Introduction:
This paper, “Liquidity in Fixed Income Markets: A Comparison of the Bid-Ask Spread in Corporate, Government and Municipal Bond Markets”, is written by Sugato Chakravarty and Asani Sarkar. The U.S. bond market being the largest market in the world (valued at more than $10 trillion), this paper focuses principally on the factors determining the realized bid-ask spread in the U.S. Corporate, Government and Municipal bond markets for the years 1995 to 1997. The bid-ask spread is the difference between the average buy price and average sell price per bond for one day (Chakravarty & Sarkar, 1999).

The three bond markets mentioned above are highly important markets in terms of the share they have in the dollar value of the U.S. debt markets which is an impressive two-thirds. A lot of different factors have varying effects on these markets. Credit risk on the bonds, trading activity, transparency of the market and issuer-specific characteristics are different for corporate, government and municipal bonds. The way in which these factors affect the individual markets is presented first and foremost in the paper and this is one of the things which makes the literature easier to understand later on in the paper and shows the superiority of this particular research paper over other similar attempts (Chakravarty & Sarkar, 1999).

It is after this that a comprehensive study of the bid-ask spread is conducted separately for each of the three markets. Here, a number of important determinants are unveiled, the most vital one being liquidity, which is shared by all three markets. A more detailed analysis will be done in the following sections, but it was found that some of the other determinants for either of the markets are trading volume, risk factors, remaining-time-to-maturity, after-tax bond yield and the age of a bond. After this, all the results are accumulated and a common model is proposed. When a pair-wise comparison of the markets was done, it was seen that corporate bond spread does not differ from government bond spread; municipal bond spread is higher than corporate bond spread by 8 cents and municipal bond spread is higher than government bond spread by 9 cents (Chakravarty & Sarkar, 1999).

Model and Method of Analysis:

This cross-market comparison of the bid-ask spreads was performed with the differing impact of the control factors in mind. Credit risk is virtually non-existent for government bonds and highest for corporate bonds, with municipal bonds ranking somewhere in the middle. U.S. governments bonds and Treasury securities also enjoy the highest rate of trading activity. This market is also the most transparent, an issue of concern for the corporate bond market and the conditions still said to be improving for the municipal bond market. Hence, these different characteristics of the three markets count for the differences in bid-ask spread (Chakravarty & Sarkar, 1999).

Data Collection:
To arrive at the substantial conclusions mentioned above and in the earlier section, an exhaustive collection of data was first undertaken. Data was collected from a variety of sources. The researchers obtained a records of bonds purchased and sold during the time period of 1995 to 1997 from Capital Access International. These records had actually been filed by insurance companies who were required to file their securities transactions. This is the fundamental data set which has been made use of in this study. The sample which the researchers got consisted of 453,481 individual transactions in the Corporate, government and Municipal bond markets, each one complete with identification of the bond, the total dollar value of the transaction, number of bonds traded and the nature of the transaction: whether it was a buy or sell order.

This may have been the main information used, but the researchers also purchased additional information about these bonds such as the credit rating of each bond, credit sector of issuer, issue date and maturity date.
The occurrence of errors in the final result was kept to a minimum by deleting possible erroneous observations, such as transactions where the date is an estimate or where ratings information is not available. Hence, after all this filtering, the final data set comprised of 152,452 transactions in corporate bonds, 54,518 in government bonds and 83,395 in municipal bonds (Chakravarty & Sarkar, 1999).

Emperical Determinants:
A study conducted by Merton (1973) is quoted where the value of corporate debt depended largely on the risk-free rate, provisions built into the bond such as the coupon rate and call provisions, and the default risk. For equity markets, Chakravarty and Sarkar (1999) also expect the bid-ask spread to be related to the price of the bond, and this way, to the determinants listed above.
They decided to control the default risk by introducing dummy variables depending on the credit ratings and the yield spread. The coupon rate and the risk-free rate were not tampered with. Since the bid-ask spread of a bond is related to the risk of trading it, the remaining life of a bond was used as a measure of its price sensitivity. In turn, the trading risk of a bond is also related to its expected liquidity: the greater the liquidity, the easier it is to buy and sell the bond urgently; and here, trading volume was used as a proxy for liquidity.
Different dummy variables were introduced in the study owing to the different categories prevalent in the market sector such as finance companies for corporate bonds and health care bonds in the municipal bonds market. A dummy variable, 1997, was also formulated to control the changes in market structure. If transparency of the market undergoes change, it affects the bid-ask spread in either direction and this determinant is counted for by giving 1997 the value ‘1’ if the transaction occurred in 1997 and zero otherwise (Chakravarty & Sarkar, 1999).

Comparison with other Literature in this area:

In this paper, the liquidity of U.S. Corporate, Government and Municipal bonds was studied and compared for the years 1995 to 1997. It was found that for bonds of $100 par value, the average bid-ask spread is highest for municipal bonds at 22 cents, slightly different for corporate bonds at 21 cents and lowest for government bonds at 11 cents. Generally, bonds with lower credit ratings had a higher bid-ask spread. For the corporate and municipal markets, this average spread was a lower value in 1996 than in 1995, and then in 1997 than in 1996 (Chakravarty & Sarkar, 1999).

