Monetary Liquidity, Market Liquidity, and Financial Intermediation Ujjal Chatterjee University of Wisconsin – Milwaukee, Lubar School of Business, P.O. Box 742, Milwaukee, WI 53201, e-mail: [email protected] . and Yong-Cheol Kim University of Wisconsin – Milwaukee, Lubar School of Business, P.O. Box 742, Milwaukee, WI 53201, Phone # 414-229-4997, e-mail: [email protected]

August 2010

1

Monetary Liquidity, Market Liquidity, and Financial Intermediation ABSTRACT This study investigates the link between the monetary liquidity or the 'macro-liquidity' measured from aggregate balance-sheets variables of the financial intermediaries and the market liquidity or the 'microliquidity' in the U.S.

The study shows that the financial intermediaries' balance-sheets variables contain

information that have predictive power for the equity and the bond market liquidity and that the predictive information content is not uniform across the financial intermediaries and sample periods

It is also shown

that the liquidity constraints of the mark-to-market financial intermediaries such as the securities brokers and dealers than the liquidity constraints of the relationship-based liquidity providers such as the commercial banks have had far greater explanatory power for the market liquidity. The results show embedded 'commonality in asset-liquidity' factor in the aggregate balance-sheets variables of the financial intermediaries.

2

1. Introduction Market liquidity as a measure of asset market liquidity like stocks and bonds is considered to be an additional important factor in explaining stock and bond returns and volatility, and the role of market liquidity is supported both in empirical and theoretical research. However the market liquidity even at this micro level is a much broader concept than other factors related to specific asset or asset classes incorporated in Fama-French asset pricing model. As such, market liquidity at the micro level is related to a broader measure of liquidity at a macro level. Monetary liquidity as a measure of liquidity at macro level covers wide spectrum of liquidity chain from an immediate and direct financial institution like mutual fund to the economy wide factors that are related to the general economy and Federal Reserve monetary policy. Financial intermediaries went thru stages of metamorphosis and evolve and change rapidly to take advantage of changing economic environment to take advantage of regulatory changes, economies of scale and scope, etc. Especially after the repeal of Glass Steagal Act in 1999 and development of financial engineering blurs the distinction of traditional commercial banks and other non-regulated financial intermediaries dissected into specialized intermediation like maturity, credit risk and the supply of liquidity. The aftermath of the „collapse of the financial system‟ renewed interests on the role of the financial intermediaries in the recent financial crisis and the policymakers around the globe are still pouring in billions of dollars to provide „the much needed liquidity‟ to rescue their respective financial systems from the financial abyss.

The meltdown

of the financial intermediaries was accompanied by the crashes of the stocks and corporate bonds markets and „the asset-liquidity‟ evaporated.

How „the much needed liquidity‟ and „the asset-liquidity‟ are related?

The objective of this paper is to investigate the relation by looking at the aggregate balance-sheets variables of the financial intermediaries and the assets‟ market liquidity; specifically, the objective is to predict the market liquidity from the aggregate balance-sheets variables of the financial intermediaries. 3

The main contribution of the paper is analyzed the commonality or the systematic component of assets‟ liquidity in the balance-sheets variables of the financial intermediaries. Financial intermediaries in the study incorporate a wide range of financial intermediaries, and the analysis aggregates financial intermediary based on the operational function of the institutions. In addition the analysis dissects balance sheet information into assets, liabilities and leverage and related each of them to the measure of market liquidity from stocks and bonds. Though the existing literature identifies some of the systematic components of the asset liquidity, except in a very few academic research initiatives, the primary sources of the systematic monetary liquidity and their relation to the asset liquidity remains largely unexplored.

This paper

endeavors to fill the gap by studying the assets‟ liquidity from a view-point that goes right into the core of the liquidity dynamics in that it connects the liquidity of the financial intermediaries, the liquidity providers, to the financial assets' liquidity. Second contribution of the paper is to expand the liquidity reaction from both downstream chain from macro level liquidity provider to market liquidity and upstream from market liquidity to monetary liquidity based on wide range of financial intermediaries.1

To keep discussions straight and to the point, we start by defining the aggregate-balance sheet variables of the financial intermediaries as 'the monetary liquidity' and the aggregate liquidity of the financial assets such as the stocks and the bonds as 'the market liquidity.' While the monetary liquidity is the money flow from the financial intermediaries to the economy, the market liquidity is the transactions-level liquidity of the financial assets. Putting the notion differently, the monetary liquidity is the „macro-liquidity‟ and the market liquidity is the „microstructure-liquidity‟. 1

In addition to proposing different measures of the liquidity, a large body of microstructure research focuses on the relevance and the pricing of the liquidity risks. While the existing literature view the assets' market liquidity from a host of theories such as 'inventory cost', 'asymmetric information', and 'search cost' etc., the fundamental question regarding the sources and the evolution of the liquidity in the economy is yet to be fully addressed.

4

The objectives of earlier research, in the market-micro structural framework, have been, among other things, evolved to explain the asset market liquidity from the macroeconomic perspectives, and found significant role of macro events. Early empirical works focused on stock market only and showed that changes in stock market liquidity proxies by the spread, the depth, and the trading activity for the U.S. equities are affected by the volatility, the market return, and the macroeconomic announcements. (Chrodia, Roll, and Subrahmanyam, 2001). Liquidity is not stuck in one asset market, rather it will move around as long as investors are able to buy and sell large quantities of assets quickly and at a fair market value. It is found that not only there are co-movements in the stock and the Treasury market‟s liquidity but also both asset liquidity is related to the 'macro' liquidity or 'money flows' that is proxied by the bank reserve, the fed fund rates, and the mutual fund investments (Chordia, Sarkar, and Subrahmanyam, 2003). The research has expanded to investigate both cross-sectional comovements but also intertemporoal association between market liquidity and macroeconomic fundamentals.

(Chordia, Sarkar, and Subrahmanyam, 2003, Naes et

al. 2010, and Fujimoto, 2004). For example, the systematic market liquidity is related to the macroeconomic sources such as the fed fund rates, the growth rates of the consumer price index, and the industrial production in the U.S. (Fujimoto, 2004). In addition, the direction of relationship is not only from macro fundamentals to market liquidity, but also it works in reverse as well. Relating stock market liquidity to business cycle, Naes, Skjeltorp, and Ø degaard (2010) reveres the logic and they found that the asset market liquidity has prediction power of macroeconomic indicators such as the GDP of the U.S. and Norway.

In the macro-liquidity literatures, the financial frictions in supplying credit through financial 5

intermediaries as a middle stage of liquidity chain has been a fertile area of research. Specifically, liquidity at mutual fund investments, along with other gross macroeconomic factors, are found to be the determinant of the monetary liquidity (Chordia et al.,2003) The theoretical model of Brunnermeir and Pedersen (2009) provides the link between the asset‟s market liquidity and trader's funding liquidity and they show that the traders provide market liquidity as long as the traders have the funding to provide the market liquidity, that the market liquidity is correlated with volatility and that the market liquidity follows the movement of the market and the funding condition. Macro liquidity can be measured directly by analyzing the balance sheet of financial intermediaries. The aggregate growth-rate of the balance-sheets variables of the financial intermediaries is the „liquidity‟ by noting “......a natural definition of liquidity as the rate of growth of aggregate balance sheets.

In more concrete terms, we can define liquidity as the rate of

growth of repos and other forms of collateralized borrowings are the tools that financial institutions use to adjust their balance sheets.”( Adrian and Shin, 2008). Expanding the findings that the variations in the „repo‟ and „commercial papers‟ funding have predictive power of the innovations in the implied volatility (Adrian and Shin (2007), Adrian, Moench, and Shin (2009) report that the aggregate balance-sheets information of the financial intermediaries such as the leverage of the securities brokers and dealers predict the U.S. equity and bond portfolio returns and that the balance-sheets variables also explains macroeconomic aggregates such as GDP and inflation. The central idea of these strands of research is that the aggregate balance-sheets variables contain information that could shed light on the credit supply conditions and that these variables could be viewed as the 'liquidity'. With this background of recent researches and observations, the link between the monetary liquidity and the market liquidity is explored. The paper lies at the confluence of the market-microstructure and the 'macro-liquidity' literatures. This paper takes these two strands of researches and endeavors to link the two to 6

explain the liquidity dynamics in the economy. The type, operational function and the scope of business of financial intermediaries have expanded and specialized over time in response to regulatory changes and market opportunities until the financial crisis. As the size and importance of shadow banking system that is outside of central bank regulation and protection grows, the function of liquidity transmission of financial intermediaries is important in understanding the role of liquidity transformations by looking at individual institution separately and combination of financial intermediaries in aggregate. The distinction between this study and the prior studies is that the liquidity providers' liquidity as the macroeconomic sources of liquidity is considered. In essence, this paper goes one step down the liquidity-chain by looking at the monetary liquidity available in the market rather than by looking at the macroeconomic sources of the asset market liquidity as is done in the existing literature. Other aspect of this study lies in the comprehensive nature of its treatment of 'money-flow' by including major liquidity providers such as the commercial banks, the pension funds, asset-backed securities issuers, and insurance companies etc. (see Table I). In addition, the role of any specific financial intermediary is not investigated in linking it to the market liquidity, i.e. 'ceteris paribus' conditions are largely relaxed. This approach allows for finding the relative importance of one 'money-flow' channel over others.

