Watch out for the impact of Scottish independence opinion polls on UK s borrowing costs

Transcription

1 Wach ou for he impac of Scoish independence opinion polls on UK s borrowing coss Cosas Milas (Universiy of Liverpool; cosas.milas@liverpool.ac.uk) and Tim Worrall (Universiy of Edinburgh; im.worrall@ed.ac.uk) 1. Inroducion As he dae of he Scoish independence referendum approaches, he debae of boh sides of he campaign has increasingly focused on he economic consequences for an independen Scoland and he Res of he UK. Perhaps surprisingly, very lile (if any) aenion has been given o he implicaions of Scoish independence polls for curren economic developmens and in paricular wha Scoish opinion polls imply for curren borrowing coss in he UK. This noe ess he impac of opinion poll resuls on he spread beween he UK 10-year governmen bond yield and he UK 5-year governmen bond yield over and above he impac of oher economic fundamenals. We esimae ha a 1 percenage poin increase in he Yes raing relaive o he No raing increases he 10-year borrowing coss relaive o he 5 year borrowing coss by up o 4 basis poins; in fac, a smaller No lead leads o a bigger impac. Our mehodology, daa and empirical esimaes are repored in Secion. Secion 3 discusses our findings and concludes.. Mehodology, daa and empirical esimaes To es he impac of opinion poll resuls on he spread beween he UK 10-year governmen bond yield and he UK 5-year governmen bond yield over and above he impac of oher economic fundamenals, we esimae he following non-linear model using daily daa over he period from 31 January 01 o 6 March 014: UKspread _10 _ 5 year USspread _10 _ 5 year UKliquidiy _10 _ 5year 0 US liq * ( YES,1 _ YES, _ ) HMT _ nc dpoll Yes lead Yes lead HMT annou u (1) where UKspread_10_5year is he spread beween he UK 10-year governmen bond yield and he UK 5-year governmen bond yield and USspread_10_5year is he spread beween he US 10-year governmen bond yield and he US 5-year governmen bond yield. Liquidiy consideraions (risk) in he UK (denoed by UKliquidiy_10_5year ) are proxied by he difference beween he 10-year bidask spread and he 5-year bid-ask spread. dpoll is a dummy variable aking he value of 1 if a poll resul is released and 0 oherwise. The Yes_lead variable refers o he Yes lead in he opinion polls. Under he assumpion ha boh β YES,1 >0 and β YES, >0, his model suggess ha when a poll opinion resul is released, he impac of Yes_lead is posiive and increases wih he size of he Yes_lead

2 opinion poll resul. HMT_announc is a dummy variable aking he value of 1 on January , and 0 oherwise. I capures Treasury s pledge o guaranee all of Briain s deb even if Scoland voes o leave he UK. u are independen and idenically disribued shocks. UK bond yields are rerieved from he Bank of England s websie. US bond yields are rerieved from he websie of he Federal Reserve Bank of S. Louis and he bid-ask spreads come from Daasream. The UK polling repor liss 54 opinion polls conduced since he sar of 01 (available from: hp://ukpollingrepor.co.uk/scoish-independence-referendum). These show ha he Yes suppor has lagged behind he No suppor by an average of 15.5 percenage poins over he period. The Yes campaign poin ou ha gap beween Yes and No has narrowed and, in paricular, has halved from 4 poins in November 013 o 1 poins in March 014. Figure 1 plos he daa. Figure 1: UK and US daa, 31 January 01-6 March I I I II II II III III III IV IV IV I I I II II II III III III IV IV IV I I I UKspread_10_5_year (percenage poins; Lef Hand Side) USspread_10_5_year (percenage poins; Lef Hand Side) UKliquidiy_10_5_year (percenage poins; Lef Hand Side) YES_lead poll (percenage poins; Righ Hand Side) Noe: Liquidiy risk is defined in erms of he percenage Bid-Ask spread: Ask10 year Bid 10 year Ask5 year Bid 5year 100* 100* 0.5*( Ask Bid ) 0.5*( Ask Bid ) 10 year 10 year 5year 5year of he 10-year and 5-year UK governmen bond, respecively., where Ask and Bid refer o he Ask and Bid price

3 OLS esimaes of (1) are repored in Table 1. Table 1: OLS Esimaes of model (1). Sample: 31 January 01-6 March 014. Dependen Variable: UKspread_10_5year Mehod: Leas Squares HAC sandard errors & covariance (Barle kernel, Newey-Wes fixed bandwidh = ) Variable Coefficien Sd. Error -Saisic Prob. Inercep USspread_10_5year UKliquidiy_10_5year dpoll *Yes_lead dpoll *Yes_lead E HMT_announc R-squared Mean dependen var Adjused R-squared S.D. dependen var S.E. of regression Akaike info crierion Sum squared resid Schwarz crierion Log likelihood Hannan-Quinn crier F-saisic Durbin-Wason sa Prob(F-saisic) Wald F-saisic Prob(Wald F-saisic) Noe: Newey-Wes Auocorrelaion and Heeroskedasiciy robus sandard errors are repored. From Table 1, we esimae ha a 1 percenage poin increase in he Yes raing relaive o he No raing lifs he 10-year borrowing coss relaive o he 5-year borrowing coss by 3.6 basis poins. The impac is quie small and he -raio is only saisically significan a he 11% level (he quadraic erm is also posiive bu saisically significan only a he 17% level). Therefore, he evidence ha markes are pricing opinion poll resuls in Scoland (over and above he posiive impac of he US spread variable and he posiive impac of liquidiy) consideraions is weak. Noice also from Table 1 ha he spread beween he 10-year and he 5-year yield shrinks in response o Treasury s pledge (on January 13, 014) o guaranee all of Briain s deb even if Scoland voes o leave he UK. Model (1) assumes ha financial markes are aking noice of an opinion poll resul only when his is released. Anoher possibiliy is ha once an opinion poll resul is released, his is evaluaed by he markes unil he following opinion poll oucome becomes available. To accoun for his possibiliy, we esimae model () below

