The Role of Economic Uncertainty in UK Stock Returns

We investigate the role of domestic and international economic uncertainty in the cross-sectional pricing of UK stocks. We consider a broad range of financial market variables in measuring financial conditions in order to obtain a better estimate of macroeconomic uncertainty compared to previous literature. In contrast to many earlier studies using conventional principal component analysis to estimate economic uncertainty, we construct new economic activity and inflation uncertainty indices for the UK using a time-varying parameter factor-augmented vector autoregressive (TVP-FAVAR) model. We then estimate stock sensitivity to a range of macroeconomic uncertainty indices and economic policy uncertainty indices. The evidence suggests that economic activity uncertainty and UK economic policy uncertainty have power in explaining the cross-section of UK stock returns, while UK inflation, EU economic policy and US economic policy uncertainty factors are not priced in stock returns for the UK.


Introduction and Literature Review
Our study investigates the role of economic uncertainty in stock returns in an asset pricing framework. Specifically, we study the UK stock market. After the global financial crisis from 2008, followed by serial crises in the Euro area and partisan policy disputes in the United States, there has been much debate on policy uncertainty.
For example, the Federal Open Market Committee (2009) and the IMF (2012IMF ( , 2013 suggest that economic recessions during the period 2007-9 and slow recoveries thereafter partly resulted from uncertainty about US and European monetary, fiscal and regulatory policies (see also Baker, Bloom and Davis (2016)). We are interested in examining investors' required rates of return on assets of varying sensitivity to uncertainty in response to this shifting economic uncertainty over time.
We examine stocks' sensitivity to economic uncertainty and study whether this sensitivity, or uncertainty risk, plays a role in predicting the future cross-section of stock returns in the UK. We estimate economic uncertainty in two aspectsmacroeconomic uncertainty and economic policy uncertainty (EPU). Many earlier macroeconomic uncertainty pricing studies such as Jurado, Ludvigson and Ng (2015) and Bali, Brown and Tang (2016) do not distinguish between output and inflation uncertainty. We consider uncertainty in the real macroeconomic environment as (i) economic activity uncertainty (EAU), also called output uncertainty and (ii) inflation uncertainty (IU), also called price uncertainty. We define macroeconomic uncertainty as the unforecastable component of output and inflation. We construct the EAU and IU indices ourselves using a time-varying parameter factor-augmented vector autoregressive (TVP-FAVAR) model.
To account for economic policy uncertainty (EPU), we employ the United Kingdom, United States and Euro Area economic policy indices (i.e., UK EPU, US EPU and EU EPU indices) of Baker et al. (2016) to investigate whether domestic and international economic policy uncertainty can be used to predict UK stock returns.
Based on newspaper coverage frequency, the Baker et al. (2016) EPU indices are developed to capture uncertainty about who will be economic policy decision makers, when and what economic policy will be implemented and what the economic effects of policy action (or inaction) will be. In other words, EPU indices differ from EAU and IU indices by focusing on shifts in economic policies rather than predicting Electronic copy available at: https://ssrn.com/abstract=3355648 macroeconomic indicators. An increase in policy uncertainty may not necessarily indicate greater difficulty in forecasting macroeconomic variables.
We then estimate stock sensitivity to the EAU index, IU index and three EPU indices and discover that the EAU and the UK EPU have power in explaining the cross-section of UK stock returns. Thus, our study not only provides stock market participants with new measures of macroeconomic uncertainty (i.e., our newly constructed EAU index and the IU index) but also presents theoretical and empirical support for incorporating economic uncertainty into investors' information sets in making investment decisions.
Traditional asset pricing models expect that average stock returns are linked to some well-known stock characteristics or risk factors, such as market, size, value, momentum and illiquidity risk factors (Jensen (1968), Fama and French (1993), Carhart (1997) and Pastor and Stambaugh (2003)). There is also some theoretical and empirical evidence that time variation in the conditional volatility of the unpredictable component of a wide range of economic indicators, i.e., macroeconomic shocks, is related to asset returns (Gomes, Kogan and Zhang (2003), Bloom (2009) and Jurado et al. (2015)). Motivated by this aforementioned evidence, Turan, Stephen and Tang (2016) quantify a macroeconomic uncertainty risk factor for the US stock market using the macroeconomic uncertainty index of Jurado et al. (2015).
