The 2016 United Kingdom EU membership referendum

The United Kingdom’s referendum of June 23, 2016 asked “Should the United Kingdom remain a member of the European Union or leave the European Union?”, and 17,410,742 voters (51.9%) chose Leave against 16,141,241 (48.1%) for Remain. The European Union Referendum Act 2015 that authorized the vote did not say what would follow, so the Supreme Court held in January 2017 that an Act of Parliament was needed before the government could begin withdrawal; the UK left on January 31, 2020. This page covers how the vote was designed and administered and what each side argues it teaches, not whether leaving was the right choice.

In 2016, voters across the UK were asked whether to stay in the European Union or leave. More people chose Leave. But the law that set up the vote did not say what had to happen next. That led to years of argument in Parliament and the courts about how, when, and on what terms to leave. Questions about campaign spending and public money also came up on both sides.

For other referendums in this collection, see the Scottish independence referendum, New Zealand’s 2020 referendums and the comparative overview. For general problems with referendum design, see question design.

Defined terms

The 2016 referendum involves several terms with specific legal meanings.

• Referendum Act: the European Union Referendum Act 2015, which received Royal Assent on December 17, 2015.⁠²
• Article 50: the provision of the Treaty on European Union under which a member state that decides to withdraw “in accordance with its own constitutional requirements” notifies the EU, after which the EU treaties “shall cease to apply” to that state within two years.⁠⁵
• Prerogative power: a power exercised by ministers on behalf of the Crown without a specific Act of Parliament; the government proposed to use prerogative powers to give notice under Article 50.⁠⁵
• Advisory versus binding: a binding referendum changes the law or requires a specified step by its own terms; an advisory one does not. The Supreme Court held that the 2015 Act “simply provided for the referendum to be held without specifying the consequences”.⁠⁵
• Designated lead campaigner: the group the Electoral Commission recognizes as the lead campaign for one outcome, eligible for a public grant of up to £600,000 and broadcasts if both sides are designated.⁠³
• Purdah (section 125): restrictions on government publication in the period before a referendum poll, under section 125 of the 2000 Act, which the 2015 Act allowed ministers to modify by regulation.⁠³

What happened

The 2016 referendum followed a dated sequence of legislative, electoral and judicial steps.

• May 27, 2015. The Queen’s Speech announced a bill “to hold a referendum on whether the United Kingdom should remain a member of the European Union,” to take place no later than December 31, 2017.⁠⁴
• September 1, 2015. The government accepted the Electoral Commission’s recommendation to replace a yes/no question with a Remain/Leave choice.⁠^6,7
• December 17, 2015. The European Union Referendum Act 2015 received Royal Assent.⁠²
• April 6, 2016. The government announced a leaflet setting out its view, to be delivered to over 27 million homes at a total cost of £9.3 million.⁠^8,9
• May 23, 2016. HM Treasury published an analysis of the immediate economic impact of a vote to leave.⁠¹⁰
• May 27, 2016. The UK Statistics Authority’s chair publicly criticized continued use of a “£350 million” weekly figure for UK contributions to the EU.⁠¹¹
• June 23, 2016. Polling day: Leave 51.9%, Remain 48.1%.⁠¹
• January 24, 2017. The Supreme Court, 8 to 3, held that an Act of Parliament was required to authorize notice under Article 50.⁠⁵
• 2017. Parliament passed the European Union (Notification of Withdrawal) Act 2017, section 1(1) of which provides that “The Prime Minister may notify, under Article 50(2) of the Treaty on European Union, the United Kingdom’s intention” to withdraw.⁠¹²
• March 29, 2017. The Prime Minister’s letter to the President of the European Council notified the UK’s intention to leave.⁠¹³
• July 17, 2018. The Electoral Commission announced that it had found “significant evidence of undeclared joint working” between Vote Leave and BeLeave, fined both, and referred matters to the police.⁠¹⁴
• July 2019. BeLeave’s founder won his appeal against a £20,000 Electoral Commission fine in the Mayor’s and City of London Court.⁠¹⁵
• September 24, 2019. The Supreme Court unanimously held that the Prime Minister’s advice to prorogue Parliament was unlawful and void.⁠¹⁶
• January 31, 2020, 11:00 p.m. “Exit day” under the European Union (Withdrawal) Act 2018 as amended.⁠¹⁷

How it worked

The 2016 referendum’s mechanics were set by the 2015 Act and the general referendum law of 2000.