The main result is that even after controlling for one of the primary drivers of the market liquidity, the volatility, the balance-sheets variables of the market-based financial intermediaries that operate in a markedto-market basis significantly help understand the market liquidity by analyzing a large set of financial intermediaries‟ aggregate balance-sheets information. The contention is that the financial constraints of the financial intermediaries are dependent on the macroeconomic environment they operate in and their balancesheets variables such as the leverages or the assets-growth are plausible proxies for the monetary liquidity. 7

By incorporating financial intermediary's funding information from their balance-sheets, we get a closer look at the sources of the monetary liquidity in determining the systematic component of the market liquidity. This approach is diametrically opposite to the approaches employed by most of the earlier researches that used gross macroeconomic factors, such as the GDP growth, as the sources of the systematic components of the market liquidity. Thus, this study takes the direct approach in finding the systematic component of the market liquidity. While doing so, the study also reveals the role of the financial intermediaries in determining the market liquidity. The plan of the article is as follows: Section 2 provides the description of the empirical methodology, section 3 describes the data. Section 4 describes the monetary and market liquidity proxies, Section 5 presents empirical results, Section 6 summarizes, and Section 7 concludes.

2. Empirical Methodology We conduct multivariate regression analysis of four liquidity indices: three for the monetary liquidity as explanatory variables and one for the market liquidity as the dependent variable.

For the monetary

liquidity indices, the aggregate balance-sheets variables of the major financial intermediaries are used. As for the market liquidity index, several liquidity proxies for the stocks and the bonds portfolios are included. While the stocks liquidities are measured from the widely used existing liquidity proxies, to proxy for the corporate bonds liquidity a new corporate bonds liquidity measure is proposed (in section 4.2.2). In the following sections alternate monetary and the market liquidity proxies are used partly as a horserace among different proxies, but more importantly to identify the nature of different proxies. The general format of the predictive regression with one of the market liquidity measures as the dependant variable and the monetary liquidity measures as the explanatory variables is: 8

 

 

Y   Y  Z    t t t  j t t t t t t  i

Where,

Y

t

(1)

is a scalar and represents one of the market liquidity measures of the market liquidity index,

is the lagged variable of

Y

t

that describes the market liquidity processes,

of the monetary liquidity indices, and



t

is a vector of control variables;

Z

t



Y

ti

is a vector and represents one

t

is a constant, and



t

is the

error term. While 'j' varies from 0 to 5, „i‟ depends on the underlying market liquidity processes. The 

parameters  , t

interested in



t

t

, and

t

are to be estimated. Though

t

is important and discussed in details, we are

.

3. Data Description The quarterly-samples under investigation are from the March 1986 to December 2009; the sample period selection was entirely based on the availability of data (for example, Moody‟s corporate AAA and BAA rated bond index yields data dates back to 1986). The whole sample is from 1986 to 2009 and the sub-sample is from 1998 to 2009 to accommodate tick size changes in 1998 and the more detailed rationale for choosing the sub-sample is discussed in Section 4.2.1.

In addition, whenever applicable, separate

analysis is done for the 1986-2006 sample to exclude data from the recent financial crisis. The financial intermediaries data is obtained form the federal flow of funds from the Federal Reserve Bank of New York. While the stocks liquidity is calculated from the daily stocks data from the CRSP, the bonds liquidity is calculated from the Moody‟s corporate AAA and BAA rated bond indices yields and the data was taken from the Federal Reserve Bank of St. Louis. We also include the „market implied volatility‟, the Chicago Board of Options Exchange (CBOE) volatility index, VIX and VXO, a variant of VIX, in the analysis. While VXO data is available since 1986, the VIX data dates back to 1993; as a result, VXO is used for the 1986-2009 sample analysis and VIX is used for the 1998-2009 sub-sample analysis. For brevity, we 9

use the term „VIX‟ for both the VIX and the VXO indices throughout the rest of the paper. We also use the Treasury 10-year constant maturity bonds and 3-months Treasury-bill data from the St. Louis Fed.

4. Liquidity Measures The measurement of the monetary liquidity and the market liquidity is an elusive task and the researchers over the years looked at different proxies for both the monetary and the market liquidities.

In

the one hand, in determining the systematic sources of the monetary liquidity, proxies such as the growth rate of the industrial production, the civilian unemployment rate, the growth rates of the consumer price index, the index of sensitive materials prices, and the federal funds rate, had been used (see Fujimoto, 2004). On the other hand, in determining the market liquidity, a host of market liquidity measures based on different hypothesis of the origin of the market liquidity exist. Next two sections describe the monetary and the market liquidity proxies that are used in this study.

4.1. Monetary Liquidity Measures The role of financial intermediaries in providing liquidity is well established but the challenge remains in identifying the financial intermediaries that are directly linked to the assets' market liquidity. We include the aggregate balance-sheets variables of the financial intermediaries to proxy for the monetary liquidity and investigate how the fluctuations in the balance-sheets variables are related to the market liquidity. We create three indices from the balance sheets of the five categories of the traditional channels of the monetary liquidities: the commercial banks, the mutual funds, the securities brokers and

10

dealers, retirement accounts and insurance companies, and shadow-banks.2

The above composition is

similar to the liquidity benchmark used in Adrian et. al (2009). Briefly described, the shadow-banks are the financial intermediaries, which borrow short and lend long mainly through commercial papers and the securities brokers and dealers are the financial institution that buy and sell securities, hold an inventory of securities for resale, or do both, through short-term borrowing and lending through repos and reverse-repos; other channels of monetary liquidities are selfexplanatory. Though the shadow-banks function like commercial banks, they are neither regulated nor chartered. The inclusion of the securities brokers and dealers and the shadow-banks are essential in the study. First, these financial intermediaries are known to operate on a marked-to-market basis. Second, the aggregate assets of these two classes of financial intermediaries are simply too large to ignore. For example, the aggregate assets of shadow-banks are more than the aggregate assets of the mutual funds (see Table 1). While the securities brokers and dealers directly contributes to the stocks and bonds liquidity, the shadowbanks' importance lies in their intimate relation to the market liquidity in that their primary sources of funding and operation (commercial papers and securitization, respectively) depend on the overall market liquidity conditions. However, we need to pay attention to the operational differences of the financial intermediaries. Each and every financial intermediary functions in different operational regimes and as a consequence, the balance-sheets of each financial intermediary are unique.

For example, while the commercial banks are

fragile in their capital structure in that they struggle to match the duration of the assets and the liabilities of their balance-sheets, the capital structure of the securities brokers and dealers are more agile in that they

2 See Pozsar, Adrian, Ashcraft, Boesky, “Shadow Banking”, Federal Reserve Bank of New York, Report #458, 2010.

11

respond to and driven by the market conditions.

Another example, while the financial intermediary such as

the mutual funds has no liabilities, the funding companies have net-zero balance-sheets.

Given such

differences in balance-sheets of the financial intermediaries, the empirical study is conducted in three stages separately: asset growth, liability growth and leverage. In the first stage of the analysis, the relationship between the aggregate assets-growth of the financial intermediaries and the stocks and the corporate bonds liquidity is investigated. The analysis is based on the hypothesis that the asset-growth of the financial intermediaries is one of the determinants of the market liquidity.

The advantage of having the aggregate assets-growths as explanatory variables is that it allows us

to include each and every financial intermediary in the analysis. The value-weighted components of the index are presented in Table 1 Panel B. In the second stage, the liabilities side of the balance-sheets of the financial intermediaries is investigated. While this analysis excludes some of the financial intermediaries such as the mutual funds, the relationship between the liabilities side of the balance-sheets and the market liquidity may provide an alternative perspective on the liquidity dynamics.

The value-weighted components of the second index are

presented in Table 1 Panel C. In the final step, the relationship between the leverage, defined as the differences between the assets and the liabilities, of the financial intermediaries and the market liquidity is investigated. Though the analysis would exclude a number of financial intermediaries such as the funding companies or the asset-backed securities issuers, this part of the analysis is crucial in that it provides us with the critical information about the balance-sheets management of the financial intermediaries in response to the market liquidity conditions. The value-weighted components of the third index are presented in Table 1 Panel D.

12

[Table 1 of the monetary liquidity index here]

4.2. Market Liquidity Measures To address the liquidity dynamics in a realistic sense, we look at the market liquidity holistically.

It

is no denying that there are multitudes of channels of „money-flow' from the financial intermediaries to the economy and the securities market. liquidity.