Based on the inter-temporal capital asset pricing model (ICAPM) of Merton (1973) and Campbell (1993Campbell ( , 1996, an increase in economic uncertainty reduces future investment and consumption as investors may save more in order to hedge against potential future downturns in the economy. Simultaneously, investors are willing to hold stocks with higher inter-temporal correlation with economic uncertainty since the returns on these stocks will increase when economic uncertainty increases. Alternatively, as these stocks provide a natural hedge against economic uncertainty, they are willingly held by investors and hence have a lower required rate of return. In addition to the ICAPM framework, Ellsberg (1961) argues that when making investment decisions, investors consider not only the mean and variance of asset returns, but also the uncertainty of events which may influence the future return distribution. The experimental evidence in the Ellsberg (1961) study points out that it is important to distinguish between risk (i.e., variance) and uncertainty as people are Electronic copy available at: https://ssrn.com/abstract=3355648 more averse to unknown or ambiguous probabilities (i.e., uncertainty) rather than known probabilities (i.e., risk). Following Ellsberg (1961), studies such as Epstein and Wang (1994), Chen and Epstein (2002), Epstein and Schneider (2010) and Bianchi, Ilut and Schneider (2014), investigate the impact of economic uncertainty in asset pricing and portfolio choice. Their evidence demonstrates that investors require a higher premium to hold the market portfolio when they are uncertain about the correct probability law governing the market return. Based on all of the above discussions, economic uncertainty influences an investor's utility function and uncertainty-averse investors require an extra compensation, i.e., an uncertainty premium, to hold stocks with low covariance with economic uncertainty. An alternative explanation of this uncertainty premium is that stocks with high correlation with economic uncertainty would only attract low uncertainty-averse investors because relatively high uncertainty-averse investors tend to reduce or cease the investment in a stock if economic uncertainty is sufficiently high and investors' expectations about uncertainty are sufficiently dispersed. Thus, stocks with high covariance with economic uncertainty require a low uncertainty premium.
Motivated by the studies discussed above, Jurado et al. (2015)  i.e., the beta on the macroeconomic uncertainty index of Jurado et al. (2015) -after controlling for seven well-known risk factors 1 . They then sort individual stocks into decile portfolios by their uncertainty beta from low to high and find that the decile containing the lowest uncertainty beta stocks generates 6% more risk-adjusted return per annum than the decile with the highest beta stocks. The positive and highly significant spread between the alphas of the lowest and highest uncertainty beta portfolios suggests that (i) the macroeconomic uncertainty risk factor does have predictive power in the cross-sectional distribution of future US stock returns; (ii) when making investment decisions, the uncertain events of the future asset return distribution are also considered as well as the mean and variance of the asset returns; and (iii) uncertainty-averse investors demand a risk premium when holding stocks with negative uncertainty beta.
The Turan et al. (2016) study demonstrates that macroeconomic uncertainty risk factor plays a role in explaining the cross-section of future US stock returns.
However, the uncertainty index they employ is a factor-based estimation of economic uncertainty which selects over a hundred macroeconomic time-series (Jurado et al. 2015). It represents a rich data set of macroeconomic activity measures involving economic activity and inflation uncertainty. However, using all available information to extract factors is not always optimal in factor analysis (Boivin and Ng (2006), Koop and Korobilis (2014)). Moreover, the Turan et al. (2016) study does not distinguish between the role of economic activity and inflation uncertainty in stock return pricing. In addition, the index they selected ignores economic policy uncertainty. We examine two aspects of economic activity uncertainty as well as economic policy uncertainty separately.
Our study, addresses the aforementioned issues in three respects. First, we construct economic activity uncertainty (EAU) and inflation uncertainty (IU) indices including only variables that are theoretically justified in predicting the future economy 2 . Then, we estimate the uncertainty beta for each stock listed in the FTSE ALL Share index using 36-month rolling multivariate-regressions of excess returns on the existing risk factors (such as market, size, value, momentum and illiquidity) and also on the level of economic uncertainty in (i) UK inflation (IU), (ii) UK economic activity (EAU), (iii) UK economic policy (UK EPU), (iv) EU economic policy (EU EPU) or (v) US economic policy (US EPU). In each case, we sort stocks into portfolios by uncertainty sensitivity betas from low to high and examine whether there are return premia to post sorted uncertainty sensitive stocks.
After controlling for market, size, value and momentum risk factors of Fama and French (1993) and Carhart (1997) in both formation and holding periods, we find a statistically significant spread between the alphas of Decile 1 (i.e., lowest beta stocks) and 10 (highest beta stocks) sorted by the UK EPU beta and by the UK EAU beta.