4 UKspread _10 _ 5 year USspread _10 _ 5 year UKliquidiy _10 _ 5year 0 US liq YES,1 _ YES, _ dpoll * ( Yes lead Yes lead ) YES,3 1 YES,4 1 (1 dpoll ) * ( Yes _ lead Yes _ Yes _ lead ) HMT _ announc v HMT () In his model, financial markes are aking noice of he Yes_lead poll resul in a regime-swiching manner. If dpoll =1 (ha is, when a new opinion poll resul becomes available) he impacs are β YES,1 and β YES,. On he oher hand, If, dpoll =0 (ha is, in he absence of a new poll), he impacs are β YES,3 and β YES,4. In his laer case, markes keep an eye on he previously released resul. OLS esimaes of () are repored in Table. Table : OLS Esimaes of model (). Sample: 31 January 01-6 March 014. Dependen Variable: UKspread_10_5year Mehod: Leas Squares HAC sandard errors & covariance (Barle kernel, Newey-Wes fixed bandwidh = ) Variable Coefficien Sd. Error -Saisic Prob. Inercep USspread_10_5year UKliquidiy_10_5year dpoll *Yes_lead dpoll *Yes_lead (1-dpoll )*Yes_lead (1-dpoll )*Yes_lead HMT_announc R-squared Mean dependen var Adjused R-squared S.D. dependen var S.E. of regression Akaike info crierion Sum squared resid Schwarz crierion Log likelihood Hannan-Quinn crier F-saisic Durbin-Wason sa Prob(F-saisic) Wald F-saisic Prob(Wald F-saisic) Noe: Newey-Wes Auocorrelaion and Heeroskedasiciy robus sandard errors are repored.

5 From Table, he β YES,1 and β YES, coefficiens are saisically significan. On he oher hand, here is weak evidence of saisical significance of β YES,3 and β YES,4 which suggess ha markes give prioriy o new informaion. Focussing on he linear par, our resuls sugges approximaely a ( ) basis poins 95% confidence inerval for β YES,1 when a poll is released (in which case dpoll =1). Therefore, a 1 percenage poin increase in he Yes raing relaive o he No raing lifs he 10-year borrowing coss relaive o he 5-year borrowing coss by up o 4 basis poins (noice also he impac rises wih he size of he Yes_lead resul). Model () fis he daa beer han model (1) as i has a higher adjused R and a lower regression sandard error. 3. Discussion and conclusions Wha do we learn from all hese? This research noe aemps o examine wheher he UK erm srucure is affeced by Scoish independence opinion poll resuls. We idenify a posiive impac from he poll variable on he UK erm srucure. Our resuls sugges ha a 1 percenage poin increase in he Yes raing relaive o he No raing lifs he 10-year borrowing coss relaive o he 5- year borrowing coss by up o 4 basis poins; i also rises wih he size of he Yes_lead poll resul. This link beween he UK s cos of borrowing and Scoish independence opinion polls has imporan implicaions for curren fiscal and moneary policy decisions. The Deb Managemen Office ( Deb and Reserves Managemen repor ) is planning gil sales of 18.4bn in A significan share (5.%) of hese sales is for bonds wih mauriies of beween 1 and 7 years while 1.0% of he share is accouned for by bonds wih mauriies of beween 7 and 15 years (he remainder are longer daed and indexed linked bonds). Assuming ha he 10-year yield is a proxy for he cos of bonds (wih mauriies beween 7 and 15 years), we believe ha fiscal auhoriies should be prepared o hedge agains he possibiliy of a sudden increase in he 10-year cos (implied by he Scoish opinion poll resuls) by purchasing shorer-erm bonds a he expense of longer-erm ones. Given curren plans o sell 6.9 billion in long-erm gils a par value, and assuming hese sales are of 10-year coupon bonds, a rise of 4 basis poins in he long-erm yield, increases he coupon paymens by approximaely 8.54% or abou 19.1 million per annum. Alhough his upper esimae is a small percenage of UK GDP, i is approximaely equal o he 130 million budge announced by he governmen o help o secure and mainain criical flood defences before nex winer his (hps:// In reviewing heir monhly Quaniaive Easing decisions, Bank of England s Moneary Policy Commiee members migh also wan o keep an eye on movemens in he Scoish opinion poll resuls and indeed be prepared o prioriise repurchases of longer-erm UK bonds in an aemp o keep a lid on he longer-erm cos of borrowing.