I. Calculate the returns and estimate the volatilities; compare with the end-of-day volatilities

Firstly, to implement this comparison the log returns as well as the volatility at both minute and daily levels are calculated to see how the observation frequency affects risk measurements. Thereby, the measures for both frequencies are annualized to allow a direct comparison. The results show a striking pattern, as minute- level volatility is 20 times higher than daily for all stocks. For example, XVIVO jumps from 0.04 to 0.78, which reveals the extent of hidden risk in HFT-data. This phenomenon can be explained with the inherent characteristics of high-frequency data, which captures minute fluctuations in price and liquidity gaps. These characteristics contribute to an increased degree of variability in the observations, causing them to be more noisy. This would not be captured through the use of daily observations, as such noise would be smoothed out over an extended observation time frame. For traders, this implies that intraday risk is considerably more significant than end-of-day risk. Nevertheless, such data is essential for high-frequency traders, market makers and institutional investors, as they rely on sub daily volatilities for dynamic hedging and estimating position adjustments. Therefore, it is important to understand the effects of volatility at different time resolutions in order to avoid costly mistakes.

II. Merge the data of returns so that the observations are contemporaneous.

To ensure the validity of statistical comparisons, it is necessary that all stocks’ minute returns are aligned to occur at identical timestamps, as contemporaneous data is imperative for accurate statistical inference. Aligning timestamps ensures that all stocks are compared at the same moments. Without such alignment, artificial correlations could be introduced, yielding misleading conclusions and potentially inflating or suppressing the results.

III. Estimate the β’s using XACT as a proxy for the market. How does the β’s compare to the estimate from end-of-day data?

Using XACT OMXS30 as the market benchmark in this study, all stocks are found to be ”defensive”, as can be seen in Table 2.VI all betas are below one. In the context of financial markets, a beta of one indicates that stock movements are parallel with the benchmark. Conversely, the observed low beta values at minute-level reflect the dominance of idiosyncratic over systematic risk, indicating that Market-wide information diffuses gradually. Additionally, as beta estimates collapse the R2 becomes small, implying that systematic risk is relatively low. Therefore, high-frequency returns contain more noise and less correlation with market benchmark movements, diminishing the share of market wide movements. These independent swings imply that market hedges matter less, as idiosyncratic risk can be seen to dominate for all stocks.

The boxplots in Figure 2.VI visually confirm that volatility is significantly higher with minute-level sampling, which reinforces the earlier statistical comparison. Together, with beta figures in Table 2.2 this illustrates how volatility spikes are much more frequent at minute-level frequencies and reduce the predictive power of the market benchmark. This finding affirms that intraday turbulence is something that longer horizons fail to capture, highlighting the importance of precise timestamp alignment. Therefore, traders operating on short timeframes should not rely on long-term beta or volatility estimates. For high-frequency strategies, idiosyncratic risk dominates, and models must incorporate other market-specific signals. This is directly relevant for asset managers, risk teams, and algorithmic traders who rely on accurate short-term volatility forecasts.