In addition, there is not a single measure that proxy for the market

For the tractability and the ease of measurement, we consider the liquidity of both the equity and

the corporate bonds to proxy for the market liquidity by creating a market liquidity index that includes the stocks and the corporate bonds liquidities. The driving criteria for this selection are the available data and the existing market-microstructure literature. The components of the market liquidity index are shown in Table 2 and are discussed in the next two sections: section 5.2.1 describes the stocks liquidity measures and the section 5.2.2 describes the corporate bonds liquidity measures.

4.2.1 Stocks Liquidity Measures While choosing measures of equity market liquidity proxies, we are impartial to any of the measures that are available and employ several commonly used liquidity measures to ascertain that the link between the monetary liquidity and the market liquidity is independent of the market liquidity measure.

The

measures included are the Roll's (1984) measure of effective bid-ask spread (hereafter ROLL), the relative spread (hereafter RS), and the Amihud’s (2002) illiquidity ratio (hereafter ILL). Note that all these liquidity proxies mentioned above measure illiquidity. The liquidity measure of Roll (1984) is a canonical model of the dealer market with fixed cost and it is an estimate of the implicit spread. The variations of this model are present, in disguise, throughout the 13

market-microstructure literature. Under the assumption that there exists a constant effective spread, the liquidity of a security 'i' is captured in D i , t 1 Roll   Cov (  r ,  r  i , t i , d , t i , d , t  1 ) d  1 D i , t

(2)

The market liquidity can then be represented as 1N Roll  Roll  t i,t Ni1

Where,

 rt

(3)

is the difference in return between two successive periods, D is the number of days, N is the

number of securities. The second measure, the relative spread is based on the trading cost and is calculated as the ratio of the bid-ask spread to the midpoint price of a security and is calculated as for a security 'i': D i , t ( price  price ) 1 i , d , t RS   i , t bid highest ask lowest d  1 D ( 0 . 5 price  0 . 5 price ) i , t i , d , t

(4)

1N RS  RS t Ni1 i,t

(5)

ask bid

The market liquidity is

Where D is number of trading days,‟d‟ is the day when bid and ask is calculated, and N is the number of securities. The third measure, the Amihud’s illiquidity measure is based on the price impact to the order flow and is calculated as the ratio of the price movement to the trading volume and is defined as:

the market liquidity is

i, t |R | 1D i , d , t ILL   i , t d  1 D ( P * VOL ) i, t i , d , t i , d , t

(6)

1N ILL   ILL t i,t Ni1

(7)

14

Where, | R i ,t |

Pi ,t

and

VOL

i ,t

is the absolute return, price, and the volume of security i on date d and N is the

number of securities. The standard specifications used for all measures are as follows: 1) Calculated daily and then aggregated. 2) Stocks with trading days greater than 15 per month are considered. 3) Only NYSE and AMEX stocks are included (NASDAQ data includes inter-dealer trades that would inflate the actual trading volume and hence excluded). 4) Share price less than $5 and more than $1000 are excluded. There are several caveats worth mentioning.

First, the 'apparent' coarseness induced by the

estimation of the market liquidity from the daily data is defended in Goyenko, Holden, and Trzcinka (2009) as „‟we generally conclude that liquidity measures based on daily data provide good measures of highfrequency transaction cost benchmarks (i.e. liquidity measures do measure liquidity)‟‟. Second, Chordia, Roll, and Subramanyam (2001) report that the daily market spread declined after the 1997 tick-size reduction. As a result, we study the whole sample for the period 1986-2009 to get an overview of the liquidity dynamics and then, study the sub-sample for the period of 1998-2009.

The selection of the sub-sample

1998-2009 isolates, if any, issues induced by the tick-size reductions.

4.2.2. Corporate Bonds Liquidity Measures In addition to the stock liquidities, the corporate bond liquidity as one of the market liquidity proxies is also considered. Given the macroeconomic nature of this article, and not to deviate from the objectives of the study, we concentrate only on Moody's AAA and BAA corporate bond yields to find the corporate bonds liquidity. The vantage point of this selection is that the bonds included in the sample are all seasoned 15

issues with a duration of around 30 years. Moody's also drops bond from the above two indices if bonds are susceptible to redemption or rating downgrades.

Thus, the selection of these two corporate bond classes

have the advantage from market-microstructural point of view in that the difficulties in calculating the corporate bond liquidity such as the 'off-the-run illiquidity' and the 'duration induced liquidity differences' etc. are avoided.3 The expansive literature on the corporate bond liquidity measures exists and as is with the stocks liquidity measures, the measurement of the corporate bond liquidity is rather subjective. The traditional direct measure of liquidity is often not suitable and/or unreliable for the corporate bonds and indirect proxies are often used (see Houweling, Mentink and Vorst, 2004). Houweling, et al. considers nine different proxies for the corporate bond liquidities and report that these measures are not significantly different.

Bao, Pan,

and Wang (2009) propose a corporate bonds liquidity proxy that closely resembles Roll (1984) measure of the liquidity. To capture the corporate bond liquidity, a variant of one of the proxies used by Houweling et al. is used. The variant given in equation (8), to our knowledge, is not used before to proxy for the corporate bond liquidity. The proxy is based on the same argument of 'inventory cost' that market-microstructure literature utilizes.

The inventory cost is higher if there is uncertainty in the predictability of the future

yields and thus, the 'dispersion of yield' proxies the illiquidity: the higher is the dispersion in yield, the lower is the liquidity. The proxy is denoted as yRoll, since the fundamental of the liquidity proxy in equation (8)

3

We do recognize that the inclusion of the 30-year Treasury bonds liquidity would have been an apt choice. The issue is that the 30-year Treasuries were

withdrawn in 2002 and were reintroduced in 2006. As a result, a full sample (1986-2009) or the sub-sample (1998-2009) of 30-year Treasuries is not available. Given the fact that Moody's AAA and BAA are of 30-year duration and are included in the analysis, the Treasury bonds with lower duration may not be a suitable option for a comparative study purpose. Moreover, 20-year Treasury daily data is available only from 1993; finding bonds liquidity from lower frequency monthly data, which is available for the whole sample, may not capture the liquidity dynamics well. control variables, I do not opt for 10-year Treasury in finding bonds liquidity measure.

16

Additionally, as I use 10-year Treasury yield in „Term‟, one of the

is still in the spirit of the Roll's 'effective bid-ask price' and yRoll is defined as yRoll  Cov (  y , y ) t t t  1

Where,

yt

(8)

is the change in the yields between the day „t‟ and „t+1’. The difference between the equation

(8) and equation (2) is the 'return' in Roll (1984) is replaced with the 'yield'. The proxy proposed in equation (8) is also similar to Houweling et al. 'dispersion of yield'. While Houweling et al. used intra-day yields to find the 'dispersion of yield' proxy, we use inter-day yields to calculate yRoll; in Bao et al., the 'yield' of (8) becomes the 'price'. Using daily yields of the Moody's AAA and BAA bonds indices, we calculate yRoll daily for each bond index and then, aggregate it over 3 months to calculate the quarterly liquidity measures for the respective bond categories. The inclusion of two different bond ratings and the respective liquidities, which we refer to as

yRoll

AAA t

and

yRoll

BAA t

, affords me to conduct additional robustness checks. The market

liquidity index that includes the corporate bond liquidities and the stocks liquidities described in the previous section is presented in Table 2.

[Table 2 of the market liquidity index here]

The contemporaneous nature of the assets‟ liquidity and the volatility can be addressed by the inclusion of control variables that captures the asset volatility. The importance of controlling for volatility changes for explaining liquidity changes is pointed out succinctly, since ''Intuitively, even in more general economics, anything that causes asset risk to rise will steepen the demand curves.” (Johnson,(2008).

We

include the 'market implied volatility index' (VIX), calculated from the S&P 500 stock index option prices to 17

control for the near-term market anticipated volatility for stock prices and hence, the stock liquidity. As for the changes in the liquidity of the bond prices, we include the term spread (henceforth, Term), which acts as an implied volatility in the fixed income securities market and Term is calculated as the difference between the yield on the 3-month Treasury-bill and the yield on a 10-year Treasury bond index.

5. Empirical Results Empirical results are presented as follows. Section 6.1 specifies the transformations of time-series data for stationarity and also presents Granger causality test along with the most important of the causality results.

Section 6.2 presents the regression results of aggregate assets-growth of the financial

intermediaries and the market liquidity index. Section 6.3 presents the regression results of aggregate liabilities-growth of the financial intermediaries and the market liquidity index. Section 6.4 presents the regression results of aggregate leverages-growth of the financial intermediaries and the market liquidity index.

5.1. Stationarity and Causality We start the empirical analysis by creating a value-weighted index of the monetary liquidity using the benchmark monetary channels.

