Policy uncertainty and international financial markets: the case of Brexit

ABSTRACT

This study assesses the impact of the Brexit probability on both the UK and on international financial markets, for the first and the second statistical moments. As financial markets are by nature highly interlinked, one might expect that the uncertainty engendered by Brexit also has an impact on financial markets in several other countries. We first estimate the time-varying interactions between UK policy uncertainty, which to a large extent is attributed to uncertainty about Brexit and UK financial market volatilities. Second, we use two other measures of the perceived probability of Brexit before the referendum, namely daily data released by Betfair and results of polls published by Bloomberg. Based on these data sets, and using both panel and single-country SUR estimation methods, we analyse the Brexit effect on levels of stock returns, sovereign CDS, 10-year interest rates in 19 predominantly European countries, and those of the British pound and the euro. We show that Brexit-induced policy uncertainty will continue to cause instability in key financial markets and has the potential to damage the real economy in both the UK and other European countries. The main losers outside the UK are the ‘GIIPS’ economies: Greece, Ireland, Italy, Portugal and Spain.

KEYWORDS:

• Brexit
• causality tests
• financial instability
• uncertainty
• spillovers

JEL CLASSIFICATION:

• E44
• F36
• G15

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ For a survey of the related arguments, see, for instance, London School of Economics (2016).

² In the following, we do not discuss the various arguments surrounding immigration and fiscal costs. For a broad survey on the potential economic impacts of Brexit, see International Monetary Fund (2016).

³ The UK is the world leader in fixed-income and derivatives transactions and far ahead of EU peers in private equity, hedge funds and cross-border bank lending (Bank of England 2016). The UK’s insurance industry is the largest in Europe and the third largest in the world.

⁴ An alternative might be the Norwegian model (EEA) or the Swiss model.

⁵ This view is backed by empirical results underscoring that the reduction in trade barriers due to EU membership has increased UK incomes (Crafts 2016; Campos, Coricelli, and Moretti 2014).

⁶ LSE (2016) concludes that the UK is already deregulated and a more skilled workforce and a better infrastructure are more potent sources of further productivity enhancements.

⁷ Gros (2016), however, puts the assessment of the literature reviewed in Section II into perspective and states: ‘(b)eyond a weaker pound and lower UK interest rates, the referendum has not had much of a lasting impact. Financial markets wobbled for a few weeks after the referendum, but have since recovered. Consumer spending remains unmoved’. While it is true that consumer spending stayed rather untouched, we report in this section that business and consumer confidence went down. See also our remarks in Section IV.

⁸ See http://www.policyuncertainty.com/index.html.

⁹ For more details about the construction of daily volatilities, please refer to Alizadeh, Brandt, and Diebold (2002).

¹⁰ We have decided to consider FTSE 250 prices instead of FTSE 100 since the first might be a better gauge of domestically oriented share prices than the FTSE 100 which is dominated by multinationals of which some have little exposure to the UK economy (Sheffield 2016).

¹¹ In cases when the index was equal to 0, we have replaced it with the value from the previous day.

¹² Empirical realizations of the VIX index, intraday high and low values of FTSE250 and the GBP/USD exchange rate are obtained from the Datastream database.

¹³ Alternatively, Hafner and Herwartz (2006b) propose a concept of impulse response functions tracing the effects of independent shocks on volatility and then consider the effect of historical shocks, such as ‘Black Wednesday’ and an announcement by the European Community finance ministers on 2 August 1993, on the foreign exchange market. However, we believe that the identification of a ‘Brexit shock’ is not trivial and should not be constrained only to the day of the announcement of the referendum results but should include the days preceding the referendum as well. Moreover, the approach taken in our article allows us to take into account the time-varying volatility of multivariate financial time series.

¹⁴ VAR specification tests are plausible and available upon request. Different Cholesky orderings do not change the signs and the significance of the impulse responses.

¹⁵ For the rolling estimations, we have set a rolling window of 500 working days and a forecast horizon of 10 working days. As robustness check, we performed estimations with different lag length, rolling windows and forecast horizons – the results remain and are available upon request.

¹⁶ For an overview of existing studies dealing with the interlinkages between stock prices and exchange rates, refer to Caporale, Hunter and Menla Ali (2014).

¹⁷ Additionally, we performed several estimations using 12 pm (GMT) values and obtained nearly identical results.

¹⁸ Further information can be found here: http://www.bloomberg.com/graphics/2016-brexit-watch/.

¹⁹ The values are based on the search topic: ‘United Kingdom European Union membership referendum, 2016’, which combines several different research requests corresponding with the Brexit topic. The following additional options are used: Search Category: ‘News’, Search: ‘News-Search’.

²⁰ For a detailed derivation of this aspect, see Pankratz (1991).

²¹ The same argument also applies to the other estimations in this section.

²² In August 2016, the BoE decreased the bank rate to 0.25% justifying its decision by potential effects of the Brexit vote on future inflation and growth.

²³ The GIIPS states comprise Greece, Ireland, Italy, Portugal and Spain.

²⁴ Ratings are taken from Fitch Ratings. The AAA group contains Canada, Denmark, Germany, The Netherlands, Norway, Sweden, Switzerland and the United States. The second group contains only the so-called GIIPS states.

²⁵ In case of the Euro, we take German 10-year yields as a proxy of the ‘European’ interest rate. However, we do not find different results when Dutch, French or Finnish yields are used.