIA’s Definition of “Crime of Child Abuse”

Matthews was ordered removed under 8 U.S.C. § 1227(a)(2)(E)(i), which

provides that, “[a]ny alien who at any time after admission is convicted of . . . a

crime of child abuse, child neglect, or child abandonment is deportable.” The

8
statute does not define crime of child abuse, neglect, or abandonment, but the BIA

has “interpret[ed] the term ‘crime of child abuse’ broadly to mean any offense

involving an intentional, knowing, reckless, or criminally negligent act or

omission that constitutes maltreatment of a child or that impairs a child’s physical

or mental well‐being.” Matter of Velazquez‐Herrera, 24 I. & N. Dec. 503, 512 (B.I.A.

2008). The BIA held that,

At a minimum, this definition encompasses convictions for offenses
involving the infliction on a child of physical harm, even if slight;
mental or emotional harm, including acts injurious to morals; sexual
abuse, including direct acts of sexual contact, but also including acts
that induce (or omissions that permit) a child to engage in
prostitution, pornography, or other sexually explicit conduct; as well
as any act that involves the use or exploitation of a child as an object
of sexual gratification or as a tool in the commission of serious crimes,
such as drug trafficking.

Id. The BIA’s definition also covers child endangerment offenses where no actual

harm or injury occurs, so long as the state statute requires a sufficient risk of harm

to a child. Matter of Soram, 25 I. & N. Dec. 378, 381–83 (B.I.A. 2010). In Soram, the

BIA held that a Colorado child endangerment statute that state courts had

interpreted to require a “reasonable probability” of injury to a child was

categorically a crime of child abuse, neglect, or abandonment. Id. at 385–86. The

BIA noted that, given the wide degree of variation in state child endangerment

statutes, “a State‐by‐State analysis is appropriate to determine whether the risk of

9
harm required by the endangerment‐type language in any given State statute is

sufficient to bring an offense within the definition of ‘child abuse’ under the

[INA].” Id. at 383.

Importantly, this is not the first time we have reviewed the BIA’s conclusion

that a state child endangerment offense is a removable crime of child abuse,

neglect, or abandonment under 8 U.S.C. § 1227(a)(2)(E)(i). In 2015, employing the

two‐step Chevron framework,3 we found the phrase “crime of child abuse, child

neglect, or child abandonment” to be ambiguous because it is not defined in the

INA, and we deferred to the BIA’s conclusion in Soram that this phrase

encompassed at least some child endangerment offenses. Florez, 779 F.3d at 211––

14. We noted that in 1996, when Congress added the crime of child abuse

provision to the INA, at least nine states had criminal child abuse statutes that

covered endangerment without injury. Id. at 212. We suggested, however, that a

definition that is not limited to crimes that pose a high enough risk of harm or

injury to a child may not be reasonable:

Although the BIA’s definition of ‘a crime of child abuse’ is expansive,
it is not unlimited. Soram confirms that a state child‐endangerment

3Under Chevron, courts ask, first, whether a statute is ambiguous. If it is not ambiguous, courts
must apply the statute’s clear meaning. Second, if the statute is ambiguous, courts defer to a
permissible interpretation of the statute by the administrative agency responsible for its
enforcement. 467 U.S. at 842–43.

10
 statute qualifies as a ‘crime of child abuse’ under the INA only if it
requires, as an element of the crime, a sufficiently high risk of harm to
a child . . . this limitation ensures that the BIA’s treatment of state
child‐endangerment statutes remains within the realm of reason.

Id. The BIA has agreed with this position. In Mendoza Osorio, the agency

“recognize[d] that there are child endangerment statutes that do not require a

sufficiently high risk of harm to a child to meet the definition of child abuse,

neglect, or abandonment under the [INA].” 26 I. & N. Dec. at 711. As an example,

the BIA pointed out that a conviction for misdemeanor child endangerment in

California is not a crime of child abuse under the INA because that statute

“punishes ‘conduct that creates only the bare potential for nonserious harm to a

child.’” Id. (quoting Fregozo v. Holder, 576 F.3d 1030, 1037–38 (9th Cir. 2009)); see

also Cal. Penal Code § 273a(b) (“Any person who, under circumstances or

conditions other than those likely to produce great bodily harm or death, . . .

willfully causes or permits that child to be placed in a situation where his or her

person or health may be endangered, is guilty of a misdemeanor.” (emphases

added)).

Whether a New York misdemeanor conviction for endangering the welfare

of a child is a crime of child abuse remains an open question of law in this Circuit.

Florez had conceded that NYPL § 260.10(1) fell within the BIA’s definition, and

11
therefore we did not need to decide the issue in that case. Florez, 779 F.3d at 209–

10, 212.4 In Matthews’s case, the BIA followed Mendoza Osorio, a precedential

decision holding that NYPL § 260.10(1) is a crime of child abuse. 26 I. & N. Dec. at

705–12. Focusing on the prong of the statute at issue here, “taking action that is

likely to be harmful to a child’s welfare,” the BIA concluded that a conviction

under the statute “requires a showing that the defendant knew that his actions

were likely to result in physical, mental, or moral harm to a child.” Id. at 705–06.

“To act ‘knowingly’ under New York law, the defendant must have been ‘aware’

of the nature of his conduct and of the fact that his actions had the potential for

harm.” Id. at 706 (quoting People v. Johnson, 95 N.Y.2d 368, 372 (2000)). “There

must also be proof that the harm was ‘likely to occur, and not merely possible.’”

Id. (quoting People v. Hitchcock, 98 N.Y.2d 586, 591 (2002)). The BIA reasoned that

§ 260.10(1) was categorically a crime of child abuse, neglect, or abandonment

because a conviction required “a knowing mental state coupled with an act or acts

creating a likelihood of harm to a child.” Id. The BIA addressed several New York

cases applying § 260.10(1) and noted that: (1) state courts regularly dismissed

4In an unpublished decision that pre‐dated Soram, we expressed skepticism about whether NYPL
§ 260.10(1) was a crime of child abuse, but we remanded for the BIA to define a “crime of child
abuse” in the first instance. Guzman v. Holder, 340 F. App’x 679 (2d Cir. 2009) (summary order).

12
endangerment charges or overturned convictions after determining that a

defendant’s conduct posed less than a likelihood of harm; and (2) the cases in

which the New York courts upheld endangerment charges as facially sufficient, or

upheld a § 260.10(1) conviction, involved factual circumstances that posed a

likelihood of physical, mental, or moral harm and thus fell within the BIA’s

definition. Id. at 707–12. The BIA declined to consider redacted complaints and

charging documents in its interpretation of the scope of the New York law on the

ground that these documents were not accompanied by any proof that the charges

led to a conviction. Id. at 707 n.4.

III. Matthews’s Arguments on Appeal

Matthews has two arguments. First, he challenges our decision in Florez and

argues that we must revisit the Chevron issue in light of a recent Supreme Court

decision, Esquivel‐Quintana v. Sessions, 137 S. Ct. 1562 (2017), which declined to

afford Chevron deference to the BIA’s interpretation of another INA provision

involving “sexual abuse of a minor.” Second, he argues that NYPL § 260.10(1)

does not fit within the broad definition of a crime of child abuse, neglect, or

abandonment that we accepted in Florez because New York court decisions and

charging documents show that a conviction for endangering the welfare of a child

13
under NYPL § 260.10(1) requires only a slight chance of nonserious harm to a

child. We address these arguments in turn and conclude that both lack merit.

A. Revisiting Florez

“In our Circuit, panels are ‘bound by the decisions of prior panels until such

time as they are overruled either by an en banc panel of our Court or by the

Supreme Court.’” United States v. Gill, 748 F.3d 491, 502 n.8 (2d Cir. 2014) (quoting

United States v. Wilkerson, 361 F.3d 717, 732 (2d Cir. 2004)). However, a panel may

overrule a prior decision if “there has been an intervening Supreme Court decision

that casts doubt on our controlling precedent.” Gelman v. Ashcroft, 372 F.3d 495,

499 (2d Cir. 2004) (quoting Union of Needletrades, Indus. & Textile Emps. v. INS, 336

F.3d 200, 210 (2d Cir. 2003)). Matthews argues that the Supreme Court’s 2017

decision in Esquivel‐Quintana casts doubt on Florez’s method of statutory

interpretation (the traditional two‐step Chevron analysis) and its conclusion that

Soram’s definition of a crime of child abuse was a permissible construction of 8

U.S.C. § 1227(a)(2)(E)(i).

In Esquivel‐Quintana the Supreme Court interpreted 8 U.S.C.

§ 1101(a)(43)(A), which covers the aggravated felony offense of “sexual abuse of a

minor.” Esquivel‐Quintana, 137 S. Ct. at 1567. The issue in that case was whether

14
state statutory rape laws that defined a “minor” as a person under the age of

eighteen fell within the INA’s definition of sexual abuse of a minor. Id. at 1568.5

The Supreme Court held that they did not because the federal generic age of

consent in 1996 (when this INA provision was enacted) was sixteen. Id. at 1569–

72. The Supreme Court looked to several different sources in reaching this

conclusion. First, the structure of the INA—specifically, the fact that the sexual

abuse of a minor offense was designated an “aggravated felony” and placed “in

the same subparagraph as murder and rape”—“suggest[ed] that sexual abuse of a

minor encompasses only especially egregious felonies.” Id. at 1570 (internal

quotation marks and citations omitted). Second, another provision of federal law,

18 U.S.C. § 2243, that criminalized “sexual abuse of a minor or ward,” was limited

to victims under age sixteen. Id. at 1570–71. The Court found it especially

instructive that Congress enacted 18 U.S.C. § 2243 “in the same omnibus law that

added sexual abuse of a minor to the INA.” Id. (citing Omnibus Consolidated

Appropriations Act, 1997, §§ 121(7), 321, 110 Stat. 3009–31, 3009–627). Third, the

5Esquivel‐Quintana had pleaded no contest to a California statute criminalizing “unlawful sexual
intercourse with a minor who is more than three years younger than the perpetrator” and
defining a “minor” as “a person under the age of 18 years.” 137 S. Ct. at 1567 (quoting Cal. Penal
Code § 261.5(c)). Thus, the minimum conduct criminalized under this statute would have been
consensual sexual intercourse between a seventeen‐year‐old and a twenty‐one‐year‐old. Id.