When adding the illiquidity risk factors of O'Sullivan (2014, 2015) and O'Sullivan (2014, 2017), the EAU factor is still statistically significant while the UK EPU becomes insignificant for the large group of FTSE All Share stocks but remains significant for the subset of FTSE 250 stocks. This evidence suggests that our UK EPU and EAU risk factors further improve our understanding of stock pricing in the UK stock markets. To our knowledge, this paper is the first to incorporate economic policy uncertainty into stock pricing and is the first to estimate economic activity and inflation uncertainty risk factors for the UK stock market.
Our paper proceeds as follows. Section 2 describes the datasets employed in the study. The estimates of the economic activity uncertainty index and the inflation uncertainty index are presented in section 3. Section 4 and 5 discuss economic uncertainty pricing. Section 6 concludes.

Data and Variable Definitions
In this section, we describe our data set and explain the selection of financial variables to forecast economic activity and inflation. Our sample period is from January 1996 to December 2015. We restrict our uncertainty pricing analysis to UK stocks that were listed in the FTSE All Share index historically. Stock return data is taken from the London Share Price database (LSPD). The LSPD Archive file records historically when a stock was a constituent of the FTSE All Share, FTSE 250 and FTSE 100.
When estimating the economic uncertainty beta, we control for market, size, value, momentum and illiquidity risk factors. In our multifactor pricing models, the risk factor benchmark portfolios to proxy market, size, value and momentum risks are obtained from Xfi Centre for Finance and Investment (XiFI) University of Exeter and described in Gregory, Tharyan and Christidis (2013). Portfolios to proxy illiquidity risk are provided by and O'Sullivan (2014, 2017).

Market Portfolio and Risk Factors
We use the FTSE All Share index to proxy the market portfolio and the monthly return on three-month Treasury bills is taken as the risk free rate. The size factor benchmark, (small minus big stocks, SMB), is calculated by forming a portfolio each month that is short the upper 50% of the largest 350 firms in the FTSE All Share and long the remaining FTSE All Share stocks and holding for one month before reforming. The value factor benchmark portfolio, (high book-to-market minus low book-to-market stocks, HML), is calculated from the largest 350 firms in the FTSE All Share by each month forming a portfolio that is the monthly return on the highest 30% of stocks by book-to-market ratio (BTM) minus the monthly return on the lowest 30% of stocks by BTM and holding for one month. The momentum factor benchmark portfolio (MOM) is also formed from the largest 350 firms in the FTSE All Share monthly by ranking stock returns over the previous eleven months. A factor mimicking portfolio is constructed by going long the top performing 30% of stocks and short the worst performing 30% of stocks over the following month. All portfolios are value weighted using the market capitalisation of each stock, as discussed in Gregory et al. (2013).  provide evidence that characteristic illiquidity risk and systematic illiquidity risk are priced in UK stock returns. For this reason, we also add the  benchmark illiquidity factors to our factor models. The Electronic copy available at: https://ssrn.com/abstract=3355648 illiquidity characteristic mimicking portfolio is developed by sorting all stocks into decile portfolios based on their liquidity as measured by quoted spread. Equally weighted decile portfolio returns are calculated over the following one-month holding period and the process is repeated over a one-month rolling window. The illiquidity characteristic mimicking portfolio is the difference between the return of the top decile (low liquidity stocks) and bottom decile (high liquidity stocks. The benchmark portfolio of the systematic illiquidity factor is established by sorting stocks into equally-weighted decile portfolios according to their sensitivity to systematic (marketwide) liquidity. Portfolios are reformed every month, and the factor mimicking portfolio is constructed as the difference between the high-sensitivity and lowsensitivity portfolios.

Predictors of Inflation and Economic Activity
In constructing the macroeconomic uncertainty index, Jurado et al. (2015) assume a rich data environment and employ over one hundred macroeconomic series. Then they use principal component analysis to extract principal components which are used to forecast macroeconomic variables of interest. However, as demonstrated by Boivin and Ng (2006) and Koop and Korobilis (2014), using all available data to extract factors is not always optimal in principal component analysis. As mentioned, we divide macroeconomic uncertainty into activity uncertainty and inflation uncertainty.
In order to select variables to predict economic activity and inflation, we use the Bank of England's diagrammatic representation of the monetary policy transmission mechanism (June 2012).
