3.2. R&D volatilityThe treatment of

3.2. R&D volatilityThe treatment of accounting R&D in the UK follows the SSAP 13. This guideline allows firms to write off their R&D expenses in the year they occur or to capitalize them in certain circumstances, such as when the reliable measurement of future benefits is possible. However, Nixon (1997) provides evidence that accountants in the UK prefer to expense R&D costs due to the uncertainty in estimating future benefits, with a large majority of the accountants (81%) writing-off these expenses immediately. Following Kumar and Li (2016), we estimate the R&D volatility measure with the available information on a firm's annual R&D expense in the income statement.Mudambi and Swift (2014) measure R&D volatility as unexpected and extreme changes in R&D in a GARCH specification over a ten-year window. Although this approach is feasible in OLS-based tests where long-term R&D changes can be included in the regression as a variable, it bears disadvantages in a panel setting. Indeed, ten-year rolling windows for measuring R&D volatility might lead to lower statistical variance over a period of observation. More importantly, the effects of R&D decisions are less likely to persist over a long period of time.Therefore, we rely on a simpler measure of R&D volatility, which Mudambi and Swift (2011) also propose. Further, we use the coefficient of variation, that is, the standard deviation of residuals in the trend of R&D expenditures over the past five years:R&D Volatility=Si/ xiwhere S i is the standard deviation of the residuals from regression R & D it = α 0i + β 1i t + ε i from t to t − 4, and x i is firm i's mean R&D expenditure over the five-year rolling windows.