15
Court “look[ed] to state criminal codes for additional evidence about the generic

meaning of sexual abuse of a minor,” and found that “[a] significant majority of

jurisdictions thus set the age of consent at 16 for statutory rape offenses predicated

exclusively on the age of the participants.” Id. at 1571. The Court only briefly

addressed Chevron deference at the end of its decision. Id. at 1572. The

Government urged that any ambiguity in the statute should be resolved in its favor

under Chevron, while Esquivel‐Quintana had argued that any ambiguity should

be construed in his favor under the rule of lenity. Id. The Court stated that there

was “no need to resolve whether the rule of lenity or Chevron receives priority in

this case because the statute, read in context, unambiguously forecloses the

Board’s interpretation. Therefore, neither the rule of lenity nor Chevron applies.”

Id.

In Esquivel‐Quintana, the Supreme Court did not reject Chevron outright or

mandate any particular approach to statutory interpretation. At most, the

Supreme Court’s decision reminds courts to use all available tools of statutory

construction in order to discern Congress’s intent before concluding that a

statutory term is ambiguous and deferring to the implementing agency’s

interpretation. Cf. Chevron, 467 U.S. at 842 (“If the intent of Congress is clear, that

16
is the end of the matter.”); Mizrahi, 492 F.3d at 158 (“If the statutory language is

ambiguous, . . . we resort first to canons of statutory construction, and, if the

statutory meaning remains ambiguous, to legislative history, to see if these

interpretive clues clearly reveal Congress’s intent.” (internal quotation marks,

citations, and brackets omitted)).6 Because the Supreme Court’s holding in

Esquivel‐Quintana was narrow and its decision did not relate to the INA’s crime of

child abuse provision or mandate any particular approach to statutory

interpretation, it does not cast doubt on Florez.7 Cf. Correa‐Diaz v. Sessions, 881 F.3d

523, 528 (7th Cir.), cert. denied, 139 S. Ct. 224 (2018) (“[W]e do not believe Esquivel‐

Quintana’s limited holding overruled all of this Court’s previous decisions

6A June 2018 Supreme Court decision analyzing a different INA provision also concluded that
“the Court need not resort to Chevron deference” if “Congress has supplied a clear and
unambiguous answer to the interpretive question at hand.” Pereira v. Sessions, 138 S. Ct. 2105,
2113 (2018). Justice Kennedy’s concurrence expressed concern that the decisions by the Courts of
Appeals on that issue were too quick to defer to the BIA under Chevron, and were engaging in a
“cursory analysis” rather than using ordinary tools of statutory construction to discern
Congress’s intent. 138 S. Ct. at 2120 (Kennedy, J., concurring). Justice Kennedy suggested that
the Supreme Court “reconsider, in an appropriate case, the premises that underlie Chevron and
how courts have implemented that decision.” Id. at 2121. However, in the absence of such
reconsideration, we remain bound by Chevron.
7Matthews overemphasizes Esquivel‐Quintana’s survey of state criminal statutes, which was only
one of several sources the Supreme Court looked to in that case. See 137 S. Ct. at 1571 n.3 (agreeing
with the Government that “this sort of multijurisdictional analysis” is not required and stating
that “[i]n this case, state criminal codes aid our interpretation of ‘sexual abuse of a minor’ by
offering useful context.”). Because the INA’s crime of child abuse provision is not an aggravated
felony (a designation that carries serious consequences such as barring most forms of relief from
removal), and there is no other provision of federal law that defines a crime of child abuse, a
significant part of Esquivel‐Quintana’s statutory analysis is inapplicable to this case.

17
deferring to [the] BIA’s interpretation of ‘sexual abuse of a minor.’”). We conclude

that Florez remains binding, and for this reason, we continue to defer to the BIA’s

definition of crime of child abuse, neglect, or abandonment as including child

endangerment offenses that pose a sufficiently high level of risk of harm to a child.

The next question is whether New York’s child endangerment law fits within that

definition.

B. The Categorical Approach

As noted above, the BIA and federal courts agree that state child

endangerment statutes must require something more than a mere possibility of

nonserious harm to qualify as a crime of child abuse, neglect, or abandonment

under the INA. See Florez, 779 F.3d at 212 (noting that the BIA’s definition “is not

unlimited” and “requires, as an element of the crime, a sufficiently high risk of

harm to a child”); Fregozo, 576 F.3d at 1037–38 (holding that California’s

misdemeanor child endangerment statute was not a crime of child abuse under

the INA because it did not require “any particular likelihood of harm to a child”);

Mendoza Osorio, 26 I. & N. Dec. at 711 (discussing California statute and agreeing

that it is not categorically a crime of child abuse). The next issue, then, is whether

New York’s child endangerment law requires a level of risk of harm to a child that

18
is significant enough to qualify as a crime of child abuse, neglect, or abandonment

under the INA.8

In determining whether NYPL § 260.10(1) is a crime of child abuse, we

employ the “categorical approach,” looking only to the text of NYPL § 260.10(1)

and New York’s interpretation of that statute to determine whether there is a

categorical match with the BIA’s definition. Florez, 779 F.3d at 209–10. We use the

categorical approach whenever the relevant INA provision “makes aliens

removable based on the nature of their convictions, not based on their actual

conduct.” Esquivel‐Quintana, 137 S. Ct. at 1567–68. The child abuse provision, 8

U.S.C. § 1227(a)(2)(E)(i), like several other provisions of the INA, “premises

removability not on what an alien has done, or may have done, or is likely to do

in the future (tempting as it may be to consider those factors), but on what he or

she has been formally convicted of in a court of law.” Gertsenshteyn, 544 F.3d at 145;

see also Velazquez‐Herrera, 24 I. & N. Dec. at 513 (discussing BIA’s application of the

categorical approach to § 1227(a)(2)(E)(i)).

8We review the BIA’s interpretation of NYPL § 260.10(1) de novo, without according Chevron
deference, because state criminal law falls outside the BIA’s area of expertise. See Gill v. INS, 420
F.3d 82, 89 (2d Cir. 2005).

19
 Under this categorical approach, “[a]n alien’s actual conduct is irrelevant to

the inquiry, as the adjudicator must ‘presume that the conviction rested upon

nothing more than the least of the acts criminalized’ under the state statute.”

Mellouli v. Lynch, 135 S. Ct. 1980, 1986 (2015) (quoting Moncrieffe v. Holder, 569 U.S.

184, 190–91 (2013)). We must identify the least culpable conduct that is punishable

under NYPL § 260.10(1) and determine whether that conduct falls within the BIA’s

definition of a crime of child abuse, neglect, or abandonment. However, the

Supreme Court has cautioned that the categorical approach’s “focus on the

minimum conduct criminalized by the state statute is not an invitation to apply

‘legal imagination’ to the state offense; there must be a ‘realistic probability, not a

theoretical possibility, that the State would apply its statute to conduct that falls

outside the [federal] definition of a crime.’” Moncrieffe, 569 U.S. at 191 (quoting

Gonzales v. Duenas‐Alvarez, 549 U.S. 183, 193 (2007)); see also United States v. Hill,

890 F.3d 51, 59 (2d Cir. 2018).

The Supreme Court first announced the realistic probability standard in

2007 in Duenas‐Alvarez, 549 U.S. 183. Duenas‐Alvarez argued that his state statute

of conviction criminalized acts outside the generic definition and thus was not

categorically a removable theft offense. Id. at 190–94. In rejecting this argument,

20
the Court shifted the burden of proof from the Government to Duenas‐Alvarez to

demonstrate that, despite the apparent categorical match between the state statute

and the generic federal definition of theft, there was a realistic probability that the

state would apply the state statute to conduct outside the generic definition. Id. at

193.

In Moncrieffe, the Supreme Court again referenced the “realistic probability”

component of the categorical approach. 569 U.S. at 206. The Court held that a

Georgia statute criminalizing possession with intent to distribute marijuana was

not categorically an aggravated felony drug trafficking offense because it

criminalized conduct that was punishable as a misdemeanor as well as a felony

under federal law. More specifically, the Georgia statute made it a crime, inter alia,

to possess with intent to distribute small amounts of marijuana, while federal law

required a sale for the crime to qualify as a felony. Id. at 193–95. The Government

argued that the Court’s holding would “frustrate the enforcement of other

aggravated felony provisions, like § 1101(a)(43)(C), which refers to a federal

firearms statute that contains an exception for ‘antique firearms.’” Id. at 205. The

Court rejected the Government’s argument, stating that, given that “Duenas‐

Alvarez requires that there be ‘a realistic probability . . . that the State would apply

21
its statute to conduct that falls outside the generic definition of a crime[,]’ [t]o

defeat the categorical comparison in this manner, a noncitizen would have to

demonstrate that the State actually prosecutes the relevant offense in cases

involving antique firearms.” Id. at 205–06 (internal citation omitted).

Other than in Duenas‐Alvarez and Moncrieffe, the Supreme Court has not

expanded or again discussed the “realistic probability” approach in the context of

determining removability. Importantly, the Supreme Court has not yet

“specif[ied] what type of evidence may be used to satisfy the ‘realistic probability’

requirement.” Nicanor‐Romero v. Mukasey, 523 F.3d 992, 1005 (9th Cir. 2008),

overruled on other grounds by Marmolejo‐Campos v. Holder, 558 F.3d 903 (9th Cir.

2009); see also Jean‐Louis v. Att’y Gen., 582 F.3d 462, 481–82 (3d Cir. 2009). Although

it is the Government’s burden to prove removability by clear and convincing

evidence, 8 U.S.C. § 1229a(c)(3)(A); Woodby v. INS, 385 U.S. 276, 277, 286 (1966),

Moncrieffe and Duenas‐Alvarez suggest that it is a noncitizen’s burden to establish

a realistic probability of being convicted for conduct outside the federal definition,

at least in cases where the state and federal statutes appear to be a categorical

match. See also Hylton v. Sessions, 897 F.3d 57, 63 (2d Cir. 2018) (concluding that

the BIA erred in applying “realistic probability” test where statutes were not a

22
categorical match, but explaining that where there is a match, the noncitizen can

“show a ‘realistic probability’ that ‘the State would apply its statute to conduct

that falls outside the generic definition of a crime.’” (quoting Duenas‐Alvarez, 549

U.S. at 193)). Because we conclude, as set forth below, that the state and federal

statutes are a categorical match, Matthews has the burden to show that New York

applies NYPL § 260.10(1) to conduct that falls outside the BIA’s definition of a

crime of child abuse.

C. The BIA’s Definition of a Crime of Child Abuse is a Categorial
Match with NYPL § 260.10(1)

Section 260.10(1) criminalizes, as relevant here, “knowingly act[ing] in a

manner likely to be injurious to the physical, mental, or moral welfare of a child

less than seventeen years old.” New York defines the requisite mens rea,

“knowingly,” as follows: “A person acts knowingly with respect to conduct or to

a circumstance described by a statute defining an offense when he is aware that

his conduct is of such nature or that such circumstance exists.” NYPL § 15.05(2).

