This work addresses two important issues of investing through asset managers: similarities in the investment philosophies of low cost funds and share restrictions of hedge funds. For the low cost funds, I create a framework to reveal their investment philosophies and study the resulting predictability of the aggregate fund trading actions. First, I develop a new methodology, which, using discrete trading observations, quantifies the fund's preferences towards available factors and their values. The approach enables us to classify quantitative factors into "Action" or "Attention" types. The latter is used to identify whether the fund is using a given factor to make trading actions or merely as a filter to concentrate its attention on a subset of stocks. I apply the model on the US mutual fund holdings data and find that 82.7% of the US mutual funds have a significant preference towards certain factor regions. For the funds which existed in the period from 1994(q1)-2014(q4), 6-month momentum is the most popular "Action" factor (used by 28% of the funds). Whereas, the most popular "Attention" factors are turnover and size (used by 43 and 36% of the funds respectively). I find that a fund's preference towards a factor value might change depending on the quantities of other factors. In particular, within different factor deciles (clusters), fund's preferences towards the same factor might be completely opposite to each other. After, I create a theoretical model, where agents follow pre-defined investment philosophies and make trading decisions based on the changes in the underlying "Attention" and "Action" factors. The model is developed to work in a framework where funds have a finite-dimensional source of public information. I adjust the model for possible trading style changes and use it to predict the next quarter trades for each fund. I aggregate those predictions to test the model on the US mutual fund holdings data. I find that for stocks with large institutional holdings, the change of the holdings between adjacent quarters can be predicted with the model. The results provide evidence against the commonly accepted hypothesis which states that the mutual fund herding happens because the funds follow each other's trades. The Action|Attention model provides an alternative explanation, where an "unintentional herding" happens because of the similarities of the trading philosophies of the low-cost funds. For Hedge Funds, I create a framework of optimal portfolio construction that can incorporate the costs of re-balancing constraints and share restrictions. I do that by transforming a constrained portfolio construction problem into an unconstrained one by penalizing the expected returns of the underlying assets. The methodology is applied to computing the lockup premium of hedge funds in Markov-Switching and transaction cost frameworks. In contrast to the approaches I find in academia, I argue that the hedge fund lockup illiquidity should be modeled as a lost investment opportunity premium. I compute the premium for an experimental data set.