The balance-sheets variables of each group are aggregated for each-

period (quarter) and corresponding value-weights are assigned for each category based on the total value of the variables of the entire financial intermediary under investigation. This time-varying weights account for the time-varying growth of variables in each category. Then, we transform the data as a growth rate of the value-weighted variables. For example, the assets-growth variables of the transformed data are represented as dBank, dMutual, dBrokerDealer, dRetire, and dShadow.

Another example, the value-weighted

transformed leverages-growth is similarly represented; for example, value-weighted leverage growth of the 18

securities brokers and dealers is represented as dBrokerDealerlev.

These transformations result in the

stationary time-series data for the financial intermediaries' balance-sheets variables. A transformation of the market liquidity measures using the difference of the log of each measure is done to ascertain stationarity of the time-series data. The ROLL, the RS, yRoll_AAA, yRoll_BAA, and the ILL are represented as dRoll, dRS, dyRoll_AAA, dyRoll_BAA, and dILL respectively.

Similarly, we

transform VIX and Term , by taking the difference of the log, to dVIX and dTerm. With the exception of dBrokerDealer, Granger causality tests for up to 10 lags show all the variables Granger cause each other.

The Granger causality result of dBrokerDealer with the market liquidity

measures are shown in Table 3.

While the securities brokers and dealers' leverage and the market

liquidities Granger cause each other, other balance-sheets variables show mixed causalities. With the exception of dyRoll_BAA, the market liquidity measures seem not to cause the assets-growth or the liabiltiesgrowth of the securities brokers and dealers. This result seems to suggest that the market liquidity determines how much leverage the securities brokers and dealers take, but it can't explain the assets or the liabilities growth. To what extent the traders‟ funding or the leverage explains the market liquidity?

This question is

investigated, along with the other determinants of the market liquidity, in the next three sections.

[Table 3 of the Granger causality here]

5.2. The assets-growth of the financial intermediaries and the market liquidity In this step of the analysis equation (1) is used and market liquidity proxies dRoll, dRS, dyRoll_AAA, dyRoll_BAA, and dILL, one at a time, is regressed as the dependent variable and the monetary liquidity proxies dBank, dMutual, dBrokerDealer, dRetire, and dShadow as the explanatory variables. The control variables are dVIX and dTerm. For brevity, the predictability of the market liquidity dRoll and dyRoll_AAA 19

in terms of the monetary liquidity proxies are reported (the rest of the results are available upon request).

[Table 4 – the relation between dRoll and the asset-growth here] [Table 5 -- the relation between dyRoll_AAA and the asset-growth here]

The regression results are organized in two sub-sections: one describes the liquidity dynamics and the other describes the market-microstructural implications of the results.

5.2.1 The Liquidity Dynamics First, irrespective of the choice of the market liquidity measures, the result is overwhelmingly in favor of the dShadow, the aggregate assets-growth of the shadow-banks. Also note the consistent negative sign of the coefficients of dShadow and the larger coefficients of dShadow during the sub-sample 1998-2009 compared to the coefficients during the whole sample 1986-2009. Second, the dBrokerDealer also enters the result but not as strongly as dShadow.

While the relationship of dBrokerDealer with the stocks

liquidities is negative, with the bonds liquidity the relationship is mixed. As for the predictability of the market liquidities from the aggregate assets-growth variables of the rest of the financial intermediaries, the result is not conclusive in that the statistically significant coefficients have alternating positive and negative signs. Though the shadow-banks do not directly contribute to the stocks and bonds market liquidity because they, in general, do not hold any stocks or bonds, the inverse relationship between the aggregate assetsgrowth of the shadow-banks and the market liquidity index is nevertheless important.

20

We have to recognize

that the market liquidity index used in this study is just a proxy for the market illiquidity4 and the market liquidity index was designed to reflect the transaction-cost and/or price-impact of trading volume of a risky security in the market.

As a result, the relationship has a broader implication in that the market illiquidity

makes it harder for the shadow-banks to lend, which is reflected in the negative assets-growth of the shadowbanks in an illiquid market. Since the lending ability of the shadow-banks, to a large extent, is influenced by their ability to securitize the assets they hold; the broad market illiquidity affects the securitization process and hence, the lending gets affected.5 The predictability persists whether the market liquidity yardsticks are the stocks or the corporate bonds. The larger coefficients in the sub-sample 1998-2009 point to the better integration of dShadow with the market liquidity conditions. To ascertain that the above results are not influenced by the recent crisis, a separate analysis is conducted with the sample from 1986 to 2006.

Unreported results show that the predictability for the bonds

liquidity measures remain unaltered, though the value of the coefficients and the significance change and that the dShadow loses its predictability for the stocks liquidity in that it is not consistent among the stocks liquidity measures. This shows that while the integration of the shadow-banks to the bonds liquidity is consistent irrespective of the economic cycles, in economic downturns the aggregate assets-growth of the shadow-banks predict the overall market liquidity, i.e. both the stocks and the bonds liquidity, condition. The inverse relationship of the dBrokerDealer to the stocks liquidity point to the funding constraints of the securities brokers and dealers face in an illiquid market but the result is not conclusive. The reason is that the statistically significant coefficients either don't show up in both the sample and the sub-sample periods or the sign of the statistically significant coefficients are not the same. The results do not change

4 Note that all liquidity measures used in the study measures the illiquidity. 5 For example, new CMBS issuance grew from about $40 billions in 2000 to $230 billions in 2007, dropped to $1.4 billions in 2009, and could reach $10 billions in 2010 (Data source: 'CMBS Market Rises from Ashes of Collapse,' The Wall Street Journal, C6, July 21, 2010).

21

even if the sample is changed to pre-crisis 1986-2006 period. The inconclusive results in explaining the market liquidity in terms of dBank, dMutual, and dRetire are, in part, could be explained by the operational differences of these groups of financial intermediaries from the shadow-banks and the securities brokers and dealers. Duration-mismatch is a pressing concern for the banking category of the financial intermediaries and the results reflect the fragility of the banking system in that the banks have no choice but to lend in longermaturity assets as their liabilities (deposits) grow. As a result, the alternating signs of the coefficients of dBank may as well be nothing but statistical association and do not reveal any important information about the liquidity dynamics. It is also possible that the assets-growth would be negatively related to the market liquidity if growth of deposits is the reason for the assets growth as investors move funds from risky asset market, and it would be negatively related for other sources of asset growth. Given the fact that mutual funds invest primarily in the stocks and bonds, the absence of consistent predictability of the market liquidity derived from the aggregate assets-growth of the mutual funds, at first glance, is counter intuitive. The probable reasons stem from the structural make-up of this study, which is rather extensive in the sense that the 'ceteris paribus' condition is largely relaxed. Unreported results show that when other explanatory variables are omitted, the aggregate assets-growth of the mutual funds could predict the market liquidities and the assets-growth and the market illiquidity are inversely related. Additionally, there is an undercurrent of liquidity dynamics within the mutual funds category.

As

noted earlier, the mutual funds group of the index is composed of a) the money market funds b) the mutual funds, and c) the closed-end funds. Under the normal business cycles or in a buoyant market liquidity condition, while the investors pour in money into the mutual funds and closed-end funds, money flow into the money market funds remain the same or it increases. Under the financial duress, two things could 22

happen concurrently: a) corporations would horde part of their income or even, their line of credit 6 in the money market funds b) investors would substitute the à la carte equity mutual funds and the closed-ends funds for the money market funds and other safer investment vehicles. The end results are that the relationship is not conclusive and there exist a positive relational bias of dMutual with the market illiquidity condition. The result supports the phenomenon of “flight to quality or safety” that previous literature on liquidity found.. The relation of the market liquidity with the dRetire, the aggregate assets-growth of the pension funds, life insurance companies and the property and causality insurance companies is also inconclusive.

Though

this group constitutes a large share of the total 'money-flow' into the economy, most of these financial intermediaries are rather passive (though the influence of pension funds such as the CalPERS and insurance companies such as the AIG on the market is well-known)7 and their aggregate assets-growth have no predictive power for the market liquidity.

5.2.2. Market-Microstructural Implications From the market-microstructural perspective, there are several important findings. First, the results point to the commonality of the market volatility in the market liquidity irrespective of the time-variations in liquidity or the reduction in the spread after 1997. Second, the measurement of the liquidity proxy for the corporate bonds from their yields is validated and supported. The contemporaneous nature of the market liquidity and the asset volatility is strongly supported from the results (Table 4-5) in the coefficients and the corresponding statistical significance of dterm and

6 See Ivashina and Scharfstein (2009) on drawing down of line of credit right by corporations after the failure of Lehman Brothers in 2008. 7 'CalPERS effect' is hotly debated and researched; its investments in CDOs and other 'unethical investments' instruments draw attention from a wide-range of stakeholders.

AIG was 'too big to fail' in the recent financial crisis.