“Likely” is defined as “probable,” “showing a strong tendency,” or “reasonably

expected.” Black’s Law Dictionary (10th ed. 2014). The New York Court of

Appeals has interpreted NYPL § 260.10(1) as requiring both that the defendant

acted “with an awareness of the potential for harm” and “that the harm was likely

23
to occur, and not merely possible.” People v. Hitchcock, 98 N.Y.2d 586, 590–91 (2002)

(emphasis added); see also People v. Johnson, 95 N.Y.2d 368, 372 (2000) (“The

Legislature specifically recognized that behavior that was likely to produce harm

to a child’s physical, mental or moral well‐being fell within its sweep as long as

the defendant was aware of its potential for harm to a child.”); People v. Simmons,

92 N.Y.2d 829, 830 (1998) (“Actual harm to the child need not result for liability

under the statute to attach, it being sufficient that the defendant act in a manner

which is likely to result in harm to the child, knowing of the likelihood of such

harm coming to the child.”).

The case law illustrates that although the question of whether a defendant’s

conduct is likely to harm a child is highly “fact specific,” evidence of likely harm

is required to convict a defendant under NYPL § 260.10(1). Johnson, 95 N.Y.2d at

373. In Johnson, the Court of Appeals cited “social science and psychological

studies,” noting “the profound adverse effect on children who witness domestic

violence.” Id. at 372 n.*.

In People v. Simmons, the defendant was charged with “repeatedly directing

vulgar remarks of a sexual nature” to a child roughly two years old. 92 N.Y.2d at

830. The New York Court of Appeals recognized that the “child was in the

24
formative stages of speech and learning, had some verbal cognitive abilities and

in fact reacted to defendant’s vulgar queries, by answering ‘yes.’” Id. Under these

circumstances, “[t]he jury therefore may reasonably have concluded that the

totality of defendant’s remarks, repeated to the child over a six‐week period at a

crucial stage in her intellectual and social development, would have combined to

create a likelihood of harm, regardless of the child’s current level of understanding.”

Id. at 831 (emphasis added). The court recognized that jurors could “draw[] upon

their common human experience and commonsense understanding of the nature

of children” to conclude that the defendant’s “remarks were likely to have caused

the child harm.” Id. (emphasis added). Our own jurisprudence similarly allows

jurors to rely on common sense and experience to draw inferences such as whether

a defendant acted with criminal intent. See United States v. Anderson, 747 F.3d 51,

70 (2d Cir. 2014) (“[A]lthough the government is not permitted to build a

conviction on a house of cards, neither is a jury required to leave its common sense

at the courthouse door: ‘Jurors are entitled, and routinely encouraged, to rely on

their common sense and experience in drawing inferences.’” (quoting United States

v. Huezo, 546 F.3d 174, 182 (2d Cir. 2008)).

25
 New York appellate courts have reversed NYPL § 260.10(1) convictions after

finding insufficient evidence or improper instructions regarding either a

defendant’s knowledge or an actual likelihood of physical, mental, or moral harm

to a child. See People v. Hosue, 56 N.Y.S.3d 764, 767–68 (2d Dep’t 2017) (weight of

evidence was against guilty verdict for attempted child endangerment where

defendant pushed his seven‐year‐old daughter’s mother during an altercation but

“[t]here was no evidence of a history of domestic violence between the parties”);

People v. Coveney, 21 N.Y.S.3d 523, 528–29 (2d Dep’t 2015) (vacating child

endangerment conviction because there was no evidence of any nexus between

defendants’ conduct and a likelihood of harm to the complainant’s children);

People v. Chase, 720 N.Y.S.2d 707, 709 (2d Dep’t 2000) (requiring a “true likelihood

of injury which is not ‘speculative’” and finding that the specific circumstances of

defendant’s driving while intoxicated did not establish either a likelihood of harm

or defendant’s knowledge beyond a reasonable doubt); People v. Simmons, 635

N.Y.S.2d 373, 374 (4th Dep’t 1995) (“Because the degree of culpability required by

the statute is actual knowledge, it was reversible error to charge the jury that it

could find defendant guilty based upon what he should have known.”).

26
 As the BIA discussed in Mendoza Osorio, 26 I. & N. Dec. at 707 n.3, New York

criminal courts have also dismissed child endangerment charges as legally

insufficient in cases where the prosecution failed to allege a sufficient, non‐

speculative risk of injury to a child. See People v. Ventura, 801 N.Y.S.2d 241, 2005

WL 735873, at *2 (Crim. Ct. 2005) (dismissing complaint as facially insufficient

because defendant’s act of taking another’s cell phone and throwing it against a

wall did not “pose[] a likelihood of either physical or mental harm to the eleven

and fourteen year‐olds present”); People v. Grajales, 686 N.Y.S.2d 608, 610 (Crim.

Ct. 1999) (dismissing child endangerment counts for marijuana possession because

“the potential harm to the child cannot be merely remote or speculative in

nature”).

This is not a situation, therefore, in which the state statute, on its face,

stretches further than the BIA’s definition; instead, the state statute and the BIA’s

definition appear to be a categorical match.9 Compare Mendoza Osorio, 26 I. & N.

Dec. at 706 (“These elements—a knowing mental state coupled with an act or acts

creating a likelihood of harm to a child—fit within our definition of a ‘crime of

9 To the extent that Matthews takes issue with New York’s inclusion of “moral harm” against
children, we note that the BIA’s definition also encompasses “convictions for offenses involving
. . . mental or emotional harm, including acts injurious to morals.” Velazquez‐Herrera, 24 I. & N.
Dec. at 512.

27
child abuse, child neglect, or child abandonment.’”), with Mathis v. United States,

136 S. Ct. 2243, 2251 (2016) (“[T]he elements of Mathis’s crime of conviction . . .

cover a greater swath of conduct than the elements of the relevant [federal] offense

. . . . Under our precedents, that undisputed disparity resolves this case.”), and

Hylton, 897 F.3d at 63 (“[T]he statutory language itself . . . creates the realistic

probability that a state would apply the statute to conduct beyond the generic

definition.” (quoting Ramos v. U.S. Att’y Gen., 709 F.3d 1066, 1072 (11th Cir. 2013)).

D. Matthews Has Not Demonstrated a Realistic Probability that the
Statute is Applied More Broadly than the BIA’s Definition

Where a statute is not facially overbroad, the realistic probability approach

requires a noncitizen to demonstrate “that the State actually prosecutes the

relevant offense in cases” that fall outside the federal definition. Moncrieffe, 569

U.S. at 206. Matthews must “at least point to his own case or other cases in which

the state courts in fact did apply the statute in the special (nongeneric) manner for

which he argues.” Duenas‐Alvarez, 549 U.S. at 193.

Matthews does not argue that the facts underlying his convictions do not

match the BIA’s definition of a crime of child abuse. Rather, he argues that New

York courts have imposed liability for child endangerment in situations that may

be illustrative of “bad parenting” or other “minor missteps,” but do not actually

28
pose a likelihood of physical, mental, or moral harm to a child. Petitioner’s Br. at

50. He points to two categories of cases: “home alone” cases and cases involving

the commission of criminal activity, such as controlled substance use or

possession, in the presence of children.

For the “home alone” cases, Matthews and the dissent cite People v. Reyes, in

which a New York criminal court denied a defendant’s motion to dismiss a child

endangerment charge as facially insufficient. 872 N.Y.S.2d 692, 2008 WL 3010044

(Crim. Ct. Aug. 6, 2008); see N.Y. Crim Proc. Law § 100.40(4)(b) (providing that a

misdemeanor complaint is sufficient on its face when its factual allegations

“provide reasonable cause to believe that the defendant committed the offense

charged.”). In Reyes, the complaint alleged that a mother left her four‐year‐old

child home alone for about fifteen minutes in order to get groceries; although the

mother stated that the child was asleep when she left and she remained within

earshot, the police officer who found the child stated that the child was awake and

opened the door for the officers, who were strangers to the child. 2008 WL

3010044, at *1–2. The court determined that the allegations in the complaint made

out a sufficient likelihood of harm, citing to a case where a four‐year‐old started a

fire when left home alone for about ten minutes, and concluded that factual issues

29
such as whether the child was asleep and whether the defendant remained within

earshot were better left for trial. Id. at *3 & n.1. The court stated:

We hold that the issue of whether there is some minimum time that a
child must be left alone in order to hold a defendant liable under PL
§ 260.10(1) is ill‐suited for resolution on a motion to dismiss for facial
insufficiency. Among the factors which would appear appropriate
for consideration are the age of the child, the length of time involved,
the maturity of the particular child, and the reason why the child was
left alone.

Id. at *1. If the case proceeded to trial, the prosecution would have been put to its

burden to convince a jury that the defendant’s conduct resulted in a likelihood of

harm to her child and that the defendant knew that her conduct was likely to result

in injury to her child. Reyes and other “home alone” cases illustrate that New York

courts are cognizant of the line between “bad parenting” and child endangerment,

and provide prosecutors and juries with guideposts for determining when the

factual circumstances of a case establish culpability under NYPL § 260.10(1).10

10Similarly, although New York criminal courts are somewhat divided on whether the possession
and use of marijuana and other controlled substances pose a sufficient risk of harm to qualify as
child endangerment, the cases distinguish between situations in which children are actually
exposed to controlled substances, and those in which the defendant made an effort to hide or
conceal the illicit substances. Compare Grajales, 686 N.Y.S.2d at 609–10 (dismissing endangerment
charges on the basis that “the mere presence of mari[j]uana alone” in an apartment with children
present is not enough to show a likelihood of harm), with People v. Kennedy, 904 N.Y.S.2d 577, 579
3rd Dep’t 2010) (affirming child endangerment conviction based on teenage daughters’ repeated
exposure to cocaine residue and paraphernalia in defendant’s home). Matthews, moreover, has
not identified any case involving an actual child endangerment conviction that was based upon
an overbroad application of the statute to conduct involving the use or possession of a controlled

