Explanation
In the case of mean reversion, the actual VaR would be lower than that estimated using the square root of time rule. This is because gains over a period would be followed by losses so that the price can revert to the mean.
In such cases, the autocorrelation between subsequent periods is effectively negative. This means the combined VaR over the periods would be lower.
In the case of positive autocorrelation, the actual VaR would be higher than that estimated using the square root of time rule for exactly the opposite reason than that described for the mean-reverting case.
(Recall that Variance (A + B) = Variance(A) + Variance(B) + 2*Correlation*StdDev(A)*StdDev(B). In cases where correlation is zero, the variance can simply be added together (which is the case for iid observations). In cases where the correlation is negative, the combined variance (and therefore standard deviation and also VaR) will be lower; and where correlation is positive, the combined variance (and therefore standard deviation and also VaR) will be higher.) Therefore statement II is correct, and so is statement III. Choice 'c' is the correct answer.

Explanation
Once we have a number of distributions in the model space, the task is to select the "best" distribution that is likely to be a good estimate of true severity. We have a number of distributions to pick from, an empirical dataset (from internal or external losses), and we can estimate the parameters for the different distributions.
We then have to decide which distribution to pick, and that generally requires considering both approximation and fitting errors.
There are three methods that are generally used for selecting a model:
1. The cross-validation method: This method divides the available data into two parts - the training set, and the validation set (the validation set is also called the 'testing set'). Parameter estimation for each distribution is done using the training set, and differences are then calculated based on the validation set. Though the temptation may be to use the entire data set to estimate the parameters, that is likely to result in what may appear to be an excellent fit to the data on which it is based, but without any validation. So we estimate the parameters based on one part of the data (the training set), and check the differences we get from the remaining data (the validation set).
2. Complexity penalty method: This is similar to the cross-validation method, but with an additional consideration of the complexity of the model. This is because more complex models are likely to produce a more exact fit than simpler models, this may be a spurious thing - and therefore a 'penalty' is added to the more complex models as to favor simplicity over complexity. The 'complexity' of a model may be measured by the number of parameters it has, for example, a log-normal distribution has only two parameters while a body-tail distribution combining two different distributions may have many more.
3. The bootstrap method: The bootstrap method estimates fitting error by drawing samples from the empirical loss dataset, or the fit already obtained, and then estimating parameters for each draw which are compared using some statistical technique. If the samples are drawn from the loss dataset, the technique is called a non-parametric bootstrap, and if the sample is drawn from an estimated model distribution, it is called a parametric bootstrap.
4. Using goodness of fit statistics: The candidate fits can be compared using MLE based on the KS distance, for example, and the best one selected. Maximum likelihood estimation is a technique that attempts to maximize the likelihood of the estimate to be as close to the true value of the parameter. It is a general purpose statistical technique that can be used for parameter estimation technique, as well as for deciding which distribution to use from the model space.
All the choices listed are the correct answer.

Explanation
The probability of default would make the expected value of the future cash flows from both the corporate bond and the risk free bond identical. If p be the probability of default, the cash flows from the risky corporate bond would be
= (cash flows in the event of default x probability of default) + (cash flows without default x (1 - probability of default))
=> 5%*0 + (1 - 5%)*(1 + 12%) = (1 + Rf).
therefore Rf = 6.4%
(In reality investors would demand a 'credit risk premium' over and above the expected default loss rate. They are unlikely to be happy with just being compensated with exactly the expected default loss rate plus the risk-fre rate because the expected default loss rate itself is uncertain. They would demand some premium over and above what the default rate alone might mathematically imply above the risk free rate. In this question, this credit risk premium is ignored.)

Explanation
The correct answer is choice 'd'
For settlement of interbank payments, there are broadly two kinds of systems: RTGS (Real Time Gross Systems) and Deferred Net Systems (DNS). RTGS process payments in real time, settlement by settlement, and each transaction is settled by the a clearing institution (mostly the central bank) on a gross basis without regard for other settlements affecting the counterparty. DNS systems, on the other hand, allow for debiting or crediting the accounts of counterparties at periodic intervals after netting all payments paid or received since the last settlement. The exact timing of the payments does not matter so long as a bank has sufficient funding on a net basis at settlement time. Implicit in the DNS system is the extension of credit and liquidity by the central bank to the participating banks as it is possible for a bank to issue payment instructions even without having funds so long as they can arrange for such funds prior to settlement at the end of the day. In RTGS, a bank needs to have funds to make a payment at any point, and cannot make a payment against moneys expected to be received later intra-day. RTGS systems therefore need more liquidity on the part of the participants, and consume far more liquidity than DNS arrangements. Of course, the 'liquidity' of the DNS arrangement has a cost - which is that someone is taking up settlement risk, and invariably it is the central bank. If a bank under DNS fails to settle, its transactions have to be 'unwound', ie all payments made by it have to be rolled back. This can cause other banks to trip, causing further unwinding transactions. RTGS systems do not carry this risk. Therefore statement I is not correct as RTGS consume more liquidity than DNS arrangements.
Statement II is correct. US Fedwire or European TARGET are RTGS while CHIPS is a DNS based payment system.
Statement III is not correct. Current Basel requirements do not require any disclosure in respect of liquidity risk management. A consultative paper was issued by BIS in Dec 2009 for comments from members, but it is far from final. The BIS is still reacting to the liquidity issues that arose during the 2007-09 credit crisis.
Statement IV is not correct as a CFP is like a disaster recovery plan for liquidity, ie it helps a bank plan for and think about what steps would be taken to deal with a liquidity disaster situation. It does not provide any access to central bank financing.

Explanation
The correct answer is Choice 'c'. The other choices do not refer to 'regulatory arbitrage'.