23

dVIX. This result is consistent with the market-microstructure researches and with the theoretical prediction of Brunnermeir et al. (2009) that volatility and the liquidity are cointegrated. The results also show the apt choice of the control variables. The result is robust in that, while VIX largely explains the stock liquidities, the Term largely explains the corporate bonds liquidity.

Another important observation is that the corporate bonds liquidities, though

not consistently, not only depend on the dTerm but also on the dVIX. The result seems to imply that while the overall riskiness of the market (i.e. VIX) is reflected in the corporate bonds liquidities, the term-spread (i.e. Term) does not have statistically significant impact on the liquidity of the stocks. Bao et al. (2009) also find the similar relationship between the corporate bonds liquidity and the VIX. This result brings several interesting points to the table: a) while Bao et al. (2009) arrive at the above conclusion based on a 'high-frequency' transaction-level Trace dataset, this study finds the relation based on 'low-frequency' Moody's corporate bonds dataset, b) while the corporate bonds liquidity proxy of Bao et al. (2009) is based on the 'bond-price', this study relies on the 'bond-yield' in deriving the corporate bonds liquidity proxy. Hence, the conjecture, as explained in section 5.2.2, that 'dispersion in yield' of 'lowfrequency' data may capture the corporate bonds liquidity seems to have foundational basis. Additionally, the results support the findings of Goyenko et al. (2009) that liquidity measures based on daily data 'do measure' liquidity and are equivalent to their high-frequency counterpart. However, the Trace dataset used by Bao et al. is available since 2002 and the sample period for this study runs from 1986 to 2009 and both studies might have been influenced by the recent financial crisis. To account for this, a separate study with the sample from 1986 to 2006 is conducted, i.e. the recent crisis period is excluded, and unreported results show that the corporate bond liquidities are not related to VIX. This phenomenon could be explained by the fact that while in the normal business cycles, corporate bonds traders 24

considers the riskiness of the fixed-income securities, in a market downturn of '2007-2008 magnitude' corporate bonds are priced to the riskiness of the whole market i.e. the riskiness of both the stocks and the bonds are priced. In other words, in a market downturn, investors would shuffle their investment portfolio to bond portfolio for a safety. Other important observation is that the similarities in magnitude of the coefficients of the control variables, i.e. the VIX or the Term, in both the sample and sub-sample periods.

In spite of the declination in

the market spread after the 1997 tick-size reduction, the relation of the assets' volatility to the assets' liquidity remained unaltered, which substantiates the contemporaneous nature of the volatility and the liquidity. Based on the results it may be concluded that the tick-size reduction or the market spread declination or the timevariations in the market liquidity have minimal effect on the contemporaneity of the volatility and liquidity. To summarize the result based on the assets-growth variable of the financial intermediaries, we find evidence that a) irrespective of the market liquidity benchmarks, the aggregate assets-growth of the shadowbanks have predictive power for the market liquidity, b) the aggregate assets-growth of the securities brokers and dealers weakly predicts the stocks liquidity and have no predictive power for the bonds liquidity c) the market volatility largely explains the market liquidity (for a 1% change in liquidity, volatility accounts for 0.2% to 0.73%).

5.3. The liabilities-growth of the financial intermediaries and the market liquidity As noted earlier, in this part of the analysis the liabilities side of the financial intermediaries is investigated. This selection naturally excludes one of the most important financial intermediaries, the mutual funds group from the study. The same equation (1) is used and the components:

Z

t

vector has the following

dBank, dBrokerDealer, dRetire, and dShadow (see Table 1) and the nomenclatures of the 25

explanatory variables are the same as are in previous section. The dependent and control variables remain the same as are in Section 6.2.

For brevity, the results are not presented but available upon request.

Once again, the predictability of the market liquidity in terms of the dShadow is noticeable. The result is not as conclusive in that the predictability of dRoll, dRS, and dyRoll_BAA in terms of dShadow is absent. However, whenever dShadow explains the market liquidity, the sign of the coefficients are positive and this implies and, as expected, the liabilities-growth is positively related to the market illiquidity conditions.

As for the relation between the market liquidity and other financial intermediaries‟ liabilities-

growth, the results are inconclusive.

5.4. Financial intermediaries’ leverage-management and the market liquidity As noted earlier, the leverage is defined as the difference of the assets and the liabilities. The issue with a study with only leverage is that financial intermediaries in the shadow-banks components have net zero-leverage i.e. their balance-sheets is balanced and shadow-banks can not be included in the study. However, from the balance-sheet management perspective, the financial intermediaries included in the shadow-banks are probably the most efficient in that they completely match their assets and the liabilities in terms of maturity, credit risk or liquidity.

Nevertheless, for a comparative study of the liquidity

management of the financial intermediaries and the liquidity dynamics, the aggregate leverage-growth is probably the most important information.

In the one hand, the study sheds light on the market-based

financial intermediaries‟ balance-sheets management.

On the other hand, for relationship-based

intermediaries, the leverages-growth may provide the dynamics of the balance-sheets management under the market liquidity constraints. To account for the above observations, we go one-step down the balance-sheets information by 26

including the leverages of the individual financial intermediaries and naturally, non-zero leverage is only considered. This results in the exclusion of the shadow-banks and the mutual-funds. Moreover, even though the pension funds accounts may have unbalanced (non-zero) balance-sheets, they are excluded, on the ground that the pension funds for the most part is passive. The same equation (1) is used with

Z

t

vector having the following components: dCommercialBanklev, dCreditUnionslev, dSavingInstituteslev, dBrokerDealerlev, dLifeInsurancelev, and dPropertyInsurancelev, where, as usual,‟d‟ implies the transformed data of the leverages.

The dependent and control variables remain the same as are in previous

sections.

[Table 6 – the relation between dRoll and the leverages-growth here] [Table 7-- the relation between dyRoll_AAA and the leverages-growth here]

The analysis of individual financial intermediaries and their leverages-growth unfolds that the dBrokerDealerlev and dSavingInstitutelev enter the regressions results consistently. The predictability of the market liquidity from the leverages-growth of rest of the financial intermediaries is rather weak in that the predictability is not present in both the sample and the sub-samples. While the relation of dCommercialBanklev and dCreditUnionslev to the stocks liquidity, though inconclusive, is generally negative, the relation to the corporate bonds liquidity is positive. The dLifeInsurancelev, and dPropertyInsurancelev generally show a positive relation to the market illiquidity condition but not all five market liquidity yardsticks can be explained by these variables. We would devote rest of this section on dBrokerDealerlev and dSavingInstitutelev, the balance-sheet variables that enters the regression results consistently. 27

Note that irrespective of the market liquidity measures, the leverage of the dBrokerDealerlev is in general inversely related to the market illiquidity and this is consistent with the prediction of Brunnermeir et al. (2009) that the funding constraints of the traders, in part, dictates the market liquidity. As noted earlier, the funding source of the securities brokers and dealer are the repos and the reverse-repos and the market illiquidity adversely affects the repo market and this in turn reduces short-term borrowing and lending ability of the securities brokers and dealers and the leverages contracts. Also note the similarities in the magnitude of the coefficients of dBrokerDealerlev, which implies that the relationship remained unaltered throughout the sample and the sub-sample periods. This points to the consistency of the securities brokers and dealers leverage in predicting the time-varying market liquidity. While dSavingInstitutelev is negatively correlated with the stocks market liquidity, the relation of the leverage of the savings institutions to the bond market liquidity is primarily positive. Also note the larger coefficients of dSavingInstitutelev during the 1998-2009 period compared to the coefficients of the whole sample. To interpret the result, we step back and point out the key historical perspectives on the savings institutions. The savings institutions are primarily known to be funded by the deposits they receive in the form of CDs and their assets-side are generally composed of the real-estate investments and other loans and leases. Thus, the mode of funding for the savings institutions is very different from the funding sources of the commercial banks, credit unions (both of which, along with CDs, are also funded by other sources) and other market-based financial intermediaries.

If the seed of the recent financial crisis lies in the „sub-prime

mortgage crisis‟, the financial crisis in the early 1990s could be traced back, to some researchers, to the savings institutions‟ imprudent investments in the real-estates.

In addition, the savings institutions‟

duration-mismatch issue was also cited to be the reason for the early 1990's financial crisis. 28

The forefront

of the savings institutes‟ funding was the „deposit-brokers‟, who attracted potential customers with „highyield‟ CDs defying underlying market-implied yields.

To make things worse, there were no „floor‟ for how

much of the deposit could be generated through the „deposit-brokers‟ channel.

The result was detrimental

to the basic structure of the savings institutions: along with the inherent duration-mismatch issues that other relationship-based financial intermediaries face, the savings institutions were forced to investment in unsound real-estate projects to counter the explosion in the liabilities-side.