30
 The dissent, by contrast, highlights Reyes and other decisions in which New

York trial courts upheld the sufficiency of criminal misdemeanor complaints

brought under Section 260.10(1) for the proposition that the state courts are

applying the statute inconsistently and, accordingly, more broadly than the federal

definition. While we appreciate the dissent’s concern that some statements by trial

courts may not correctly interpret the state statute, there is no question that the

statute, as interpreted by the New York Court of Appeals, is a categorical match

with the BIA’s definition, and there is no New York appellate decision cited by the

dissent that has upheld a conviction that sweeps more broadly than the BIA

definition. The dissent notes correctly that the Supreme Court has yet to articulate

the contours of the “realistic probability” test to argue that our analysis is not so

limited, but it is for this precise reason that we find ourselves constrained from

reading these cases as contravening the interpretation espoused by New York’s

highest court.11

substance. The only controlled substance case he relies on involved a trial court’s denial of a
motion to dismiss. See People v. Alvarez, 860 N.Y.S.2d 745 (Crim. Ct. 2008). In Alvarez, the court
relied on the “strong odor of mari[j]uana throughout the apartment,” which implied that the
defendant had been smoking marijuana while the children were present, as an aggravating factor
that distinguished the case from others involving the mere possession of marijuana. Id. at 747–
49.
11In Armed Career Criminal Act (“ACCA”) cases, moreover, which also employ the categorical
approach and realistic probability test, the Supreme Court and other Courts of Appeals have

31
 For similar reasons, we decline to rely upon charging documents to alter our

analysis. Amici have submitted several redacted arrest reports, complaints, and

misdemeanor informations that charge defendants with violating NYPL

§ 260.10(1) for conduct such as: driving with a suspended license with a child in

the car, shoplifting while caring for an infant or toddler, leaving a nine‐year‐old

and sleeping five‐year‐old in a car for ten minutes with the windows open, yelling

and knocking items from a shelf in the presence of two children, and smoking

marijuana in a public playground while children are playing. Matthews, amici,

and the dissent argue that these charging documents show a realistic probability

that New York prosecutes conduct that only poses a minimal risk of harm to

children.

The Government does not contest the authenticity of these documents,

although in Matthews’s case the IJ and BIA relied on Mendoza Osorio, a

recognized that federal courts are bound by the highest state court’s interpretations of state law.
See Johnson v. United States, 559 U.S. 133, 138 (2010) (holding that although the meaning of ACCA’s
requirement of “physical force” is a question of federal law, the Court was “bound by the Florida
Supreme Courtʹs interpretation of state law, including its determination of the elements of” the
Florida battery statute at issue); United States v. Holloway, 630 F.3d 252, 259–60 (1st Cir. 2011)
(citing Johnson for the same proposition); see also United States v. Southers, 866 F.3d 364, 368 (6th
Cir. 2017) (rejecting notion that defendant’s reliance on Tennessee intermediate appellate court
decisions could satisfy “realistic probability” test to the extent that those “decisions appl[ied] the
Tennessee robbery statute in a manner . . . inconsistent with Tennessee Supreme Court
precedent.”).

32
precedential case in which the BIA refused to consider charging documents

without proof of a subsequent conviction, reasoning that “a judge or jury may

have found that the facts as charged were insufficient to support a conviction.” 26

I. & N. Dec. at 707 n.4. The fact that Matthews did not submit these documents to

the IJ or BIA prevents us from reviewing them because in the immigration context,

we are strictly constrained to the administrative record. 8 U.S.C. § 1252(b)(4)(A)

(“[T]he court of appeals shall decide the petition only on the administrative record

on which the order of removal is based.”).12

Even if we could consider these charging documents, we are not convinced

that they would change the outcome of this case. On the one hand, we recognize

the concern expressed by Matthews, amici, and the dissent that prosecutors may

be overcharging conduct under the child endangerment statute, and that the vast

majority of child endangerment charges result in guilty pleas and minimal

sentences. The dissent argues that we can consider these charging documents, in

12We have taken judicial notice of material and undisputed changes in country conditions that
occur between the BIA’s decision and our review, see Hoxhallari v. Gonzales, 468 F.3d 179, 185 n.5
(2d Cir. 2006) (per curiam), but that exception does not apply here. And Matthews and amici have
not provided any information about the court proceedings in which these arrest reports and
misdemeanor charges were filed, such that we might take judicial notice of those proceedings.
Cf. Kramer v. Time Warner Inc., 937 F.2d 767, 774 (2d Cir. 1991) (“[C]ourts routinely
take judicial notice of documents filed in other courts, . . . to establish the fact of such litigation
and related filings.”).

33
combination with a high incidence of guilty pleas, to infer that defendants are

pleading guilty to charges that are broader than the BIA’s definition. Amici,

however, have not submitted any evidence that these charging documents led to

convictions (whether through guilty pleas or trial), or that any court upheld the

allegations in these documents as facially sufficient. Without specific evidence

linking charging documents to guilty pleas, we have no way of knowing the

circumstances of the individual offenses that caused the prosecutors to bring the

charges or the defendants to plead to those charges. Moreover, this is not a

situation where we are without guidance because, as we have previously

explained, New York’s highest court has clarified the least‐acts‐criminalized by

NYPL § 260.10(1) and requires a likelihood of harm to a child to support a

conviction. For this reason, we agree with the BIA’s decision in Mendoza Osorio

and conclude that the charging documents and the cases Matthews and the dissent

point to are not enough to show a realistic probability that NYPL § 260.10(1) covers

conduct that falls outside the BIA’s definition of a crime of child abuse.13

13The dissent would grant Matthews’s petition on the additional and independent ground that
the BIA’s current definition of “child abuse” cannot be applied retroactively to Matthews’s 2002
and 2003 state convictions. Counsel, however, has not raised this issue, and we therefore do not
reach it. Nor do we foreclose the possibility that another petitioner might challenge an order of
removal premised on convictions that predated the BIA’s contemporaneous definition in a proper
case.

34
 CONCLUSION

Because we hold that NYPL § 260.10(1) is a categorical match with the BIA’s

definition of “crime of child abuse,” to which we owe Chevron deference, and

because Matthews has not established a realistic probability that the statute

encompasses conduct that stretches further than the BIA’s definition, we DENY

the petition for review.

Similarly, although we appreciate the policy concerns that Matthews, amici, and the dissent have
thoughtfully highlighted—such as the risk of separating families or disqualifying domestic
violence victims from obtaining immigration relief—these concerns are not present in Matthews’s
case, and he has not presented evidence that such concerns have materialized in the context of
New York’s child endangerment law.

35
CARNEY, Circuit Judge, dissenting.

I respectfully dissent.

New York Penal Law (“NYPL”) § 260.10(1) provides in relevant part, “A

person is guilty of endangering the welfare of a child when: (1) He or she

knowingly acts in a manner likely to be injurious to the physical, mental or moral

welfare of a child less than seventeen years old. . . .” The Board of Immigration

Appeals (“BIA”) has found the far‐ranging and vaguely contoured terms of this

state misdemeanor child endangerment statute to be categorically equivalent to

the similarly ambiguous federal definition of the phrase “crime of child abuse,

child neglect, or child abandonment.” Under the Immigration and Nationality

Act (“INA”), non‐citizens convicted of crimes falling under this federal definition

are both removable from the United States and absolutely barred from seeking

many forms of discretionary relief. 8 U.S.C. § 1227(a)(2)(E)(i); id. § 1229b(b)(1)(C).

Our court has previously deferred to and affirmed the agency’s legal

interpretations of the federal statute. See Florez v. Holder, 779 F.3d 207 (2d Cir.

2015). Given the dramatic evolution of the BIA’s interpretation of the ambiguous

federal law over the past thirty years, from requiring an intentional act involving

some “form of cruelty to a child’s physical, moral or mental well‐being,” In re

Rodriguez‐Rodriguez, 22 I. & N. Dec. 991, 996 (BIA 1999), to requiring only an act

that creates a “reasonable probability that the child’s life or health will be

endangered,” Matter of Soram, 25 I. & N. Dec. 378, 385 (BIA 2010), one might
question the decision to defer. But I agree with the majority that we are bound by

that decision.

Even accepting the BIA’s interpretation of the INA, however, the agency’s

decision to equate New York “child endangerment” law categorically with INA

“child abuse”—a decision first set forth in Matter of Mendoza Osorio, 26 I. & N.

Dec. 703 (BIA 2016), and now applied to Matthews’s case—does not in my view

hold under the analysis prescribed by the Supreme Court in Moncrieffe v. Holder,

569 U.S. 184 (2013), and, before that, in Gonzales v. Duenas‐Alvarez, 549 U.S. 183

(2007). Under these binding Supreme Court precedents, we must assess whether,

in applying its criminal law to the “least of the acts criminalized” under the state

statute, there is a “realistic probability” that a state could successfully prosecute

an individual for conduct that falls outside the federal definition. A review of

state trial court opinions and publicly available data reveals that individuals are

convicted under New York’s child endangerment statute for conduct that, even if

richly deserving of criticism, nonetheless falls outside the INA’s generic

definition of a “crime of child abuse” as conceived of by the BIA.

For example, in New York, individuals have been convicted under

§ 260.10(1) for leaving a sleeping child unattended in an apartment for at least

fifteen minutes, People v. Reyes, 2008 WL 3010044, 872 N.Y.S.2d 692 (Crim. Ct.

2008), repeatedly directing vulgar remarks at a toddler, People v. Simmons, 92

N.Y.2d 829 (1998), and possessing marijuana in their residence in proximity to

children, People v. Alvarez, 860 N.Y.S.2d 745 (Crim. Ct. 2008). See Guzman v.

Holder, 340 F. App’x 679, 682 (2d Cir. 2009) (summary order) (observing that “the

2
New York provision is extraordinarily broad”). Contrary to the expectation

expressed in Florez that a conviction under the state statute will be supported by

a “sufficiently high risk of harm,” 779 F.3d at 212,—a standard that itself floats,

unmoored, on the fickle sea of child‐rearing conventions—state prosecutions

have gone forward and convictions have been secured without evidence of the

risk or nature of the harm posed by the ostensibly criminal conduct. See Part I,

below. Similarly undercutting the BIA’s analysis, the lesser boundaries of the

state statute as applied to one significant category of prosecutions are unclear

even to the state adjudicators: as New York trial courts have repeatedly

commented, “[t]here is no clear consensus among New York cases regarding the

sufficiency of an accusatory instrument charging Endangering the Welfare of a

Child regarding leaving a child alone at home or in a vehicle.” People v. Hot, 2018

WL 652480, at *2, 94 N.Y.S.3d 539 (Crim. Ct. 2018).

In my view, these facts, supported by reliable data about how often the

state prosecutes the crime and obtains guilty pleas that are punished by non‐

carceral sentences, take the state crime significantly out of alignment with the

INA’s definition of child abuse. They demonstrate that, for immigration law

purposes, the crimes are not categorically equivalent.

