In the paper 'The Information Content of ICO White Papers' by David Florysiak and Alexander Schandlbauer (2019), the authors evaluate if the information disclosed in white paper successfully signal the ICO issuers quality-type. To do so they extract the textual information content decomposed into standard or informative content components - standard being related to terms used in concurrent and same industry white papers and informative terms being not explained by recent and industry white papers. Also, the study evaluates predictions consistent with signaling in different economic setting ie. pre-ICO, ICO and post-ICO phases and future distinguish between the effects in hot and cold markets. They hypothesize that high-quality ICO issuers have an incentive to separate themself form low-quality ICO issuers and therefor will signal their quality-type with a higher informative content white paper. Based on that they hypothesize that high-informative content ICOs is associated with lower information asymmetry and hence less rating disagreement across rating experts in the pre-ICO phase. In the ICO Phase the authors predict that greater informative content ICOs leads to increased probability of success in funding (hence high-quality type ICOs are more likely to get funded). Future in the post-ICO phase it is hypothesized that there is a positive relationship between informative content and exchange-listing probability. More so it is hypothesized that high-quality type issuers that signal their type through greater informative content also underprice their ICO-issue leading to a positive relation between informative content and underpricing. Do to high asymmetric information in ICO markets, values of low-quality ICOs are likely overpriced while high-quality ICOs are likely underpriced. Due to a expected positive relation between informative content and trading volume the discrepancy between price and quality-type is expected to correct over time, leading to higher expected returns for high-quality ICOs. The paper finds that an increase in informative content leads to a decrease in rating deviations. While insignificant for the entire sample, higher informative content during hot market phases increases the likelihood of the ICO being successfully funded. While in the post-ICO phase the authors do not find a significant relationship between informative content and listing probability. For the overall sample the authors find a very strong relationship between informative content and underpricing. Higher standard content is associated with higher cumulative one week return, becomes insignificant after one month and reverses after three months. No positive relationship between cumulative returns and informative content across the one week, one month and three months test periods, which could be a sign of strong and persistent mispricing in the short run. Differentiating between hot and cold market phases the authors find associations with informative content are mostly only present and reinforced in cold market phases but less or not present in hot markets.
Set up data
Note* My log return for the series had funny looking numbers on the first day for all portfolios, therefor I moved the period backwards by one day resolving the problem. Then I lather on remove the extra date, to get the correct return series for the period.
2.a.i - Portfolio All ICOs
2.a.ii - Portfolio Top informative content ICOs
2.a.iii Portfolio Top standard content ICOs
In the above descriptive statistic table the mean daily log return, standard deviation (volatility), range (min, max) and quantiles is reported for the 3 portfolios. The standard content portfolio has the highest daily mean log return (0.0095), next all ICOs portfolio (0.0067) and lastly informative content ICOs portfolio (0.0040). The standard content portfolio also has the highest volatility over the daily log returns (0.1043), next the informative content ICOs portfolio (0.0851) and lastly the all ICOs portfolio with the lowest volatility over the daily log returns (0.0676). Looking at the range (max - min) the standard content portfolio has the highest range, then the informative content ICOs portfolio and lastly the all ICOs portfolio with the smallest range, in agreement with the volatility measures. the quartiles indicate that 25%, 50% and 75% of the daily log returns for each portfolio lies below these values hence 25% of daily returns for the All ICO portfolio < -0.0238, 50% < 0.0061, 75% < 0.046 and so on for all portfolios. interestingly - looking at the median (50%) - the informative ICOs portfolio with 0.000793 which indicate that for the informative content ICOs portfolio 50% of daily return lies below 0.000793 which is small compared to 0.00606 and 0.00153 for the all ICOs and standard content portfolio respectively.
Based on the above descriptive statistic results I expect standard ICOs portfolio return to be more volatile hence riskier compared to the other 2 portfolios, after that the informative content ICOs portfolio and lastly all ICOs portfolio being the least risky. Based on mean daily return I expect the standard content ICOs portfolio to produce the highest overall return, then all ICOs and informative content ICOs portfolio respectively
Had I not looked at the above results I would have expected that high-quality ICOs, who signal their quality-type through informative content white papers, being less volatile, and low-quality ICO with high standard content with papers to be more volatile making low-quality ICOs more riskier. Therefor the portfolio with top 100 content ICOs I would expect to show the lowest volatility, after that the portfolio based on all ICOs and lastly the top 100 standard portfolios to show the highest volatility. Reason being that low-quality ICOs is bought through hype rather than the underlying business and hence being more volatile and riskier to invest in. The results could indicate in line with (Florysiak and Schandlbauer 2019) that the marked is hot and high content ICOs is undervalued.