The introduction of the

regulations,8 the numerous forced close-downs, and the government bailouts, resulted in the reductions in the number of the savings institution from 3740 in 1986 to 1173 in 2009.9 Given the above background, the larger coefficients of dSavingInstitutelev in the 1998-2009 sample seems to reflect the reduction in the market-frictions in that the savings institutions took advantage of the innovations in the financial products. The duration-matching issues were also probably aided by the lower volatility in the interest rates during the 1998-2009 periods. Overall, the result implies that the savings institutions' aggregate balance-sheets are better integrated with the market liquidity conditions during the 1998-2009 periods. The better integration in the sub-sample could also be explained from the recent financial crisis and this is discussed later in the article. While the negative correlation of dSavingInstitutelev with the equity market liquidity seems to suggest the expansion in the liabilities-side relative to the assets-side of the balance-sheets, the positive relation of dSavingInstitutelev with the corporate bonds liquidity implies just the opposite. Judging from the depositors‟ perspective or the liabilities-side, in the one hand, as the stocks market illiquidity increases, investors seeks a safer haven in the savings deposits and in the other hand, with the bonds market illiquidity, hence variability in the savings deposits yields, investors seems to redeem their positions or simply 8 9

Financial Institutions Reform, Recovery, and Enforcement Act of 1989 Source: FDIC report on the savings institutions

29

disintermediation occurs.

Judging from the assets side, the negative coefficients seem to suggest that the

stocks market illiquidity negatively impacts the assets value but the positive-relation of dSavingInstitutelev can‟t be logically explained in that it implies that the variability in the bonds market yields increases the assets value. Putting the results together and from the savings institutions perspective, we can argue that: a) the stocks market illiquidity commoves with both the assets and the liabilities side of the balance-sheets b) the corporate bonds illiquidity negatively impacts the liabilities-side. How did the recent financial crisis affect the relationship?

With a sample from 1986-2006,

unreported results show that, for the most part, the predictability of the market liquidities for dSavingInstitutelev remain unaffected and this could be explained by the fact that the Saving Institutions did not and/or could not respond to the market illiquidity conditions in the recent crisis. However the relationship of dBrokerDealerlev to the market liquidity is not absolute in the 19862006 samples: only the market liquidity predictability based on the Roll, RS, and yRoll_BAA are present. Putting the leverages-growth variables of financial intermediaries in perspective, this is the most important finding. This point to the securities brokers and dealers‟ alignment to the market illiquidity in that the adverse market liquidity conditions affected their leverages considerably and all five market liquidity measures were negatively related to the leverages. The securities brokers and dealers' aggregate leverages-growth provide a cleaner signal of the monetary liquidity conditions in that their balance-sheets variables are primarily driven by the short-term borrowing and lending and the relationship-based transactions are practically absent. As a result, the fluctuations in their aggregate leverages are found to be intimately related to the fluctuations in the market liquidity and are more pronounced during the financial crisis.

30

6. Summary of the results The important empirical results of the previous three sections could be summarized as follows: a) The market-based financial intermediaries‟ aggregate balance-sheets variables predict the market liquidity in that the funding constraints of the traders represented by the leverages-growth of the securities brokers and dealers explain the market liquidity. b) The aggregate assets-growth of the shadow-banks predicts the market liquidity and the integration of the shadow-banks to the market liquidity improved in the 1998-2009. c) The leverages-growth of the savings institutions, which are not a pure play market-based financial intermediary, has the predictive power for the market liquidity and the relationship is stronger in the 1998-2009. d) There is commonality factor embedded in the aggregate balance-sheets variables of three out of the five financial intermediaries considered in this study. e) The market volatility largely explains the market liquidity (for a 1% change in liquidity, volatility accounts for about 0.20% to 0.73%). f)

While the stocks liquidity is explained by the stocks market volatility, the corporate bonds liquidity may depend on both the stocks and the bonds market volatility. The relationship is unaffected by the 1997 tick-size reduction or subsequent declination of the spread or the timevariations of the market liquidity.

g) The corporate bonds liquidity can be proxied by the 'daily dispersion in yield' provided the bonds are of the similar maturities and seasoned.

7. Concluding Remarks 31

The primary contributions of this paper are two folds.

First, the paper provides the link between the

monetary liquidity and the market liquidity in a unified framework by including a large number of 'macroliquidity' channels and an index of 'micro-liquidity' measures drawn from the stocks and bonds universe. Second, the paper finds one of the dimensions of the commonality in 'microstructure liquidity' in the variations of the aggregate balance-sheets variables of the financial intermediaries.

Thus, the paper

contributes not only to the understanding of the market-frictions in credit-supply but also to the broader understanding of the market-microstructure. While the liquidity dynamics between the macro-liquidity and the micro-liquidity so far have been studied either by considering the gross macro-variables such as the GDP or by including the „money-flow‟ from a sub-set of financial intermediary, this article provides a broader perspective on the liquidity dynamics by including the „money-flow‟ from a large number of financial intermediaries. The robustness of the results is ensured by the inclusion of a large set of proxies for both the monetary and the market liquidities. An additional contribution to the market-microstructure literature in that the results provide an alternative measure of the corporate bonds liquidity based on the 'daily dispersion in yields' although the outcome is primarily driven by the data. The robustness was ascertained by substantiating the results with the findings of the existing literature on the corporate bonds liquidity measures. There are several policy implications of the results. First, the result may find applications in macro-prudential policy issues. The 'liquidity dry-up' conditions that are typically associated with the financial crises could be addressed by preemptive financial policies that focus on the financial intermediaries that are intimately linked to the market liquidity. For example, one of the plausible short-term financial policies geared towards the financial crises may involve recapitalizing financial intermediaries such as the securities brokers and dealers that are at the core of delivering the market liquidity. 32

The balance-sheets variables of both the shadow-banks and the savings institutions are found to be linked to the market liquidity. The operational structure of both the financial intermediaries are similar in that both borrow short-term (the shadow-banks mainly through commercial papers and the savings institutions mainly through one-year CDs) and lend long-term. The primary difference between the two is that the savings institutions are regulated and the shadow-banks are not. From regulatory perspective this result may provide important input in future banking regulations/policies. One application may involve revisiting, re-tweaking, and applying the regulations that did the 'right-thing' for the savings institutions to other regulated financial intermediaries. Moreover, the existence of shadow-banks is primarily driven by the absence of these intermediaries to follow regulations that are imposed on the commercial banks. The results suggest that the shadow-banks were able to respond to the fluctuations in the market liquidity by avoiding regulations. This may provide additional input to the future banking policies. Future paths of research may concentrate on 'high-frequency' version of this study to get a shorterhorizon perspective on the liquidity dynamics.

Interesting topics to follow up would be: a) to conduct

upstream analysis o liquidity chain i.e. to analyze the reverse relationship from financial intermediary to macro factors such as discount windows and monetary policies. b) to count economics of crisis, as financial crises can be understood as

liquidity crises, c) to analyze the post-regulation „apparent operational

efficiency‟ of the savings institutions, d) to study the relation between each of the financial intermediaries to the market liquidity, and, e) to find the key similarities and the differences between the shadow-banks and the savings institutions and this may help formulate future policies targeting the financial intermediaries.

33

References 1) Adrian, Tobias, Moench, Emanuel, and Shin, Song Hyun, “Financial Intermediation, Asset Prices, and Macroeconomics Dynamics”, Federal Reserve Bank of New York, Working Paper, 2009. 2) Adrian, Tobias, and Shin, Song Hyun, ''Liquidity, Monetary policy, and Financial Policy,'' Current Issues in Economics and Finance, Federal Reserve Bank of New York, 2008. 3) Adrian, Tobias and Shin, Song Hyun, “Money, Liquidity and Monetary Policy,” American Economic Review, 2009. 4) Amihud, Yakov, Mendelson, Haim, and Pedersen, Lasse Heje, “Liquidity and Asset Prices,” Foundation and Trends in Finance, 2005. 5)

Bao, Jack, Pan, Jun, and Wang, Jiang , “Liquidity of Corporate Bonds,” Working Papers, MIT, 2010.

6) Brunnermeier, Markus K., and Pedersen, Lasse Heje, ''Market Liquidity and Funding Liquidity'', Review of Financial Studies, 2009. 7) Chordia, Tarun, Sarkar, Asani, and Subrahmanyam, Avanidhar, “An Empirical Analysis of Stock and Bond Market Liquidity,” Review of Financial Studies, 2005. 8) Chordia, Tarun, Roll, Richard, and Subrahmanyam, Avanidhar, “Market Liquidity and Trading Activity,” Journal of Finance, 2001. 9) Douglas, Diamond, and Dybvig, Philip, “Bank Runs, Deposit Insurance, and Liquidity,” Journal of Political Economy, 1983. 10) Douglas, Diamond, and Rajan, Raghunath, ''Liquidity risk, liquidity creation and financial fragility: A theory of banking'', Journal of Political Economy, 2001. 11) Fujimoto, Akiko, ''Macroeconomic Sources of Systematic Liquidity,'' working paper, Yale University, 2004. 12) Gorton, Gary, and Winton, Andrew, “Financial Intermediation,” University of Pennsylvania, Financial Institutions Center, 2002.