The results of this study also showed that liquidity was shared by all three markets as an important determinant of bid-ask spread. As trading volume increases, the bid-ask spread decreases. Since government bonds are virtually risk-free, the risk aspect is dominant in the municipal and corporate bond markets. In both these markets, the bid-ask spread increases in the remaining-time-to-maturity of the bond. For corporate bonds, spread increases with credit risk and age, while for municipal bonds, spread increases with after-tax bond yield. It was also found that in a study of these three years, the municipal bond spread came out to be higher than the government bond spread, but the same can not be said for corporate bond spread (Chakravarty & Sarkar, 1999).

A similar study conducted by Shen and Starr (2000) proposed a model of bid-ask spread in financial markets based on it being a function of asset price variability and order flow. They concluded that the bid-ask spread usually fluctuates in a way so as to compensate the market-makers’ average cost. As price volatility, volatility of order flow and the value of the market-maker’s inventory increase, so does his average cost – and so does the bid-ask spread. It is important here to note that as the three factors (price volatility, volatility of order and value of market-maker’s inventory) increase, the overall liquidity of the market decreases. Hence, Shen and Starr (2000) also propose an inverse relationship between liquidity and bid-ask spread, as was proposed by Chakravarty and Sarkar (1999).
A lot of the other literature in this area, such as Copeland and Galai (1983) and Easley and O’Hara (1983) attribute price risk and asymmetric information as important determinants of bid-ask spreads. Inventory was found to be a determining factor in some studies while in some studies like Stoll (1978), it was found to be a non-factor. In Chakravarty and Sarkar (1999) though, the discussion very rarely discusses this aspect.

Critical Evaluation:

Methodology of the Paper:
The procedural aspect of this research will now be illuminated. First the realized bid-ask spread for every bond per day were calculated as the difference between the average selling price per bond and the average buying price for it. The results were estimates at best, because trades are done at different times during the day. Comparisons were made and it was found that the mean spread was highest for municipal bonds, followed by corporate bonds and lowest for government bonds (owing to their high liquidity).
To further ensure that accuracy was maximum, the corresponding volume-weighted dollar spreads were also compared and the results were in alignment with the raw spread results given above.
When a comparison of market sectors was taken, it was found that utility sector spreads are the highest in their sample. Industry /service sector bonds account for being 45% of all the bonds traded in this sample, and have relatively low bid-ask spreads as do banking / financial sector bonds (Chakravarty & Sarkar, 1999).

Regression Model based on GMM:
Owing to the different factors governing each bond market in different ways, a separate study was done of the set of factors which determined the bid-ask spread in each market. A technique called Generalized Method of Moments (GMM) was used to estimate the price change regression. The regression specification used equated the daily bid ask spread with the sum of the time-to-maturity of the bond in years, the time in years between the transaction date and date of issuance, the sum of the daily dollar value of purchases for the bond, a dummy variable (1997) which was given the value ‘1’ if the bond was traded in 1997 and 0 otherwise and additional dummy variables to control for credit risk according to Moody’s credit ratings (Chakravarty & Sarkar, 1999).

Results of the Regression Model:
For the corporate bond sector, the adjusted value for R-square came out to be 2.28%. Results showed that since the coefficients for Maturity and Age were positive, the spread increases as these variables increase. For the government bond sector, the adjusted R-square was mostly zero, meaning that the specification given could not be called the contributing elements to the bid-ask spread for this sector. In the municipal bond sector, the adjusted R-square was 1.87%. Spread increases with maturity while Age is not a factor here as was in the corporate bond sector. The credit sector dummy variables did not have significant coefficients (Chakravarty & Sarkar, 1999).

Other Regression Specifications and their results:
In a second regression specification, the conjecture was that since sale of corporate and municipal bonds could be information driven, this could have an impact on bid-ask spread. This was however, not proved by the model.

The third regression specification took into account the yield spread in place of the credit variables. Results showed an inverse relationship between bid-ask spread and yield spread (Chakravarty & Sarkar, 1999).

Pooled Regression:
A pooled regression model was used to test the differences in bid-ask spread in all three sectors, after controlling for volatility, credit risk and liquidity. Results showed that municipal bonds spread was higher than government bonds, but corporate and government bond spread was statistically about the same (Chakravarty & Sarkar, 1999).

Review of the Paper:
The question that this paper addresses is definitely an interesting one and one that needed to be looked into in detail. Other literature in this area has usually generalized over the different sectors in the bond market, and in this way, glossed over their differences. Many studies have been conducted relating to the bid-ask spread in financial markets but this one is an in-depth look into the determinants of bid-ask spread specifically in the corporate, government and municipal bond sectors.
This is a very relevant study because as mentioned in the paper, the spreads come out to be different from each other due to the different factors governing them in various capacities. It is important to have knowledge of these facts for a thorough analysis of the overall U.S. bonds market. The results provide a new insight to the very nature of the three market sectors examined in this paper. Liquidity did predictably come out to be the strongest determinant for all three sectors, but some factors important for one sector were not so much for another. For instance, age is a strong determining factor for corporate bonds while the municipal bond spread is largely determined also by yield, because has additional tax subsidies built into it.
I agree with the authors and their research has provided valuable depth to what was previously a general idea of the U.S. bonds market. A sector-wise study such as this one has been instrumental in giving me a thorough understanding of the many facets of each sector and the determinants of its bid-ask spread.