[ Figure 1 here] As illustrated in Figure 1, monetary policy adjusts economic activity and inflation through financial markets. In the first stage, changes in monetary policy affects four groups of variables including market interest rates, asset prices, consumer and investor confidence and exchange rates, which in turn jointly influence economic activity. Then inflation is affected by shifts in both economic activity and the foreign exchange market. Therefore, by defining macroeconomic uncertainty as the uncertainty in forecasting economic activity and in forecasting the inflation rate, we opt to use financial variables that are most relevant in monetary policy transmission in Electronic copy available at: https://ssrn.com/abstract=3355648 the estimation of macroeconomic uncertainty. In other words, instead of employing 147 financial time series as in Jurado et al. (2015), we concentrate on variables that are theoretically well justified in predicting economic activity and inflation. Table 1 lists 45 indicators under eight categories that are used in estimating our macroeconomic uncertainty indices.
In Panel A Table 1 [ Table 1 here] Because some economic activity indicators including GDP are not available on a monthly basis, we use the growth rate of the real industrial production index (source: OECD) and the unemployment rate (source: Office for National Statistics) to measure real economic activity in the UK. Percentage changes in the retail price index, the producer price index and the consumer price index (source: DataStream) are employed as three inflation measures. In other words, we use the financial variables listed in Table 1 Panels A-F and the methodology discussed later to predict the five macroeconomic indicators in Panels G-H. The unforecastable components of the output indicators and that of the inflation indicators are considered as economics activity uncertainty (EAU) and inflation uncertainty (IU) respectively.

Economic Policy Uncertainty Indices
As already mentioned, in addition to investigating the pricing ability of output and inflation uncertainty, we are also interested in examining whether policy uncertainty can be used to price stocks return in the UK. We where is an × 1 vector of normalised financial variables which are included in Panel A-F Table 1, is a vector of economic activity and inflation proxies in Panel G-H Table 1 (2014), is calculated only using the observed indicators at time .
As mentioned in the data section, we use the industrial production index and the unemployment rate to measure economic activity and employ the retail price index, the producer price index and the customer price index to assess the price level. Rather than equally weighting all individual uncertainties as in Jurado et al. (2015), we distinguish between the role of economic activity uncertainty ( (ℎ)) and inflation uncertainty ( (ℎ)). We equally weight the two resulting activity indices and the two resulting inflation indices as follows: [ Figure 3 here]

Economic Uncertainty Pricing: the Method
We now turn to investigate the role of economic uncertainty in pricing UK stocks. As already mentioned, we use three measures of economic uncertainty including (i) economic policy uncertainty, (ii) economic activity uncertainty and (iii) inflation uncertainty. The measure of policy uncertainty is provided by Baker et al. (2016) based on newspaper coverage frequency, while the economic activity uncertainty index and the inflation uncertainty index are obtained from Section 3 using the TVP-FAVAR model. Our stock sample includes all common stocks which were in the FTSE All Share Index historically.
In the first step, we construct an economic uncertainty risk mimicking portfolio.
For each measure of economic uncertainty (i.e., UK EPU, EU EPU, US EPU, EAU and IU), each month individual stock (excess) returns are regressed on the economic uncertainty measure as well as other benchmark factors for market, size, value, momentum and illiquidity risks. We estimate this OLS regression over the previous 36 months based on stocks with a minimum of 18 observations. Then we sort stocks into fractile portfolios according to their uncertainty risk, i.e., the coefficient ( , uncertainty beta) on the measure of economic uncertainty: where is the relevant economic uncertainty measure, is a matrix of other risk factors for market, size, value, momentum and illiquidity risks and , is the excess return of stock over the risk-free rate. The subscripts denote time. To construct our risk mimicking portfolios, we assign stocks to a portfolio based on the estimated beta ̂, which measures a stock's sensitivity to the measure of Electronic copy available at: https://ssrn.com/abstract=3355648 economic uncertainty, in ascending order. It is worth noting that the value of ̂ may vary from negative to positive. In other words, Portfolio 1 contains stocks with the most negative ̂ while Portfolio 10 is constituted of stocks with the highest ̂. As explained by Bali et al. (2016) using the US data, the portfolio with the most negative beta is associated with the highest risk of economic uncertainty and hence uncertainty-averse investors demand a premium in the form of higher expected return to hold this portfolio and vice versa. We calculate each portfolio return as the equally weighted average return of its constituent stocks for the following month.
Portfolios are reformed monthly. The economic uncertainty risk mimicking portfolio is constructed as the difference between the 'low minus high' portfolios, (e.g., Decile 1 minus Decile 10).