13) Gorton, Gary and George Pennacchi, "Banks and Loan Sales: Marketing Non-Marketable Assets," Journal of Monetary Economics (1995). 14) Houweling, Patrick, Mentink, Albert, and Vorst, Ton, “Comparing possible proxies of corporate bond liquidity,” Journal of Banking and Finance, 2004. 34

15) Ivashina, Victoria, and David S. Scharfstein, “Bank Lending During the Financial Crisis of 2008,” Journal of Financial Economics, forthcoming, 2009. 16) Johnson, T., "Dynamic Liquidity in Endowment Economies,” Journal of Financial Economics, 2006. 17)

Johnson, T., “Volume, Liquidity, and Liquidity Risk,” Journal of Financial Economics, 2008.

18) Pastor Lubos and Robert F. Stambaugh, “Liquidity risk and expected stock return”, Journal of Finance, 2001. 19) Pozsar, Adrian, Ashcraft, Boesky, “Shadow Banking”, Federal Reserve Bank of New York, Report #458, 2010. 20)

Randi Næ s, Johannes A. Skjeltorp and Bernt Arne Ø degaard ''Stock Market Liquidity and the Business Cycle'', Journal of Finance, Forthcoming, 2010.

35

Table 1: The Monetary Liquidity Indices and their descriptive statistics The table display the monetary liquidity measures during the sample period 1986-2009. Panel A show the components of the five financial intermediaries groups: Banks, Mutual Funds, Shadow-banks, Securities broker-dealers, and Retirement account/life insurance companies. Panel B shows the descriptive statistics of the value-weighted assets, Panel C shows the descriptive statistics of the value-weighted liabilities and Panel D shows the descriptive statistics of the value-weighted leverages for two sample sub-periods.

Panel A: Components of the financial intermediaries groups Retirement / Life Insurance

Shadow Banks

Banks

Mutual Funds

Private Pension Funds

Asset Backed Securities

Commercial Banks

Money Market Funds

Life Insurance Companies

Agency/GSE Mortgage Pools

Credit Unions

Mutual Funds

Federal Govt. Retirement Funds

Funding Corporations

Savings Institutions

CEFs and ETFs

State Govt. Retirement Funds

Finance Companies

Property and Casualty Insurance

Panel B: Descriptive Statistics of the value-weighted assets (in millions) index 1986-1997

1998-2008

Mean

Median

Std Dev

Mean

Median

Std Dev

Retirement/Life Insurance

1695087

1637398

123248

2617578

2589516

733244

Shadow Banks

212131

138089

177628

1324126

1220345

371978

Banks

476402

514572

167439

1850077

1937192

707955

Mutual Funds

8752

7392

6314

79647

63050

47282

Securities Broker Dealers

1987896

1878045

737762

4096913

4079749

527786

Panel C: Descriptive Statistics of the value-weighted liabilities (in millions) index 1986-1997

1998-2009

Mean

Median

Std Dev

Mean

Median

Std Dev

Banks

2128633

2092627

131320

3206091

3208789

785285

Shadow Banks

143516

135613

72109

984824

881682

474595

Securities Broker Dealers

10602

8445

8491

110673

87073

67701

Retirement/Life Insurance

2143398

1993082

866304

4765492

4734822

640541

Panel D: Descriptive Statistics of the value-weighted leverages (in millions) index 1986-1997

1998-2009

Mean

Median

Std Dev

Mean

Median

Std Dev

Commercial Banks

34677

34430

17003

454834

416145

272203

Credit Unions

652

498

384

1675

1769

237

Savings Institutions

4543

3502

3477

10580

10631

2185

Securities Broker Dealers

1325

1335

358

2008

1747

804

Life Insurance Companies

14225

11548

6280

29963

30448

3296

Property and Casualty Insurance

59483

56988

23946

95485

103224

24553

36

Securities Brokers & Dealers Securities Brokers & Dealers

Table 2: Summary Statistics of the Market Liquidity Measures Panel A shows the descriptive statistics for the market liquidity measures during the sample period 1986-2009. The stock liquidity measures are a) ROLL is the effective bid-ask spread of Roll (1984) b) RS is the relative bid-ask spread, c) ILL is the Illiquidity measure of Amihud (2002) (a, b, and c are calculated from the CRSP database). The corporate bond liquidity measures are a) yRoll_AAA is the liquidity measured of the Moody‟s corporate AAA rated bond index and b) yRoll_BAA is the liquidity measured of the Moody‟s corporate BAA rated bond index (data source: St. Louis Fed). Panel B shows the correlation among different market liquidity measures. All the „liquidity measures‟ measure illiquidity.

Panel A: Descriptive Statistics 19862009

19861997

Liquidity

19982009

Standard

Standard

Standard

Measures

Mean

Median

Deviation

Mean

Median

Deviation

Mean

Median

Deviation

ROLL

0.0190

0.0180

0.0053

0.0183

0.0178

0.0029

0.019744

0.018936

0.007047

RS

0.0237

0.0217

0.0075

0.0216

0.0209

0.0035

0.026086

0.02363

0.009699

ILL

0.3602

0.3259

0.1637

0.4687

0.4649

0.1435

0.241797

0.253511

0.081804

yRoll_AAA

0.0205

0.0208

0.0080

0.0172

0.0155

0.0071

0.024208

0.023252

0.007245

yRoll_BAA

0.0206

0.0202

0.0066

0.0179

0.0177

0.0060

0.023428

0.022021

0.006165

Panel B: Correlation between Market Liquidity Measures ROLL

RS

ILL

RS

0.9764

ILL

0.5054

0.3951

yRoll_AAA

0.3146

0.3249

0.1913

yRoll_BAA

0.2348

0.2420

0.1955

yRoll_AAA

0.7504

37

Table 3: Granger Causality Tests The table reports Granger causality tests between the securities broker-dealer aggregate growth of the assets, the liabilities and the leverage with a) ROLL, the effective bid-ask spread of Roll (1984) (dRoll) b) RS, the relative bid-ask spread (dRS), c) ILL, the Illiquidity measure of Amihud (2002) (dILL), d) yRoll_AAA, the liquidity measured of the Moody‟s corporate AAA rated bond index (d yRoll_AAA, )and e) yRoll_BAA, the liquidity measured of the Moody‟s corporate BAA rated bond index (d yRoll_BAA). The P-values for 10 lags for the sample period 1986-2009 is reported.

Assets

Liabilities

Leverage

Growth

Growth

Growth

dRoll

0.7682

0.8577

0.1594

dBrokerDealer

0.0000

0.0000

0.2350

dRS

0.8489

0.8577

0.3200

dBrokerDealer

0.0000

0.0000

0.1672

dILL

0.2607

0.2628

0.1490

dBrokerDealer

0.0028

0.0172

0.5697

dyRoll_AAA

0.2795

0.4706

0.0522

dBrokerDealer

0.0907

0.0171

0.9968

dyRoll_BAA

0.0128

0.0532

0.0258

dBrokerDealer

0.4591

0.1590

0.4930

a) dRoll H0:

dBrokerDealer

H0:

dRoll

does not cause does not cause

b) dRS H0:

dBrokerDealer

H0:

dRS

does not cause does not cause

c) dILL H0:

dBrokerDealer

H0:

dILL

does not cause does not cause

d) dyRoll_AAA H0:

dBrokerDealer

H0:

dyRoll_AAA

does not cause does not cause

e) dyRoll_BAA H0:

dBrokerDealer

H0:

dyRoll_BAA

does not cause does not cause

38

Table 4: The Market Liquidity and Aggregate Assets-Growth The table shows the regression results of the equation Y   t   t Yt  i   t Z t  j   t  t   t , where Y , the dependent variable is a scalar and represents one of the t t market liquidity measures; Z is the vector of explanatory variables dBank, dMutual, dBrokerDealer, dRetire, and dShadow;  is the vector of control variables t t dVIX and dTerm. The variables dBank, dMutual, dBd, dRetire, and dShadow, are the asset-weighted growth rate of the assets of Banks, Mutual Funds, Securities broker-dealers, Retirement Accounts/Insurance companies, and Shadow-banks (financial institutions that operate like banks but not regulated) respectively. The control variables are dVIX, the log differenced CBOE volatility index and dTerm, the log differenced term-spread. When VIX data is not available, VXO, an alternate measure of CBOE volatility index is used. The table shows the dependent variable dILL, which is the log difference of the market liquidity measure ILL, in terms of the explanatory and control variables. The subscript „i‟ changes from 1 to 5. The subscript „j‟ changes from 0 to 1; this selection allows us to see the dynamic relationship between financial intermediaries‟ balance-sheet and the asset liquidity. Sample period is from Jan1986 to Dec 2008; sub-samples are from Jan 1986 to Dec 1997 and from Jan 1998 to Dec 2009. T-statistics are in the parenthesis, all standard errors are Newey-West adjusted, and only statistically significant coefficients of Z and  are reported. t

t

Market Liquidity Measure

Explanatory variables

Y

t

is dRoll

Sample: 1986-2009

Sample: 1998-2009

Lags

Lags

0

1

2

3

4

5

0

1

2

3

dBrokerDealer -0.929 (-2.05)