Relying primarily on published New York state appellate decisions, the

majority declines to consider the full spectrum of publicly available data and

evidence of prosecutions under the statute. But we need not be so restricted: the

data and the lower court decisions are largely in the public record, and the texts

of those decisions that are nominally “unreported” are nonetheless publicly

3
available. Even if Matthews may not have presented these precise data and court

decisions to the BIA, in assessing the “realistic probability” of prosecution under

the statute as we apply the categorical approach, our panel may and indeed must

consider such public data that reflects the reach of the statute as it is actually

applied. These data and decisions paint a picture of prosecutions and guilty

pleas showing that the statute’s broad and ambiguous language is enforced in a

far more expansive, flexible, and subjective fashion than the reported case law

might lead one to expect. To overlook this material is to rely on a flawed

foundation in concluding that, as prosecuted, New York misdemeanor “child

endangerment” is equivalent to the INA’s definition of “child abuse.”

The results of the equation are troubling: to adopt the BIA’s conclusion of

categorical equivalence in 2019, and to apply that conclusion retroactively, is to

impose consequences on guilty pleas entered years ago that were then

unimaginable for a non‐citizen defendant. See generally Obeya v. Sessions, 884 F.3d

442 (2d Cir. 2018); Lugo v. Holder, 783 F.3d 119 (2d Cir. 2015). To be sure, the

charged crime of some who were subject to such prosecutions under § 260.10(1)

likely fell in the heartland of the generic INA definition of “child abuse.” In such

cases, that separation from even one’s own children may be the consequence of a

guilty plea is no surprise; rather, it is envisioned and may be required by both

criminal and family law. But the conclusion seems unavoidable that, for those

whose charged endangerment of a child posed more attenuated, limited, or

uncertain risks—those whose alleged neglect was leaving a sleeping child home

alone briefly, for example; perhaps even for Matthews—the calculus of entering

a guilty plea would surely have been significantly altered had the accused

4
parents known that doing so could mean risking the loss of the right to live with

their own children in the United States. Although the majority’s opinion does not

hold that the BIA may apply its ruling retroactively in all cases, today’s decision

means that those parents will need to live with fear of that new risk.

Aggravating these circumstances yet further, under the INA, individuals

convicted of a crime of “child abuse” are categorically ineligible for the

cancellation of removal that is otherwise available to non‐permanent residents. 8

U.S.C. § 1229b(b)(1)(C). This discretionary form of relief requires the petitioner to

show that removal would effect “exceptional and extremely unusual hardship”

to a U.S. citizen‐ or lawful permanent resident‐ child (or other qualifying

relative). Id. § 1229b(b)(1)(D). Today’s holding means that some individuals who

pleaded guilty to New York misdemeanor endangerment of a child, and who are

also able to prove that separation would cause “exceptional and extremely

unusual hardship” to that same child, would be unable even to seek

discretionary relief. I see no reason to conclude from the INA’s language or

history that Congress intended that result.

The majority’s affirmance of the BIA’s erroneous conclusion also opens the

door to consequences that are fundamentally at odds with Congress’s stated aim

in enacting the immigration portions of the Violence Against Women Act

(“VAWA”). These authorize immigration courts to provide discretionary relief

from removal for individuals who have been “battered or subjected to extreme

cruelty” by a U.S. citizen or lawful permanent resident. 8 U.S.C.

§ 1229b(b)(2)(A)(i). But individuals convicted of a crime falling within the INA

5
definition of “child abuse” are ineligible for cancellation of removal under these

“battered spouse or child” provisions. Id. § 1229b(b)(2)(A)(iv). This effect, too,

suggests the error of the agency’s analysis.

The majority thus errs, in my view, by denying Matthews’s petition.

I.

Gerard Matthews became a lawful permanent resident of the United States

in 1989, having emigrated to this country from Ireland.1 A.R. 97. For some period

of his residency, he has worked as a carpenter; he has consistently paid his taxes.

Id. He has been married to a U.S. citizen since 2001. Id.

In 2016, the BIA entered a removal order against him based primarily on

his convictions in 2002 and 2003 under the New York misdemeanor child

endangerment statute, NYPL § 260.10(1).2 As relevant here, § 260.10(1) proscribes

a “knowing[] act[] [done] in a manner likely to be injurious to the physical,

mental or moral welfare of a child less than seventeen years old.” For his part,

Matthews’s unhappy predicate acts were exposing himself on public streets in

the proximity of minors, in 2002 and 2003. Matthews, himself a victim of child

abuse during his youth in Ireland and who has been convicted of public

1These facts are taken from the April 7, 2016 order of the Immigration Judge finding
Matthews removable.

2A prior removal order was entered against him in 2011 based on convictions that he
sustained in the 1990s for public lewdness, a violation of NYPL § 245.00. The order was
rescinded after he successfully challenged it. See Matthews v. Holder, 590 Fed. App’x 75
(2d Cir. 2015).
6
lewdness on other occasions not involving minors, has sought and undergone

treatment for his underlying disorder. A.R. 98–99.

To prevail on his petition for review, Matthews must prove that § 260.10(1)

criminalizes conduct falling outside the INA’s definition of a “crime of child

abuse, child neglect, or child abandonment”—for present purposes, the operative

phrase in the INA. 8 U.S.C. § 1227(a)(2)(E)(i).3 He may do so by demonstrating a

“realistic probability . . . that the State would apply its statute to conduct that

falls outside the generic definition,” and pointing to “other cases in which the

state courts in fact did apply the statute in the special (nongeneric) manner.”

Gonzales v. Duenas‐Alvarez, 549 U.S. 183, 193 (2007). I believe that Matthews has

made that showing.

A. Defining the Generic INA Offense

Determining whether NYPL § 260.10(1) criminalizes conduct beyond the

generic INA offense of child abuse requires that we first identify that generic

definition.

The BIA has not made the task a simple one: its understanding of the

“generic offense” has changed significantly over the past two decades. In the late

1990s, the BIA focused on acts of “cruelty” to children, construing the INA

phrase to cover “[a]ny form of cruelty to a child’s physical, moral or mental well‐

being.” In re Rodriguez‐Rodriguez, 22 I. & N. Dec. 991, 996 (1999). In 2008, in Matter

3Section 237(a)(2)(E)(i) of the INA is codified at 8 U.S.C. § 1227(a)(2)(E)(i). For the
reader’s convenience, I refer to 8 U.S.C. § 1227(a)(2)(E)(i), the codified section, instead of
the section number in the enacted bill.
7
of Velazquez‐Herrera, 24 I. & N. Dec. 503, 512 (BIA 2008), the agency meaningfully

broadened its operative definition in several respects, holding that it covered not

only “intentional, knowing, [or] reckless” conduct, but “criminally negligent

act[s] or omission[s]” as well.4 Its 2008 decision appeared to emphasize the sexual

mistreatment of children and strongly suggested that some quantum of actual

harm or injury to a child was required:

At a minimum, this definition encompasses convictions for
offenses involving the infliction on a child of physical harm, even
if slight; mental or emotional harm, including acts injurious to
morals; sexual abuse, including direct acts of sexual contact, but
also including acts that induce (or omissions that permit) a child to
engage in prostitution, pornography, or other sexually explicit
conduct; as well as any act that involves the use or exploitation of
a child as an object of sexual gratification or as a tool in the
commission of serious crimes, such as drug trafficking.

Id.

Two years after Velazquez‐Herrera was decided, in Matter of Soram, 25 I. &

N. Dec. 378 (BIA 2010), the agency announced that the statutory definition that it

4The BIA’s discussion in Rodriguez‐Rodriguez was dictum. Nevertheless, the agency
cited its “cruelty” formulation in several non‐precedential decisions preceding
Velazquez‐Herrera. See, e.g., In re Pacheco Fregozo, 2005 WL 698590, at *1 (BIA 2005); In re
Maltez‐Salazar, 2005 WL 952489, at *1 (BIA 2005); In re Manzano‐Hernandez, 2005 WL
698392 at *2 (BIA 2005).

In turn, several Circuit courts deferred to the formulation stated in Rodriguez‐Rodriguez,
and this Court, in a footnote, suggested that it provided the operative definition. See
Ochieng v. Mukasey, 520 F.3d 1110, 1114–15 (10th Cir. 2008) (deferring to Rodriguez‐
Rodriguez); Loeza‐Dominguez v. Gonzales, 428 F.3d 1156, 1158 (8th Cir. 2005) (same);
Nguyen v. Chertoff, 501 F.3d 107, 114 n.9 (2d Cir. 2007) (noting Rodriguez‐Rodriguez
definition).
8
will apply is not limited to “those [offenses] requiring proof of actual harm or

injury to a child,” id. at 381 (emphasis added), but includes those in which the

conduct created “a reasonable probability that the child’s life or health will be

endangered,” id. at 384. Five years later, in Florez v. Holder, we accorded Chevron

deference to that expansive construction of the INA and held that a state child

endangerment statute will be held categorically to match the INA’s generic

definition if “it requires, as an element of the crime, a sufficiently high risk of

harm to a child.” 779 F.3d 207, 212 (2d Cir. 2015).

In 2016, in Matter of Mendoza Osorio, 26 I. & N. Dec. 703 (BIA 2016), the BIA

acknowledged that some state child endangerment statutes “do not require a

sufficiently high risk of a harm to a child to meet the [INA’s generic] definition.”

Id. at 711. As an example, the agency cited a California law criminalizing

“conduct that places a child ‘in a situation where his or her person or health may

be endangered,’” which the Ninth Circuit described as “punish[ing] ‘conduct

that creates only the bare potential for nonserious harm to a child.’” Id. (quoting

Fregozo v. Holder, 576 F.3d 1030, 1038 (9th Cir. 2009)). Since then, the BIA has not

spelled out what level of risk, of what type of harm, will satisfy the generic INA

definition.

Even before its decisions in Soram and Mendoza Osorio, we pointed out

(albeit in a non‐precedential order), that the agency’s standards for the level and

type of risk required by the INA’s generic definition of “child abuse” have been

evolving and expanding; its limits remain unclear. Guzman v. Holder, 340 F.