The question. The bill originally proposed a yes/no question. The Electoral Commission tested options with voters and recommended the Remain/Leave form.⁠⁷ As reported by a political blog that reproduced its statement, the Commission found that voters understood the original wording but that some campaigners and members of the public perceived it as biased because it “only sets out the ‘remain’ option,” a perception it said could affect the result’s legitimacy for those campaigning to leave.⁠⁶ The Act fixed the final question and the two answers, “Remain a member of the European Union” and “Leave the European Union”.⁠²

The franchise. The Act’s explanatory notes describe the electorate as “British, Irish and Commonwealth citizens over the age of 18 who are resident in the United Kingdom,” together with people entitled to vote in parliamentary elections, certain peers, and eligible voters in Gibraltar.⁠^3,18 The franchise followed the parliamentary model rather than the local-election model, so most citizens of other EU states living in the UK could not vote, and 16 and 17 year olds were not included.⁠³ Both choices were debated; supporters said a decision on national membership should use the national parliamentary franchise, and critics said people most directly affected were excluded.

Campaign rules. The Commission could designate a lead campaigner for each side, eligible for a grant of up to £600,000; spending limits from the 2000 Act applied, uprated for inflation.⁠³ Vote Leave was the designated lead campaigner for Leave, subject to a £7 million spending cap.⁠^14,15

Government communication. The government took a position and spent public money to explain it. Its April 2016 press release gave the leaflet’s cost as £458,500 for production, £5,947,436 for printing and delivery, and £2,894,064 for digital promotion and the website, and justified it by citing polling that “85% of the public want more information from the government itself”.⁠⁹ The government had planned to suspend purdah restrictions, which drew a rebellion by 27 Conservative MPs in June 2015, and in September 2015 said it would restore them with some exceptions.⁠^6,7

Legal effect. The Act contained no provision on the consequences of either result; challenges to the count had to be brought by judicial review within six weeks.⁠³ The Supreme Court later contrasted it with referendum statutes that “stipulated what should happen in response to the referendum result,” and observed that the 2016 referendum “was described as advisory by some ministers and as decisive by others”.⁠⁵

Evidence: measured results

The 2016 referendum’s official totals, as compiled by IFES from official returns, were:⁠¹

Votes cast divided by registered voters is about 72.2%; that ratio is calculated here from the IFES figures.⁠¹ Results differed across the UK’s nations and regions; those breakdowns were not verified for this page.

Who voted which way. Sara Hobolt’s analysis of campaign and survey data, published in the Journal of European Public Policy in 2016, found that support for leaving was higher among less educated, lower income and older voters, and among those concerned about immigration and multiculturalism, and argued that a divide between those who felt they had gained and lost from globalization was a major driver.⁠¹⁹ Hobolt’s interpretation is one of several; the statistical associations it reports are not by themselves explanations of individual motives.

Economic forecasts. HM Treasury’s May 2016 analysis projected that a vote to leave would “push the UK into recession and lead to a sharp rise in unemployment,” estimating GDP 3.6% lower after two years in a “shock” scenario and 6% lower in a “severe shock” scenario, with unemployment up about 500,000 or 800,000.⁠¹⁰ How those projections compare with what followed is disputed, partly because the analysis modeled a scenario that did not match the actual timetable, and is listed below as an unresolved question rather than settled here.

Regulatory findings. The Electoral Commission found that Vote Leave exceeded its £7 million cap through undeclared joint working with BeLeave, including a donation of about £680,000 used to buy digital advertising from AggregateIQ.⁠^14,15 The court later allowed BeLeave’s founder’s appeal against his personal £20,000 fine; the Commission said it was “disappointed” and would review the judgment.⁠¹⁵

Failures and criticisms

The 2016 referendum is criticized from several directions, and the criticisms are attributed here rather than endorsed.