1.963 (2.82)

-2.073 (-1.65)

dRetire -0.621 (-1.83)

dMutual Control Variables dVIX

5

-0.281 (-1.73)

dBank

dShadow

4

-0.316 (-2.41)

0.389 (2.13)

-1.482 (-1.76)

-1.435 (-2.15) -1.563 (-2.33)

-1.3831 (-1.69)

0.660 (9.91)

0.670 (14.21)

0.678 (12.08)

0.694 (12.89)

0.679 (12.98)

0.687 (12.09)

0.661 (8.60)

0.699 (12.32)

0.656 (8.14)

0.718 (12.91)

0.727 (13.59)

0.7338 (10.60)

0.76

0.756

0.75

0.75

0.76

0.75

0.81

0.75

0.76

0.76

0.78

0.7569

dTerm R

2

39

Table 5: The Market Liquidity and Aggregate Assets-Growth The table shows the regression results of the equation Y   t   t Yt  i   t Z t  j   t  t   t , where Y , the dependent variable is a scalar and represents one of the t t market liquidity measures; Z is the vector of explanatory variables dBank, dMutual, dBrokerDealer, dRetire, and dShadow;  is the vector of control variables t t dVIX and dTerm. The variables dBank, dMutual, dBd, dRetire, and dShadow, are the asset-weighted growth rate of the assets of Banks, Mutual Funds, Securities broker-dealers, Retirement Accounts/Insurance companies, and Shadow-banks (financial institutions that operate like banks but not regulated) respectively. The control variables are dVIX, the log differenced CBOE volatility index and dTerm, the log differenced term-spread. When VIX data is not available, VXO, an alternate measure of CBOE volatility index is used. The table shows the dependent variable dILL, which is the log difference of the market liquidity measure ILL, in terms of the explanatory and control variables. The subscript „i‟ changes from 1 to 5. The subscript „j‟ changes from 0 to 1; this selection allows us to see the dynamic relationship between financial intermediaries‟ balance-sheet and the asset liquidity. Sample period is from Jan1986 to Dec 2008; sub-samples are from Jan 1986 to Dec 1997 and from Jan 1998 to Dec 2009. T-statistics are in the parenthesis, all standard errors are Newey-West adjusted, and only statistically significant coefficients of Z and  are reported. t

t

Market Liquidity Measure

Y

t

is dyRoll_AAA

Sample: 1986-2009

Sample: 1998-2009

Lags Explanatory variables

0

Lags 1

2

3

4

5

0

1

3.300 (2.33)

0.807 (2.18) 4.921 (2.72) 3.842 (2.90)

dBrokerDealer dBank dRetire

1.983 (1.85) -2.425 (-2.40) -1.073 (-1.66)

dShadow dMutual Control Variables dVIX dTerm R

2

0.442 (2.38) 0.213 (3.22)

0.295 (4.88)

0.30

0.35

-2.519 (-2.20)

2

3

4

5

-0.948 (-1.85)

-1.983 (-1.65)

-5.596 (-2.92)

0.258 (4.40)

0.300 (1.82) 0.244 (4.24)

0.326 (1.90) 0.224 (5.01)

0.322 (2.27) 0.232 (3.60)

0.310 (4.41)

0.372 (5.31)

0.317 (4.66)

0.314 (5.55)

0.283 (6.60)

0.374 (5.72)

0.30

0.32

0.35

0.30

0.47

0.51

0.39

0.50

0.47

0.44

40

Table 6: The Market Liquidity and Aggregate Leverages-Growth The table shows the predictive regression results of Y   t   t Yt  i   t Z t  j   t  t   t , where Y , the dependent variable is a scalar and represents one of the t t market liquidity measures; Z is the vector of explanatory variables dBrokerDealer, dCommercalBank, dCreditUnios, dSavingInstitutes, dLifeInsurance, t dPropertyCasuality, which are the value-weighted growth rate of the leverages of Securities broker-dealers, Commercial Banks, Credit Unions, Savings Institutions, Life Insurance companies, and Property and Causality Insurance Companies, respectively;  is the vector of control variables dVIX and dTerm, where dVIX is t

the log differenced CBOE volatility index and dTerm is the log differenced term-spread. When VIX data is not available, VXO, an alternate measure of CBOE volatility index is used. The table shows the dependent variable dAAA, which is the log difference of the market liquidity measure AAA, in terms of the explanatory and control variables. The subscript „i‟ changes from 1 to 5. The subscript „j‟ changes from 0 to 5. Sample period is from Jan1986 to Dec 2009; sub-samples are from Jan 1998 to Dec 2009. T-statistics are in the parenthesis, all standard errors are Newey-West adjusted, and only statistically significant coefficients of Z and  are reported. t

t

Market Liquidity Measure

Y

t

is dRoll

Sample: 1986-2009

Sample: 1998-2009

Lags Explanatory variables

0

Lags 1

2

3

4

dBrokerDealerlev

5

0

1

-0.101 (-2.82)

dCommercialBanklev

2 -0.102 (-1.80)

-0.368 (-1.98)

dCreditUnionlev

-0.348 (-2.04)

dSavingInstitiuelev

-0.002 (-1.82)

dLifeInsurancelev

-0.003 (-1.71)

-0.009 (-5.39)

3

4

-0.119 (-2.40)

5 -0.094 (-2.02)

-0.469 (-2.28)

-0.607 (-1.75)

1.021 (1.67)

-0.635 (-3.76)

-0.682 (-3.18)

0.692 (16.14)

0.652 (9.77)

0.0452 (2.35)

0.0551 (2.84)

0.85

0.79

-0.463 (-2.11)

-0.302 (-3.63)

dPropetyCasualitylev

0.183 (1.81)

Control Variables dVIX

0.674 (13.82)

0.694 (13.26)

0.682 (11.75)

0.692 (12.21)

0.690 (13.27)

0.682 (13.31)

dTerm R

0.727 (8.42)

0.704 (7.23)

0.669 (11.50)

0.726 (9.80)

0.74

0.80

0.75

0.76

2

0.77

0.77

0.75

0.76

0.74

41

0.76

Table 7: The Market Liquidity and Aggregate Leverages-Growth The table shows the predictive regression results of Y   t   t Yt  i   t Z t  j   t  t   t , where Y , the dependent variable is a scalar and represents one of the t t market liquidity measures; Z is the vector of explanatory variables dBrokerDealer, dCommercalBank, dCreditUnios, dSavingInstitutes, dLifeInsurance, t dPropertyCasuality, which are the value-weighted growth rate of the leverages of Securities broker-dealers, Commercial Banks, Credit Unions, Savings Institutions, Life Insurance companies, and Property and Causality Insurance Companies, respectively;  is the vector of control variables dVIX and dTerm, where dVIX is t

the log differenced CBOE volatility index and dTerm is the log differenced term-spread. When VIX data is not available, VXO, an alternate measure of CBOE volatility index is used. The table shows the dependent variable dAAA, which is the log difference of the market liquidity measure AAA, in terms of the explanatory and control variables. The subscript „i‟ changes from 1 to 5. The subscript „j‟ changes from 0 to 5. Sample period is from Jan1986 to Dec 2009; sub-samples are from Jan 1998 to Dec 2009. T-statistics are in the parenthesis, all standard errors are Newey-West adjusted, and only statistically significant coefficients of Z and  are reported. t

t

Market Liquidity Measure

Y

t

is dyRoll_AAA

Sample: 1986-2009

Sample: 1998-2009

Lags Explanatory variables

0

dBrokerDealer

Lags 1

2

-0.176

-0.277

(-1.62)

(-1.72)

3

4

dCommercialBank

5

0

0.367

0.981

(2.09)

(2.46)

1

2

3

4

1.018

1.323

-0.948

(3.21)

(2.26)

(-1.70)

5

dCreditUnion dSavingInstitiues

0.033

0.933

(4.10)

(1.69)

dLifeInsurance dPropetyCasuality Control Variables dVIX

0.305 (1.94)

dTerm

R

2

0.330

0.297

(2.03)

(2.00)

0.236

0.191

0.204

0.240

0.223

0.221

0.323

0.276

0.341

0.265

0.285

0.324

(3.91)

(3.54)

(3.31)

(4.46)

(4.27)

(3.68)

(4.64)

(3.67)

(4.22)

(4.13)

(4.84)

(4.82)

0.28

0.29

0.35

0.30

0.29

0.30

0.48

0.42

0.47

0.48

0.48

0.40

42