App’x 679 (2d Cir. 2009) (summary order); see also Martinez‐Cedillo v. Sessions, 896

9
F.3d 979, 999 (9th Cir. 2018) (Wardlaw, J., dissenting) (outlining history of BIA’s

changing position and opining that Board’s failure “to define the precise level of

risk required” divorces its definition from statutory text and renders its

definition “judicially unadministrable and overly vague”).5 Although the

definitional task admittedly is not an easy one, the BIA neither articulates the

quantum of risk required by the generic crime nor addresses the functional

relationship between the gravity of the risk posed and likelihood of the risk’s

realization: a reasonable person might expect that the lower the harm presented

by certain conduct, the higher the likelihood need be of its occurrence to create

criminal liability. For example, while it is easy to imagine all kinds of harms,

including mortal harm, that could arise from leaving a sleeping four‐year‐old

child alone in an apartment for fifteen minutes while a parent runs to a

neighboring store for milk, the likely harm seems minimal: this is a risk that some

reasonable and caring parents will run if pressed by circumstances.

What is clear is that the BIA has deemed that the level of risk was adequate

in the Colorado statute considered in Soram to meet the generic INA child abuse

definition—a “reasonable probability that the child’s life or health will be

endangered”—but inadequate in the California statute considered in Fregozo—a

child’s “person or health may be endangered.” Recognizing the haziness of this

conception, one of our sister Circuits has recently concluded that a conviction

5The Ninth Circuit’s decision in Martinez‐Cedillo and Judge Wardlaw’s dissent in that
case were withdrawn pending en banc review. Martinez‐Cedillo v. Barr, 918 F.3d 601 (9th
Cir. 2019). The appeal was recently dismissed altogether as moot in light of the
petitioner’s death. Martinez‐Cedillo v. Barr, No. 14‐71742, 2019 WL 2136113 (9th Cir. May
16, 2019). I cite to Judge Wardlaw’s observations for their persuasive value alone.
10
under a Pennsylvania child endangerment statute that “merely requires conduct

that could threaten a child’s welfare” does not amount to a crime of child abuse

under the INA’s generic definition. Zhi Fei Liao v. Attorney Gen., 910 F.3d 714, 722

(3d Cir. 2018) (internal quotation marks omitted) (emphasis added).

With this background, I turn briefly to the text of the relevant New York

law. As noted above, NYPL § 260.10(1) provides that a defendant is guilty of

misdemeanor child endangerment when “[h]e or she knowingly acts in a manner

likely to be injurious to the physical, mental or moral welfare of a child.” Unlike

the California statute at issue in Fregozo, § 260.10(1) by its terms criminalizes only

conduct that is “likely” to create a physical, mental, or moral injury. In Florez, the

petitioner did not contest the point; accordingly, we assumed without deciding

that the New York statute was coextensive with the federal definition, 779 F.3d at

209, and therefore that the risk envisioned as a predicate to criminal liability was

“sufficiently high”—a pivotal phrase whose ambiguity is obvious.

Consequently, we must determine whether, as evidenced by actual

prosecutions and convictions, we nonetheless discern a “realistic probability”

that New York convicts individuals for conduct generating risks of harm of a

likelihood or severity below the federal threshold. See Hylton v. Sessions, 897 F.3d

57, 64 (2d Cir. 2018) (realistic probability inquiry necessary where “the elements

of the state statute alone do not provide sufficient guidance on its application”). I

therefore next address this test, and my objections to the majority’s application of

it.

11
 B. The Majority’s Application of the “Realistic Probability”
Test

As the majority observes, the New York Court of Appeals has ruled that a

conviction under § 260.10(1) requires that a defendant have acted “with an

awareness of the potential for harm,” and “that the harm was likely to occur, and

not merely possible.” People v. Hitchcock, 98 N.Y.2d 586, 590–91 (2002).6 Matthews

and his amici, however, point to numerous New York prosecutions in which the

identified risk of harm seems indeed to have been “merely possible.” Further,

New York courts appear to be divided on the likelihood of harm that the

prosecution needs to show to secure a conviction. In my view, these prosecutions

and the related convictions demonstrate a “realistic probability” that § 260.10(1),

as applied, sweeps more broadly than does the INA’s generic definition of a

crime of child abuse.

1) The “Home Alone” Cases and N.Y. Court Rulings on Motions
to Dismiss for Facial Insufficiency

The main category cited by Matthews and amici is the “home alone”

cases—in which a defendant is charged with violating § 260.10(1) for having left

a child unattended for some period of time. For example, in People v. Reyes, a

6The majority’s application of the “realistic probability” test is largely consistent with
that of the BIA in Mendoza Osorio, 26 I. & N. Dec. 703, where the agency rejected similar
arguments raised by the respondent and held that § 260.10(1) was categorically a “crime
of child abuse” under the INA. Id. at 712. Because the BIA has “no expertise in
construing . . . state criminal statutes,” however, the majority rightly subjected the BIA’s
interpretation of § 260.10(1) to de novo review. See Gill v. I.N.S., 420 F.3d 82, 89 (2d Cir.
2005). Accordingly, I refer to the majority’s conclusions rather than the agency’s in this
discussion.
12
defendant was prosecuted for “allegedly leaving her [sleeping] four year old

child in an apartment, unsupervised, for at least fifteen minutes.” 2008 WL

3010044 at *1, 872 N.Y.S.2d 692 (Crim. Ct. 2008). The defendant moved to dismiss

the complaint as facially insufficient as a matter of law, arguing that it did not

“allege facts sufficient to provide reasonable cause to believe that the defendant

committed the offense charged.” Id. at *2. The trial court disagreed, expressing

the view that “[i]t is reasonable to imagine the wide range of harm that might

befall a four year old child left alone [awake] in an apartment,” but cautioning at

the same time that factual issues such as “the age of the child, the length of time

involved, the maturity of the particular child, and the reason why the child was

left alone” were “particularly unsuitable for determination on motion, and

except in the most extreme cases, are best reserved for trial.” Id. at *1, *3.

The majority reasonably proposes that, “if [Reyes] proceeded to trial, the

prosecution would have been put to its burden to convince a jury that the

defendant’s conduct resulted in a likelihood of harm to her child and that the

defendant knew that her conduct was likely to result in injury to her child.” Maj.

Op. at 30. Therefore, it follows, Reyes illustrates that “New York courts are

cognizant of the line between ‘bad parenting’ and child endangerment, and

provide prosecutors and juries with guideposts for determining when the factual

circumstances of a case establish culpability . . . .” Id.

With respect, I must disagree with that conclusion. Although the Reyes

court’s decision to commit the judgment to a jury is understandable, the

“guideposts” that the majority sees in Reyes strike me as largely illusory. While

13
New York state courts have published dispositions resolving motions to dismiss

criminal complaints brought under § 260.10(1), as happened in Reyes, the parties

have pointed to no judicial statements resolving a sufficiency challenge to a

defendant’s conviction on “home alone” charges under § 260.10(1), and I have

identified none. This leaves the actual guideposts available for prosecutors

making charging decisions and for defendants making plea decisions in fact

encased in fog.

This vagueness persists, but not because New York only rarely prosecutes

“home alone” cases under § 260.10(1) or has only recently begun to do so.

Numerous New York courts have found criminal misdemeanor complaints

under § 260.10(1) facially sufficient for prosecution on facts reminiscent of Reyes.

See, e.g., People v. Hot, 2018 WL 652480, 94 N.Y.S.3d 539 (Crim. Ct. 2018) (toddler

left sleeping in car while mother shopped nearby); People v. Fielden, 2015 WL

4460568, 18 N.Y.S.3d 581 (Crim. Ct. 2015) (infant left awake in hotel room for one

hour); People v. Eury, 2015 WL 135097, 7 N.Y.S.3d 244 (Crim. Ct. 2015) (four or

five children aged under ten left alone, with four asleep, in apartment, for at least

40 minutes); People v. Gulab, 2009 WL 1309785, 886 N.Y.S.2d 68 (Crim. Ct. 2009)

(two children ages five and ten home alone for two hours); People v. Fraser, 2008

WL 4866334, 875 N.Y.S.2d 822 (Crim. Ct. 2008) (security officer saw infant child

in stroller in an apartment building hallway; defendant stated that she was

“down the hall watching”); People v. Watson, 700 N.Y.S.2d 651, 655 (Crim. Ct.

1999) (seven‐year‐old child home alone awake for two‐and‐a‐half hours).7 As

7Worthy of observation in connection with VAWA, with the sole exception of Watson,
only women were charged in these “home alone” cases.
14
long ago as in 1997, in People v. Cenat, 671 N.Y.S.2d 578, 580 & n.2 (Crim. Ct.

1997), a New York trial court observed that “[o]ver the past few years the

Criminal Court has seen a flood of cases charging Endangering the Welfare of a

Child for leaving children of various ages ‘home alone’” and reported at the time

that “[t]here are over 200 [child endangerment] cases pending currently, a

substantial number of which allege ‘home alone’ type allegations.”

Despite the apparent frequency of prosecutions in “home alone” cases

over an extended period, New York courts continue to express a range of

inconsistent judicial views about legal sufficiency. See Eury, 2015 WL 135097 at *3

(“[E]ven the simple question of whether [an] allegation that the defendant left a

child or children alone for a time—without any aggravating circumstance—is

sufficient has produced conflicting results.”); Watson, 700 N.Y.S.2d at 655

(recognizing a “divergence in judicial views on the matter”). Matthews and amici

provide a compelling and plausible explanation for the persistent haziness: such

cases are overwhelmingly unlikely to go to trial (and, consequently, even less

likely to present a sufficiency challenge for appellate review). For example,

according to the New York State Division of Criminal Justice Services, and as

reflected in the attached chart, in 2008—the year that Reyes was decided—4,851

people in the State were arrested for violating § 260.10(1). Approximately 33% of

15
the resulting prosecutions were dismissed.8 Of the remaining cases,

approximately 80% were resolved by guilty plea.9 Only forty‐two people

(approximately 1.3% of those whose cases were not dismissed) went to trial. Of

these forty‐two, twenty were convicted and twenty‐two were acquitted. Put

another way: of convictions obtained under § 260.10(1) in the year 2008,

over 99% were obtained by guilty plea, resulting in no articulation at all of any

“guideposts.”

The reported data are similar for the other years in the dataset, which

covers from 1990 through 2015.10 The available statistics regarding these

prosecutions thus suggest that, unless the complaint is dismissed as facially

insufficient, defendants in a § 260.10(1) prosecution go to trial in about 1% of

cases and will in approximately 80% of cases plead guilty to the charge, with no

judicial articulation of any limits on their misdemeanor crimes.

8The Government has not contested the accuracy of the statistics presented by amici,
which amici advise were obtained by amicus Immigrant Defense Project from the New
York State government pursuant to a request for information under New York’s
Freedom of Information Law. Amici Br. 23 n.8. I see no reason why they should not be
subject to judicial notice. See Fed. R. Evid. 201(b)(2) (“The court may judicially notice a
fact that is not subject to reasonable dispute because it: . . . can be accurately and readily
determined from sources whose accuracy cannot reasonably be questioned.”)

