6.1. Proceso de organización del HCG, (Hogar Comunitario
6.1.1 Paralelo
In this section, I first discuss how each UMP affected the credit risk premia, liquidity risk premia, and “fear” premia components of CDS spreads. Then, I discuss two alternative explanations of how UMP could have increased CDS spreads.
How did UMPs affect CDS spreads?
I use a two-step event-style panel regression to study the changes in CDS spreads during each UMP event. In the first step, I use Equation 1.1 to estimate rolling risk exposures regarding market credit risk, liquidity risk, and the “fear” index for each firm. As shown in Table 1.3 Panel A, the changes in the independent variables are not highly correlated, except the changes in VIX and the 10-year Treasury rate. The -0.35 correlation between changes in VIX and the 10-year Treasury rate is consis- tent with the argument that low risk-free rates are usually associated with economic downturns (Collin-Dufresne et al. 2001; Ericsson et al. 2009). The summary statistics of the estimated covariance risks (βs) are shown in Table 1.3 Panel B. On average, firm CDS spreads positively correlate with market credit risk, liquidity risk, and the “fear” index, but they negatively correlate with 10-year Treasury rates. The average positive relationships are intuitive in that the increase in market credit risk, market illiquidity, and “fear” about future growth inflate default risk and risk aversion, which in turn increase CDS spreads. The negative relationship between CDS spreads and the 10-year Treasury rate is consistent with the literature (Collin-Dufresne et al. 2001; Ericsson et al. 2009). A possible explanation for the inverse relationship is that lower risk-free rates are associated with expanding monetary policies and a weak economic outlook.
Then, in the second step, I use the estimated risk exposures in the first step to study the changes in credit risk premia, liquidity risk premia, and “fear” premia.13
The estimation results are presented in Table 1.4. Because dependent variables take the form of log changes, the coefficients reported in Table 1.4 can be interpreted as the percentage changes in risk premia. QE 1 announcements decreased the credit risk 13Because the negative empirical relationship between 10-year Treasury rate and CDS spreads does not mean that a decrease in 10-year Treasury rates increases CDS spreads (Gilchrist and Zakrajšek 2013), this paper only estimates the “risk-free rate premia” for control purposes without further analysis.
premium by 2-5% and the “fear” premia by 5-13%. The initial QE announcement decreased the liquidity premium by 20% and the interest rate risk premium by ap- proximately 7%. The first two QE 2 announcements that indicated the stimulation but not a plan increased the “fear” premia significantly by more than 10%, while the third QE 2 that specified the size of the second round of QE purchase decreased the “fear” premia. The empirical results on QE 1 and 2 suggest that the UMPs in the crisis reduced credit risk premia, liquidity premia, and “fear” premia. However, the UMPs after the crisis, especially those with unclear statements, may have delivered the information of a weak economic outlook, which resulted in higher credit spreads. The results on OT further confirmed that the UMPs during the recovery period could have persuaded investors that the economy was weaker than they had expected. Hence, the credit spreads increased during the OT announcements. In contrast to QE and OT, the FG that eliminate the uncertainty in FF rate policies for two years decreased credit risk premia and “fear” premia. The results indicate that FG was effective with little cost.14
To estimate the economic significance, this paper further decomposes changes in average CDS spreads to components due to changes in each risk premia during each announcement (Table 1.5). The results show that changes in these risk premia explain a significant fraction of average changes in CDS spreads during those expanding policy announcements (from 31% to 139%, with a mean of 60%). The increase in average CDS spreads during the 2011 FG was abnormal that it could not be explained by changes in risk premia studied in this paper.
14The estimated independent variables (risk exposures-βs) may lead to biased estimations. To deal with this issue, I re-estimate Equation 1.2 usingβs weighted by their standard errors (Table C.4), which does not change the quality of the results. Still, the coefficients in Table 1.4 may be biased toward insignificant because of attenuation errors. However, in most cases, the signs of the insignificant coefficients are in line with the arguments in this paper.
Two alternative explanations
The first alternative explanation of why CDS spreads increased on QE 2 announce- ments is that the size of the QE 2 purchase failed market expectation. This explana- tion is not supported by the results in this paper. In particular, average CDS spreads increased during the first two QE announcements, in which the Fed did not mention the size of the new stimulus. Furthermore, average CDS spreads decreased during the third QE 2 announcement, in which the Fed announced the size and the asset they would purchase. These empirical results contradict the prediction of the first alternative explanation.
The second alternative explanation is that other events may contribute to the findings in this paper. In particular, the Euro Crisis (occurring before the QE 2 announcements) and the US credit downgrade (occurring before the announcement of FG and OT) might have increased CDS spreads on average. In such cases, the changes in “fear premia” found in this paper might have just picked up the legacy of the Euro Crisis and the US credit downgrade, among other events. To mitigate this concern, this paper investigates the “fear” index during a 10-day window surrounding each UMP announcement. As shown in Figure 1.8 (a), all QE 1 announcements, except the Bernanke Speech on 12/1/2008 (QE 1_2), slightly decreased the “fear” index. The most supportive evidence that the changes in “fear premia” found in this paper were caused by UMP announcements comes from QE 2, OT, and FG (Figure 1.8 (b)- (d)). In particular, the VIX was relatively stable within 5 days before each QE 2 and OT announcements; and, in contrast, it sharply changed after these announcements. Moreover, the first FG announcement obviously reversed the increasing trend in the VIX.15 Last, the plots of the S&P 500 index mirror the plots of the VIX and support 15It is not surprising that the second OT announcement and the second FG announcement had minimal effect on VIX because these two announcements did not add much new information. The second OT announcement extended the size of the OT from $400 billion in Treasuries to $600 billion, and the second FG announcement reinforced that the low Fed Funds rate would continue for a long
the argument in this paper that UMP announcements shaped the expectations about future growth and, consequently, affected the “fear” premia in CDS spreads.
Overall, the empirical results support the first two parts of Hypothesis 1 that UMPs could reduce credit spreads by reducing credit risk premia and liquidity risk premia; they also support the third part of Hypothesis 1 that UMPs could increase credit spreads by creating a “fear” premia. The positive effects are more likely to happen in a crisis, and the negative effects tend to dominate in a recovery period.