HOUSE_OVERSIGHT_010610.jpg

We estimate a regression of weekly returns on the MXN spot exchange rate on five variables: a basket of commodity currencies, US equities, the 10-year US bond yield, the two-year interest rate differential between MXN and USD swaps, and RealClearPolitics polling average for Donald Trump in the 2016 US presidential election. The estimated equation is RON = bo + by AR +b *RPP™ + bstdy” + bat(™ — °°?) + bs*RCP + @ where A” is the MXN spot return, A" is the commodity currency basket spot return, RP” is the SP500 index return, dy’’ is the 10-year US bond yield change, “ - 7°’ is the two-year swap rate differential (lagged one period), ACP is RealClearPolitics polling average for Donald Trump, and eis an error term. The commodity currency basket consists of BRL, CLP, COP, CAD, NOK and AUD with equal weights. The full data sample has weekly frequency and runs from 6/20/2012 to 9/21/2016. The model without Trump yields an R-squared coefficient of 0.50. Including Trump’s variable increases the coefficient to 0.63, a significant improvement (Chart 2). We use data only up to 1/27/2016 for parameter estimation of the financial variables. We use the full sample to estimate the parameter on Trump’s variable due to its shorter history. Chart 1: The Trump factor on the Mexican peso Chart 2: Forecasting MXN returns with and without Trump 20 48 . oO on ay tory a) fo} NO fo} » . co p BR . » Ny is) . . ss us ow fF ann oOo © . us » x Now fk aon oO © . w NR ry » ry a as ug15 Nov15 Feb16 May16 Aug16 2014 2015 2016 ——MXN -—=——=Trump's polling average ——— Model (without Trump) -—===Model (with Trump) ===~=-MXN Source: BofA Merrill Lynch Global Research, Bloomberg Source: BofA Merrill Lynch Global Research, Bloomberg A possible 15% negative shock on MXN According to the model that does not include Trump’s polling average, we estimate that MXN should be trading at around 16.50, more than 15% stronger than current levels (spot 19.72). Chart 2 shows the possible magnitude of the Trump effect on the Mexican peso. The model including Trump’s polling average appears to do a much better job at explaining the recent behavior of the Mexican peso. However, it still suggests that the currency has depreciated more than justified by Trump’s increase in recent polls. Based on Trump’s increase to 43.9% (RealClearPolitics polling average on 21 September) from 39% (RealClearPolitics polling average on 29 June), our model suggests the Mexican peso should be trading at around 18.50, still almost 7% stronger than current levels. This, however, does not necessarily mean that MXN has oversold. The almost 5% gain in Trump’s polling average since June could have had an outsized effect on MXN if those gains took place in swing states, potentially leading to crucial Electoral College votes. Strategy: neutral MXN for now From a strategy standpoint, we recommend staying neutral the Mexican peso for now. We believe the Mexican peso would likely strengthen significantly if Hillary Clinton wins the US election. But there is still significant uncertainty. Net speculative long USD/MXN positions are somewhat stretched, but could increase much more (Chart 3). eee Bankof America 2 Liquid Insight | 22 September 2016 Merrill Lynch HOUSE_OVERSIGHT_010610