9 These numbers do not add up to 100% because 359 cases remained open without
disposition; 132 cases were not prosecuted; and 148 cases in the chart were simply
labeled “other,” without further specification.

10The full database cited by amici is available at
https://www.immigrantdefenseproject.org/wp‐content/uploads/2016/06/DCJS‐EWOC‐
data‐arrs‐dispos‐sents.pdf (last accessed June 3, 2019).
16
 These dramatic figures should not be surprising, even apart from the well‐

known frequency of plea bargains in the criminal justice system: the data further

show that, in prosecutions under § 260.10(1) from 2000 to 2015, less than 20% of

convictions resulted in any term of imprisonment at all; 43% of convictions led

only to probation or a fine; and over 35% of convictions resulted in what the state

calls a “conditional discharge,” meaning a non‐carceral sentence to which certain

conditions are attached. Amici Br. 10. Significant here, by statute, a “conditional

discharge” embodies a judicial finding that “neither the public interest nor the

ends of justice would be served by a sentence of imprisonment” or even by

probation. NYPL § 65.05(1)(a). When, in over one‐third of prosecutions brought

under the statute, the sentence for pleading guilty is of this type (and is

understood to bear no immigration consequences), a defendant has little

incentive to contest the charges or proceed to trial.

Despite these data, the majority’s approach appears to require Matthews to

show that a § 260.10(1) defendant charged with conduct outside the INA’s

generic definition could be convicted at trial and have his conviction upheld on

appellate review. Absent that appellate statement, the construction of the statute

that underlies his conviction has little import, the reasoning would go.

This approach is certainly defensible as a formal matter: after all, the New

York state courts, not its prosecutors (or the local federal courts), have the final

word on the correct construction of New York laws. But the Supreme Court has

never described the “realistic probability” test in these terms, nor in my

estimation does it make sense to require such a showing here. Rather, in Duenas‐

17
Alvarez, the Court required only a showing that “the State would apply its statute

to conduct that falls outside the generic definition.” 549 U.S. at 193 (emphasis

added). More recently, in Moncrieffe, the Court required a showing that “the State

actually prosecutes the relevant offense” in a fashion that is inconsistent with the

generic offense. 569 U.S. at 206 (emphasis added). To import into the Court’s

“realistic probability” test a requirement that the state appellate courts describe

the farthest contours of the state law’s application strikes me as both unworkable

and inappropriate, particularly in the context of a misdemeanor crime, and

where (as here) courts are unlikely ever to have the opportunity to do so. An

approach that by definition focuses on only the 0.8% of convictions that are

secured following a jury trial (and the even smaller percentage that are

subsequently upheld on appellate review) will necessarily fail to grasp “the

elements of the offense in practice.” Whyte v. Lynch, 807 F.3d 463, 469 (1st Cir.

2015).

The Supreme Court has not as yet applied, and we are rarely, if ever,

called upon to apply the “realistic probability” test to misdemeanor statutes that

criminalize conduct that is punished so leniently as the mine‐run of prosecutions

under § 260.10(1). But because the analysis that we undertake requires us to

identify “the least of the acts criminalized under the state statute,” Mellouli v.

Lynch, 135 S. Ct. 1980, 1986 (2015) (internal quotation marks omitted), and

because the immigration consequences will be so severe, we must examine all

available and reliable evidence regarding related convictions to ensure the

accuracy of the New York endangerment statute’s asserted match with the

generic federal definition.

18
 If the allegations in Reyes, discussed above, were deemed sufficient to

survive a motion to dismiss, the case suggests that some defendants (indeed,

based on the statistical evidence before us, possibly many defendants) have

pleaded guilty to similar offense conduct. The only remaining question, then, is

whether the conduct alleged in Reyes and similar “home alone” cases more

closely approximates conduct creating the “sufficiently high risk of harm to a

child” we required in Florez or the “bare potential for nonserious harm”

described by the Ninth Circuit in Fregozo and distinguished by the BIA in

Mendoza Osorio, 26 I. & N. Dec. at 711. Even drawing all factual inferences in

favor of the prosecution in these cases, in my view the nature of the allegations

presented and the judicial comments elicited seem far more consistent with the

speculative “may be endangered” standard found wanting in Fregozo (or the

“conduct that could threaten a child’s welfare” found similarly insufficient by

the Third Circuit in Liao, 910 F.3d at 722), than the “sufficiently high risk of harm

to a child” that we demanded in Florez.11

11Tending to reinforce this conclusion, New York courts have accepted remarkably slim
“risk of harm” showings in some § 260.10(1) prosecutions. For example, in People v.
Simmons, 92 N.Y.2d 829 (1998), the New York Court of Appeals upheld the
endangerment conviction of a child care worker who “repeatedly direct[ed] vulgar
remarks of a sexual nature” at a two‐year‐old child. Id. at 830. On appeal, the defendant
challenged the scant record evidence regarding the likelihood of harm inflicted on the
child. Rejecting that challenge, the court explained that, “regardless of the child’s
current level of understanding,” the jury needed no evidence on the issue of harm, but
could rely on its “common human experience and commonsense understanding of the
nature of children” to conclude that the remarks were “likely to have caused the child
harm, and that defendant knew that her remarks were likely to cause the child to suffer
harm.” Id. at 831.

19
 For these reasons, I conclude that Matthews has carried his burden of

showing a “realistic probability” that New York prosecutes conduct under

§ 260.10(1) that falls outside the INA’s generic definition of child abuse. I would

grant his petition for review on this basis.

2) Import of the Charging Documents Presented by Amici

In similar vein, Matthews’s amici have presented several arrest reports,

complaints, and misdemeanor informations—the authenticity of which the

Government does not contest—as reflecting New York State decisions to charge

various individuals with state child endangerment misdemeanor violations for

conduct such as: (1) driving with a suspended license while a minor child is in

the car (Amici App. 2); (2) “yell[ing] and swing[ing] his backpack” in a store and

knocking objects off shelves, thereby injuring a child (Amici App. 9–10); and (3)

Judge Titone observed in dissent that the prosecution itself had “put forth proof that the
young girl here did not understand appellant’s lewd questions,” and opined that the
jury had “no evidentiary basis . . . to conclude that the young girl would likely be
harmed”; rather, the “evidence pointed only to the conclusion that the child here was
nothing more than a baby who was capable of mouthing the words ‘yes, yes’ but said
this to ‘everything’ regardless of context or content.” Id. at 832 (Titone, J., dissenting).

In my view, the Simmons decision and cases of this ilk provide another persuasive
indication that state prosecutors apply NYPL § 260.10(1) outside the generic federal
definition.

20
shoplifting while accompanied by children (Amici App. 12, 14, 16).12 None of the

charged conduct is commendable, to be sure, and under the “moral welfare”

portion of the New York statute, prosecution may have been justifiable.

But the majority’s position must be that, absent any evidence that the

allegations in the charging documents ever led to convictions or were upheld by

state courts as facially sufficient, these documents, too, do not suggest that

individuals in New York have a “realistic probability” of conviction under

§ 260.10(1) for conduct falling outside the INA’s generic definition of “child

abuse.” See Maj. Op. at 32–34. Such documents alone may not be sufficient to

carry Matthews’s burden, it is fair to say. But the broad range of conduct that

they reflect as warranting state criminal charges further suggests that the statute

12The majority accepts the government’s argument that we should not consider these
charging documents because Matthews did not introduce them into the administrative
record below. See 8 U.S.C. § 1252(b)(4)(A) (“[T]he court of appeals shall decide the
petition only on the administrative record on which the order of removal is based.”).
But introducing such documents into the record below would have been futile. By the
time Matthews was ordered removed by the Immigration Judge in April 2016 based on
his two aged § 260.10(1) convictions, the BIA had already decided in Mendoza Osorio
that the New York statute was categorically a crime of child abuse. And in Mendoza
Osorio—a decision that we did not have the opportunity to review directly—the BIA
was presented with, and refused to consider, similar charging documents. 26 I. & N.
Dec. at 707 n.4. Although they were not incorporated into the administrative record, I
see no unfairness or impropriety in considering their implications. The BIA was familiar
with this public record evidence from the proceedings in the context of its Mendoza
Osorio decision, and the documents do not concern factual matters on which credibility
was an issue or testimony appropriate.

21
has been used to charge conduct less than the generic INA definition of a crime

of child abuse.13

In sum, the charging documents adduced by amici illustrate further that,

in addition to the “home alone” cases discussed above, New York State pursues

prosecutions and secures convictions under § 260.10(1) in situations that, while

doing no credit to the individuals convicted, “‘creat[e] only the bare potential for

nonserious harm to a child.’” Mendoza Osorio, 26 I. & N. Dec. at 711 (quoting

Fregozo, 576 F.3d at 1038). This, too, reinforces the conclusion that Matthews has

demonstrated a “realistic probability” that the statute is more broadly applied

than the INA offense.

II.

Beyond its unduly constrained analysis of the realities of NYPL § 260.10(1),

the BIA’s holding that Matthews is removable suffers from another troubling

flaw. The agency’s determination to apply its categorical equation of New York

child endangerment with the INA definition of “child abuse” retroactively to

Matthews and to others, newly rendering them deportable, can be expected to

have a devastating effect on those who pleaded guilty to the New York

13Although it did not survive appellate review, one particular reported conviction also
tends to confirm the impression created by amici’s submissions that the statute is
charged broadly and on occasion idiosyncratically. In this odd case, a horse‐farm owner
was convicted after a bench trial of violating § 260.10(1) “after she injected a dog with a
tranquilizer in the presence of a child.” People v. Kanciper, 954 N.Y.S.2d 146, 147 (2d
Dep’t 2012). The case history demonstrates that a prosecutor may charge (and secure
the conviction of) an individual on allegations where the harm to the child is highly
speculative.
22
misdemeanor years ago, understanding then that the likely consequence would

be only probation or conditional discharge. The parties did not pursue the

implications of this observation in their briefs on appeal here. But the agency has

not proposed that its equation of New York law and the INA’s generic definition

has only prospective force, and its actions with regard to Matthews surely

demonstrate the contrary.

We recently re‐affirmed the following statement of five factors that we

“weigh . . . to determine whether an agency may apply a new rule retroactively”:

(1) whether the case is one of first impression, (2) whether the new
rule presents an abrupt departure from well‐established practice or
merely attempts to fill a void in an unsettled area of law, (3) the
extent to which the party against whom the new rule is applied
relied on the former rule, (4) the degree of the burden which a
retroactive order places on a party, and (5) the statutory interest in
applying a new rule despite the reliance of a party on the old
standard.