In the second step, we estimate the alpha of the above risk mimicking portfolios in the following regression to examine whether the excess return of the low-beta portfolio over the high-beta portfolio can be explained by the existing risk factors (such as market, size, value, momentum and illiquidity risk factors). , = + 1 × , + 2 × + 3 × + 4 × + , [9] or , = + 1 × , + 2 × + 3 × + 4 × + 5 × + 6 × ℎ + , where , is the return on the low minus high portfolio, j  , j = 1,2…6 are the risk factor loadings and , , , , , and ℎ are the returns on the benchmark factor portfolios for market, size, value, momentum, systematic illiquidity and characteristic illiquidity risks respectively. Hence is a measure of return adjusted by the aforementioned risks and can be used as a test statistic to evaluate the predictive power of uncertainty risk.

Empirical Results: Is uncertainty priced?
If UK stocks are exposed to economic uncertainty risk and if this risk is systematic, i.e., difficult to diversify, investors would require a premium for holding economic uncertainty sensitive stocks. Consistent with the recent study of Bali et al. (2016) using US data, our results provide some evidence that uncertainty is also priced in stock returns in the UK. The results presented in Table 2-3 are obtained using the Carhart (1997)  In Panel A Table 2, we report findings on whether UK economic policy uncertainty is priced in stock returns. We find that for stocks in the FTSE All Share index, the low minus high EPU risk decile portfolios yields a four-factor alpha of 0.641% per month over the sample period (January 1997 -December 2015)significant at the 10% significance level. Alphas are also reported in Panel A for '1-15' and '1-5' portfolios where stocks are sorted into either 15 or 5 equally weighted portfolios respectively based on the beta on the UK EPU index. Interestingly, the '1-15' portfolio alpha is significant at 5% significance providing stronger evidence of pricing among stocks in the more extreme tails of the uncertainty sensitivity distribution.
[ Table 2 here] These results relate to the broad group of FTSE All Share stocks. In order to investigate whether the above findings apply equally to stocks that are more commonly analysed and traded, we repeat the above analysis separately for the subset of FTSE 250 stocks and FTSE 100 stocks. In the third column of Panel A Table 2, for the historic constituents of the FTSE 250 index, UK EPU is still priced in stock returns across all portfolios -significant at the 1% level in the case of the '1-10' and '1-5' portfolios. However, moving to FTSE 100 stocks we find that there is a notable diminution in this evidence. Because there are far fewer stocks included in the FTSE 100 index and its constituents have changed over time, we do not have return data for the '1-15' portfolio at some points in time (denoted by NaN 3 ).
In Panel B and C of Table 2, we present results from investigating whether economic policy uncertainty in the EU and US respectively plays a role in UK stock returns. Generally, there is little robust evidence in support of such a role: only domestic economic policy uncertainty is relevant. Table 2 presents results around the pricing of economic policy uncertainty. Our study also examines the pricing of macroeconomic uncertainty which is assessed by two factors in our study: economic activity uncertainty and inflation uncertainty.
These results are given in Table 3. From Panel A, relating to economic activity uncertainty, it is quite clear that for FTSE All Share and FTSE 250 stocks, the alpha of the portfolio comprised of low minus high economic activity uncertainty stocks is significantly positive at at least the 10% significance level (with the only exception being the '1-5' portfolio for FTSE 250 stocks). However, there is no supporting evidence in the case of FTSE 100 stocks. In Panel B, relating to inflation uncertainty, we see very little evidence in support of a role for inflation uncertainty in UK stock returns.
Overall, Table 3 provides supporting evidence that stocks that are sensitive to fluctuations in UK economic activity command a future return premium. This is particularly the case for FTSE All Share and FTSE 250 stocks but not for the cross section of FTSE 100 stocks. However, the results fail to document a role for stocks' sensitivity to inflation uncertainty in future stock returns. Therefore, although economic activity uncertainty and inflation uncertainty jointly contribute to the overall macroeconomic uncertainty, economic activity uncertainty is the real factor relating macroeconomic uncertainty variables to stock returns.
