Limitations and Alternative Methods

Discussion

Page 107 of 118

Discussion

Page 108 of 118 Alternatively, the value factor could have been combined with a size or BAB factor to compare more different value strategies. We combine value with only one factor in the portfolios, however, academics such as Fama and French (1993), Carhart (1997), Frazzini, Kabiler and Pedersen (2013) have shown that cross sectional variations in returns can be explained through multifactor models.

Alternatively, it would have been interesting to combine e.g. value, momentum and quality into a single portfolio. Hypothetically, as the individual factor loading would become smaller as more factors are included, we expect that the transaction costs incurred by momentum would be reduced, increasing the excess returns after cost. As value-momentum and value-quality have almost similar after cost Sharpe ratios, a lower momentum loading in a three-factor portfolio with value, momentum and quality could increase the Sharpe ratio beyond those of the examined two-factor portfolios. As we find that the US market index have outperformed the two-factor portfolios since 1985 after costs, a portfolio including the market factor would have been interesting to examine. Asness, Frazzini, Israel, Moskowitz and Pedersen (2015) finds that a strong size premium can be obtained when controlling size for junk stocks, and that this also explains interaction with size and other factors such as value and momentum. Even though we limited our self from including size as a factor, due to the academic discussion about its existence since the 1980’s, the size factor could have been relevant to include in the tested portfolios if controlled for junk stocks.

We limit our study to the US market as the amount of data exceeds the amount of available data from all other markets. Asness, Moskowitz and Pedersen (2013) finds consistent value and momentum premiums across eight different stock markets and across different asset classes. Alternatively, we could have expanded our analysis of the value, value-momentum and value-quality portfolio across more developed markets to determine if the premiums were present globally. Asness, Moskowitz and Pedersen (2013) finds evidence for value and momentum but does not study the performance of the quality factor globally. Expanding the analysis to include different markets could provide insight in the portfolios performance internationally, and uncover if the performance could be due to international risks that we are not able to observe due to focusing only on the US. Asness, Moskowitz and Pedersen (2013) finds that value assets have significant co-movement across asset classes, and momentum the same. Between value and momentum there is negative correlation across and positive correlation within different asset classes. Their findings indicate that common global risk factors exist, which we are not able to measure as this study is limited to the US stock market. Asness, Moskowitz and Pedersen (2013) finds that the global returns of the factors have little correlation with macroeconomic variables, but that liquidity risks are negatively correlation to value and positively to

Discussion

Page 109 of 118 momentum. They explicitly state that global funding liquidity risk is a partial source of these value and momentum patterns, which only can be identified when testing jointly across markets.

Alternatively, we could have examined the performance of a value, momentum and quality strategy on other asset classes than equities, such as bonds, currencies and commodities.

Another alternative method presented is on how the factors are defined. For the value portfolio we used an index based on returns from value stocks identified by their B/M ratio. As shown in Chapter 2, this measure can be derived from Gordon’s growth formula. In equation 2.1 we show how the P/E-ratio can be derived from Gordon’s growth formula as well, indicating the several measures beside B/M can be used as a measure to screen for value stocks. We have limited our study to B/M as this is the most used value identifier in academic research papers. Alternatively, value stocks could have been identified using P/E, dividend/price or cash flow/ price ratio. The momentum factor applied is based on past returns, but as stated in chapter 2.5, momentum could also have been identified from earnings momentum or adjustments in analysts forecast. The metrics used by Asness, Frazzini and Pedersen (2013) to compose the quality factor was profitability, growth, safety and pay-out ratio.

Quality measures such as return on equity, return on invested capital, solidity e.g. could have been included when screening for quality stocks. We only identify how value performed individually and combined through previously constructed factors, but do not examine how the factors are optimally identified in the context of our study. Before combining the portfolios, it would have been preferable to identify which measure would have been optimal to use in constructing the factors.

Limitations in Transaction Costs

In chapter 4.6 we illustrated how transaction cost were calculated for the analysis of the different value strategies post transaction costs. In the chapter, we mentioned that the transaction costs calculated should only be considered an estimate, as the transaction cost per trade were set at current level of 0,05%. This transaction cost was chosen because we wanted to estimate the strategies profitability with a forward-looking perspective. However, an alternative approach to the transaction cost calculation that potentially would have yielded a more realistic estimate would have been to use actual transaction costs per trade as they have been historically. Historically transaction costs have been around 1% up until the mid-60’s and then around 1% until mid-80’s, and sometimes even higher depending on the size of the trade according to Jones (2002). After the market liberalization transaction costs fell drastically and became more aligned with the number used in chapter 4.6.

Discussion

Page 110 of 118 Had we used the more precise historical numbers of transaction costs, the resulting return profile for the value strategies would have been different than the ones we found. As shown, transaction costs have an adverse effect on the return profile of the value strategies. The higher the transaction cost, the higher the adverse effect. Intuitively this means that the strategy that are most affected by a change in transaction costs will be the strategy with the most trades, as transaction costs are incurred for every trade.

With a more realistic transaction costs, all the value strategies return profiles would see a downwards draw compared to the current results found with transaction cost equal to 0,05%. Value and value-quality would be almost equally affected, but value-momentum might be so affected by the large historical transaction costs, that its historical performance would possibly show a less profitable investment strategy. The value-momentum was so affected by transaction costs, that it is not unreasonable to assume that it might had yielded a negative excess return historically. Looking forward, we expect value-momentum to be profitable, as costs and barriers have diminished.