From those who favored Remain or questioned the process. Critics point to the Electoral Commission’s findings against the lead Leave campaign.⁠¹⁴ and to the UK Statistics Authority’s statement that it was “disappointed” that suggestions continued that “the UK contributes £350 million to the EU each week,” having earlier described the gross figure as potentially misleading because it did not account for the rebate⁠¹¹ Some also argue that a constitutional change of this size should have required a supermajority or a second confirmatory vote; the Act required neither.

From those who favored Leave or questioned the government’s role. Critics point to the government spending £9.3 million of public money on a leaflet advocating one side,⁠⁹ and to official economic forecasts of recession,⁠¹⁰ that they characterized as “Project Fear.” They also point to the BeLeave founder’s successful appeal⁠¹⁵ as evidence that the regulator overreached, and argue that Parliament’s later conduct, including attempts to delay or reshape withdrawal, resisted a clear democratic instruction.

From constitutional scholars and the courts. The most widely shared criticism is structural: the statute left the consequences of a Leave vote undefined. The Supreme Court held that “the change in the law required to implement the referendum’s outcome must be made in the only way permitted by the UK constitution, namely by legislation”⁠⁵ and in 2019 it held the prorogation of Parliament unlawful.⁠¹⁶ Supporters of the outcome read these cases as courts intervening in politics; others read them as the ordinary constitution filling a gap the Act left open.

On the franchise. Excluding most resident EU citizens and 16 and 17 year olds,⁠³ drew criticism from those affected; defenders cite consistency with general elections.

Competing interpretations

Readers draw at least three different lessons from the 2016 referendum.

A clear mandate resisted. On this reading, voters gave a direct answer to a plain question, and the three and a half years between the vote and exit reflect reluctance in Parliament and parts of the establishment to accept it. The size of the vote, 17.4 million for Leave⁠¹ is central to this view.

A narrow vote on an undefined proposition. On this reading, “Leave” named a destination without a route: the ballot did not specify a future relationship, so the conflict after 2016 was over what the vote meant, not whether to honor it. The Supreme Court’s observation that ministers described the vote inconsistently as advisory and decisive.⁠⁵ is often cited.

A test of campaign regulation. On this reading, the referendum exposed weak enforcement tools: findings came two years after the vote, a fine was overturned on appeal, and no result can be revisited for campaign breaches⁠^14,15 People on both sides make versions of this argument, focused on different conduct.

What transferred and what did not

The 2016 referendum influenced debates about referendum design in the UK and elsewhere.

Transferred. The Electoral Commission’s practice of testing questions with voters and recommending balanced wording.⁠⁷ follows the approach it took for Scotland in 2014; see the Scottish independence referendum. Writing the consequences of a result into the authorizing law is the approach New Zealand used for its 2020 End of Life Choice referendum; see New Zealand’s 2020 referendums.

Did not transfer. Hobolt found no sign of a short-term spread of similar membership referendums to other EU states⁠¹⁹

Unresolved in the UK. The sources consulted here do not establish any statutory change to UK referendum law made in direct response to 2016.

Application to this framework

The 2016 referendum bears on several principles of My Vote Forward.

• A ballot item needs a defined consequence. The framework distinguishes an issue, a proposal and a ballot item, and asks every proposal to state what happens if it passes; see proposals. The 2015 Act’s silence on consequences⁠⁵ is the clearest real example of what goes wrong without that.
• Balanced wording matters. The change from yes/no to Remain/Leave⁠^6,7 is a case study for question design.
• Who may participate is a design choice with consequences. See identity and registration and jurisdictions.
• Official information and advocacy are different. A government leaflet stating a position⁠⁹ raises the same issue the framework addresses when it separates neutral releases from advocacy; see published releases and governance.
• Priorities and ballots are separate. A one-day yes/no choice cannot show how strongly people hold a view or what they would trade for it. The framework’s separation of priorities, opinions and private voting responds to that limit without claiming to resolve it.

Unresolved questions

• How should a referendum statute define consequences when the winning option can be implemented in several very different ways?
• How do the Treasury’s 2016 projections compare with measured outcomes once differences in timing and scenario are accounted for? Assessments differ and were not adjudicated for this page.
• Should campaign finance breaches discovered after a referendum have any remedy beyond fines?
• What franchise is most defensible for a decision about a country’s relationship with other states?
• How did results vary by nation and region, and how should multi-nation states weigh such differences? Official breakdowns were not verified for this page.