The above plot shows the exponentiated cumulative log return for the three portfolios with the standard content ICOs portfolio in green, the all ICOs portfolio in blue, and the informative content ICOs portfolio in orange. It looks like the informative content ICOs portfolio being the least volatile of the three portfolios which is do the y-indexing putting the three cumulative return series in the same graph hence the all ICOs portfolio is the least volatile of the three series in agreement with results and expectations from 2.a.
The above table shows the exponentiated cumulative sum of the log returns being the simple return for the three portfolios over the whole period. It shows in line with the earlier findings and expectations (2.a) that the informative content ICOs portfolio generate the lowest nominal return followed by all ICOs portfolio and lastly the standard content ICOs with the highest nominal return. It is important to keep in mind that this is simple returns and not real returns.
Portfolio All ICOs
Portfolio informative content ICOs
Portfolio standard content ICOs
In the above table annualized volatility, sharpe ratio and log returns are reported. The standard content ICOs portfolio with a 239% return has the highest return of the three portfolios, but it also has the highest volatility hence is riskier compared to the other portfolios. The all ICOs portfolio has the lowest volatility but with a return of 169% has an almost 70% higher return compared to the informative content portfolio which has a return of 100%. The volatility and returns are in agreement with the above findings and expectations (see 2.a). Looking at the sharp ratio which is a risk adjusted measure for the portfolio returns - the higher sharpe ratio the better. Based on that all ICOs portfolio is the best pick of the three.
Correlation values can range between -1 and 1 for perfect negative and perfect positive correlation. A perfect positive correlation means as one portfolio return moves either up or down the other portfolio return moves equally in the same direction. The correlation matrix shows that the three portfolios are all positively correlated, indicating all three portfolios move in the same direction together. The strongest correlation is between all ICOs portfolio and the other two portfolios (informative content ICOs portfolio (0.74) and standard content ICOs portfolio (0.65)) Which makes sense since the two portfolios bough are subsets of the all ICOs portfolio. Interestingly the informative content ICOs portfolio is more correlated to the marked than being the informative content ICOs portfolio which could indicate that the performance of the standard content ICOs portfolio is not driven by the market as a whole but rather subject to speculation indicating a hot market. The correlation between informative content ICOs portfolio and standard content ICOs portfolio is 0.56 which again makes sense since they are the most different sets compared. Based on the correlation matrix it is not possible to conclude whether the relationships presented are statistically significant.
See the description below tables and graphs in the above document.
As my expectations was based on the results from 2.a, the findings are in agreement with my expectations. Had I made my expectations before running 2.a I would have expected to se the informative ICOs portfolio being the less risky choice of the three, then all ICOs portfolio and lastly standard content ICOs portfolio. Also I would have expected informative content ICOs portfolio to generate the highest return, then the all ICOs portfolio and standard content ICOs portfolio respectively. Leading to informative content ICOs having the highest sharpe ratio, followed by all ICOs portfolio and lastly standard content ICOs portfolio with the lowest sharpe ratio. The above results almost find that, but with one important disagreement being the informative content ICOs portfolio which has the lowest return, medium volatility and smallest sharpe ratio compared with the two other portfolios.
With daily rebalancing, transactions costs and potentially taxes are anreal issue which would eat of daily positive return and make negative return more costly leading to decrease in performance for all portfolios. One way to reduce transaction cost would be to hold the ICOs over longer periods and rebalance each month, quarterly or twice a year. Based on the correlation results the informative content ICOs portfolio is more affected by market fluctuation one could short-sell informative content ICOs portfolio when the market shows signs of recession or simply liquidate positions and buy again when markets show signs of upward movement. Which leads to liquidity constraints – not all ICOs can easily be liquidated and not all investors can borrow funds at feasible interest rate potential making overall gain smaller than cost and hence, making the marked less efficient than simple market theory could lead one to believe.
In the following I descripe what formalas I use to calculate log return, volatility and sharp ratio.
To calculate the annualized daily return I simply take the avarage daily log return and multiply it with 252 trading days
To calculate the annualized volatilities I take the standard deviation of the daily log returns and multiply with the square root of 252
To calculate the annualized sharp ratio I take the avarage of daily annualized sharp ratios