[ Table 4 here] In Panel A Table 4, we are surprised to see that for FTSE All Share stocks, all the low minus high economic policy uncertainty sensitivity alphas become statistically insignificant across all fractile portfolios. This indicates that the risk premium for economic policy uncertainty risk reported in Table 3 is explained by illiquidity risk factors. However, moving to FTSE 250 stocks that are more commonly traded and exhibit less illiquidity risk, we obtain strong evidence indicating that controlling for both systematic and characteristic illiquidity risk factors together with the market, size, value and momentum factors, economic policy uncertainty in the UK is still priced in stock returns. The alpha of the '1-5' and '1-10' portfolios are significant at the 1% significance level. In Panel B Table 4, in the case of economic activity uncertainty pricing, there remains some, albeit weaker, evidence of a role for economic activity uncertainty in pricing among the broad universe of stocks. This is strongest in the case of FTSE 250 stocks (significant at the 5% significance level in the case of the '1-5' portfolio and significant at the 10% significance level in almost all other cases). Therefore, the evidence indicates that our economic activity uncertainty risk factor does play some role in UK stock returns even controlling for all the existing risk factors in .
Across all tabulated results, the emerging theme is that one-period ahead UK stock returns may be partly predicted by stocks' sensitivity to UK economic policy uncertainty and UK economic activity uncertainty over the previous three years. This findings is particularly robust among FTSE 250stocks where it persists even after controlling for illiquidity risk factors in addition to more conventional risk factors for market, size, value and momentum risks.

Conclusions
We examine the role of economic uncertainty in explaining the cross-sectional Electronic copy available at: https://ssrn.com/abstract=3355648 variation of UK stock returns. We distinguish between economic activity and inflation uncertainty by employing five separate economic uncertainty indices. We also consider economic policy uncertainty. Our study is distinguished in particular by using financial variables that are theoretically justified in forecasting the economy to construct economic activity and inflation uncertainty indices for the UK. This is quite important because using all available information is not always optimal in predicting developments in the macro economy. After controlling for market, size, book-tomarket and momentum and illiquidity risk factors, we find evidence in support of using our estimated activity uncertainty index and the Baker et al. (2016) UK economic policy uncertainty index to predict the cross-sectional variation of UK stock returns.
Our results suggest that stocks' sensitivity to both UK inflation uncertainty and foreign economic policy uncertainty is not rewarded by higher returns. UK stock market investors should concentrate on stocks that are negatively sensitive to economic activity uncertainty and/or UK economic policy uncertainty in order to earn an abnormal return.
Electronic copy available at: https://ssrn.com/abstract=3355648   The Economic Activity Uncertainty Index The Price Uncertainty Index   Baker et al. (2016) EPU index along with market, size, value and momentum factors. A stock's economic uncertainty risk is the beta on the EPU index. Stocks are sorted into either 15, 10 or 5 equal weighted portfolios based on beta and held for one month before reforming the portfolios. The time series of the low-uncertainty beta portfolio minus the high-uncertainty beta portfolio is tested against the Carhart (1997) four-factor model. Table 2 reports the alphas of these regressions with p-values in parentheses. * represents significance at 10%, ** represents significance at 5% and *** represents significance at 1%. NaN means return data on the low minus high portfolio is not available at some points in time. For all stocks in either the FTSE All Share index, the FTSE 250 index or the FTSE 100 index, each month economic uncertainty risk for stock i is estimated by regressing stock i's returns over the previous 36 months on the macroeconomic uncertainty index (economic activity uncertainty or inflation uncertainty) along with market, size, value and momentum factors. A stock's economic uncertainty risk is the beta on the macroeconomic uncertainty index. Stocks are sorted into either 15, 10 or 5 equal weighted portfolios based on beta and held for one month before reforming the portfolios. The time series of the low-uncertainty beta portfolio minus the high-uncertainty beta portfolio is tested against the Carhart (1997) four-factor model. Table 3 reports the alphas of these regressions with p-values in parentheses. * represents significance at 10%, ** represents significance at 5% and *** represents significance at 1%.NaN means return data on the low minus high portfolio is not available at some points in time. (0.0408) Note: For all stocks in either the FTSE All Share index or the FTSE 250 index, each month economic uncertainty risk for stock i is estimated by regressing stock i's returns over the previous 36 months on the economic uncertainty index (the UK EPU index or the economic activity uncertainty index) along with market, size, value, momentum and illiquidity factors. A stock's economic uncertainty risk is the beta on the economic uncertainty index. Stocks are sorted into either 15, 10 or 5 equal weighted portfolios based on beta and held for one month before reforming the portfolios. The time series of the low-uncertainty beta portfolio minus the high-uncertainty beta portfolio is tested against the  five-factor model. Table 4 reports the alphas of these regressions with p-values in parentheses. * represents significance at 10%, ** represents significance at 5% and *** represents significance at 1%.