Obeya v. Sessions, 884 F.3d 442, 445 (2d Cir. 2018) (quoting Lugo v. Holder, 783 F.3d

119, 121 (2d Cir. 2015)). In Obeya, we also clarified that, “when conducting

retroactivity analysis in the immigration context, we look to whether it would

have been reasonable for a criminal defendant to rely on the immigration rules in

effect at the time that he or she entered a guilty plea.” Id. at 448; see also Velasquez‐

Garcia v. Holder, 760 F.3d 571, 582 (7th Cir. 2014) (applying similar retroactivity

analysis). In the case of individuals who entered guilty pleas in New York under

§ 260.10(1) before today—and especially those who did so before Soram was

decided—it is easy to conclude that they could have reasonably relied on the

23
“immigration rules in effect at the time,” in Obeya’s phrase, and expected no

immigration consequences at all from their guilty plea to misdemeanor child

endangerment that garnered a sentence of probation or conditional discharge.

Here, as in Obeya, the first and fourth Lugo factors “are not seriously at

issue in the case before us”; both favor Matthews. 884 F.3d at 445. As set forth in

Part I, the interpretation of the INA’s generic definition of “child abuse” is not a

matter of first impression for the BIA, and the burden of retroactive

application—removal from the United States—is drastic. Accordingly, here, also

as in Obeya, the “heart of this case rests with the second and third Lugo factors,”

that is, whether Soram (and Mendoza Osorio) marked “abrupt departure[s]” from

the BIA’s “well‐established practice,” and whether Matthews reasonably relied

on the “former rule.” Id.

Between 1999 and 2008, as observed in Part I, the BIA’s sole statement

concerning the generic federal definition was that it encompassed “[a]ny form of

cruelty to a child’s physical, moral or mental well‐being.” Rodriguez‐Rodriguez, 22

I. & N. Dec. at 996; see also Velazquez‐Herrera v. Gonzales, 466 F.3d 781, 783 (9th Cir.

2006) (“The source Rodriguez‐Rodriguez consulted, Black’s Law Dictionary 405

(8th ed. 2004), defines ‘cruelty’ as ‘[t]he intentional and malicious infliction of

mental or physical suffering on a living creature.’”). Then, in 2008, the BIA

lowered the requisite mens rea to criminal negligence. See Velazquez‐Herrera, 24 I.

& N. Dec. at 511–12 n.13 (repudiating Rodriguez‐Rodriguez’s reliance on Black’s

Law Dictionary’s definition of “cruelty,” and holding that under “the weight of

Federal and State authority in effect in 1996 . . . criminally negligent acts

24
sufficed”). But even Velazquez‐Herrera retained the suggestion that “the infliction

on a child of . . . harm, even if slight” was a necessary element of the generic

offense. Id. at 512. Only in 2010, in Soram, did the BIA finally state its current

view that “actual harm or injury to the child” was not required as an element of

the generic crime. 25 I. & N. Dec. at 381.

In this context, where the BIA has materially changed its interpretation of

the generic offense, the second Lugo factor—whether there has been an “abrupt

departure”—weighs in favor of Matthews. Consequently, because “[t]here can be

little doubt that, as a general matter, alien defendants considering whether to

enter into a plea agreement are acutely aware of the immigration consequences

of their convictions,” INS v. St. Cyr, 533 U.S. 289, 322 (2001), the third Lugo factor

favors Matthews as well. It is fair, then, to conclude that Matthews’s 2002 and

2003 convictions should not render him deportable under the contemporaneous

BIA interpretation, on which he could have reasonably relied.

Accordingly, even if I agreed with the merits of the BIA’s recent alignment

of § 260.10(1) and the generic federal definition, I would still hold that, under

Obeya and Lugo, this rule could not be applied retroactively to Matthews and

those similarly situated. I would grant his petition for review on this ground as

well. The majority does not address this question and its opinion should not

foreclose other individuals from challenging the BIA’s retroactive application of

its definition to them.

25
 III.

Even apart from retroactivity concerns, the implications of the BIA’s

interpretation of § 260.10(1) that the majority affirms here today reach well

beyond Matthews’s case. In addition to being at odds with the “realistic

probability” analysis, the holding leads to harsh results that I believe are unlikely

to have been intended by Congress.

As the BIA recounted in Matter of Velazquez‐Herrera, when passing the

statute that added “a crime of child abuse, child neglect, or child abandonment”

to the list of offenses rendering an individual deportable, Congress “clearly

intended to single out those who have been convicted of maltreating or preying

upon children.” 24 I. & N. Dec. at 509. We have been alerted to no basis for

concluding that, in so doing, Congress also intended to render deportable parents

who, in a possible violation of a state child endangerment statute and in a

display of bad judgment, leave their child alone briefly, commit minor offenses

in the presence of a child, or engage in other conduct that might allow for the

possibility of relatively minor harm befalling a child. Surely such acts fall outside

of the general understanding of “child abuse,” even broadly interpreted.

Indeed, the BIA has adduced no evidence that Congress (in enacting 8

U.S.C. § 1227(a)(2)(E)(i)), the New York legislature (in enacting § 260.10(1)), or

the New York judiciary (in leniently sentencing § 260.10(1) violations) desired or

even recognized that permanent family separation could be triggered by such

lapses in judgment in childrearing, as blameworthy as they may be. See Martinez‐

Cedillo, 896 F.3d at 1002 (Wardlaw, J., dissenting) (“It should not be lost on us

26
that, while we fault [Respondent] for endangering his son, we simultaneously

condone the separation of a family, exiling a father of two children who has

resided in the United States lawfully for more than twenty‐five years. That

Congress did not intend such a result is apparent from these facts.”).14

Paradoxically, it is children who will suffer harm under the agency’s

interpretation of § 260.10(1)—a law intended for their protection. In addition to

subjecting lawful permanent residents to deportation, under the agency’s rule,

non‐permanent residents become not only removable but also statutorily

ineligible for cancellation of removal because of such a conviction. See 8 U.S.C.

§ 1229b(b)(1)(C) (excluding aliens “convicted of an offense under section . . .

1227(a)(2)” from such relief). Even in cases where a non‐permanent resident can

otherwise prove that removal would result in exceptional and extremely unusual

hardship to U.S. citizen or lawfully resident children, the agency’s position,

affirmed by the majority, will compel the immigration courts to deny them relief.

See id. § 1229b(b)(1)(D); Matter of Monreal, 23 I. & N. Dec. 56, 62 (BIA 2001) (to

qualify for such relief, “the hardship to an alien’s [children], if the alien is obliged

to leave the United States, must be ‘substantially’ beyond the ordinary hardship

that would be expected when a close family member leaves this country”). It

requires no exercise of judicial imagination to conclude that a child will be

harmed by being separated from a parent, who, even if imperfect, is generally

competent and caring.

14 See note 5, supra, for Martinez‐Cedillo’s subsequent appellate history.
27
 I am also concerned that another vulnerable group will be harmed by this

result as well—victims of domestic violence. Under the BIA’s interpretation, a

conviction under § 260.10(1) will also render a noncitizen ineligible for

cancellation of removal under the “battered spouse or child” provisions of

VAWA. See 8 U.S.C. § 1229b(b)(2)(A)(iv) (excluding individuals deportable under

“paragraphs . . . (2) through (4) of section 1227(a)” from relief). This provision

allows an immigration court, subject to certain conditions, to grant cancellation

of removal to individuals who have been “battered or subjected to extreme

cruelty” by a spouse who is a U.S. citizen or lawful permanent resident. Id.

§ 1229b(b)(2)(A). The statute’s text and legislative history offer strong signals that

Congress intended VAWA’s immigration‐related provisions to serve as a safety

valve aimed at preventing victims of domestic violence who depend on a spouse

for immigration status from becoming trapped in violent relationships.15 It seems

extremely unlikely that Congress intended that domestic violence victims

previously convicted of a minor child endangerment misdemeanor be barred

15The Report of the House Committee on the Judiciary accompanying the legislation
explained:

Domestic battery problems can become terribly exacerbated in marriages
where one spouse is not a citizen, and the non‐citizen[’]s legal status
depends on his or her marriage to the abuser. Current law fosters
domestic violence in such situations by placing full and complete control
of the alien spouse’s ability to gain permanent legal status in the hands
of the citizen or lawful permanent resident spouse. . . . Consequently, a
battered spouse may be deterred from taking action to protect himself or
herself, such as filing for a civil protection order, filing criminal charges,
or calling the police, because of the threat or fear of deportation.

H. R. Rep. 103‐395 at 26, 1993 WL 484760 (1993).
28
from relief when it created VAWA’s immigration‐related remedies, years before

the BIA’s definition of a crime of “child abuse, child neglect, or child

abandonment” expanded in Soram, in 2010, to reach its current capacious state.

In sum, the majority affirms an agency ruling that inflicts an array of

troubling collateral consequences on a broad class of non‐citizens. That Congress

so intended, sub silentio, by making a crime of “child abuse” a removable offense

strikes me as implausible. These considerations, too, support the conclusion

reached in Parts I and II that Matthews’s petition for review should be granted.

* * *

Our country’s immigration laws address the tension between promoting

public safety by removing non‐citizens who have violated our criminal laws in a

domestic setting, on one hand, and pursuing the “underlying intention . . .

regarding the preservation of the family unit,” Nwozuzu v. Holder, 726 F.3d 323,

332 (2d Cir. 2013) (quoting H.R. Rep. No. 82‐1365 (1952)), on the other. We have

acknowledged that one of the law’s goals is “to keep families intact where

possible.” Nwozuzu, 726 F.3d at 332; see also Ibarra v. Holder, 736 F.3d 903, 918 n.20

(10th Cir. 2013) (declining to defer to the BIA’s ruling in Soram and declaring,

“[O]ne of the purposes of the INA is ‘keeping families of United States citizens

and immigrants united,’ not just deporting people.” (internal citation omitted)).

In its incorrect ruling that New York misdemeanor child endangerment

law is equivalent to the INA’s generic definition of child abuse, the BIA ignores

the real‐world application of the New York statute. It also averts its eyes from the

29
real‐world effect of its decision: the needless and potentially permanent

separation of children from their parents. Retroactive application of the agency’s

ruling will run afoul of the principles that we established in Lugo and Obeya. Its

prospective application will inflict needless suffering on some of the most

vulnerable members of our society.

For these reasons, I respectfully dissent.

30