Work from Home Suitability and Credit Risk Assessment

Figures & data

Table 1. Summary statistics for credit rating sample.

Panel A: Descriptive Statistics
Variable	Observations	Mean	SD	P25	P50	P75
RATING	6,237	12.501	3.204	10.000	12.583	15.000
WFH	6,237	0.588	0.251	0.410	0.603	0.786
WFH_IND_ADJ	6,237	0.001	0.205	−0.124	0.011	0.135
HIGHWFH	6,237	0.128	0.334	0.000	0.000	0.000
Firm level characteristics
LOGSIZE	6,237	8.834	1.385	7.845	8.712	9.786
LEV	6,237	0.360	0.180	0.234	0.336	0.458
ROA	6,237	0.046	0.066	0.019	0.044	0.077
LOSS	6,237	0.143	0.351	0.000	0.000	0.000
ZSCORE	6,237	1.505	1.231	0.741	1.432	2.197
SUBORD	6,237	0.137	0.344	0.000	0.000	0.000
CAP_INTEN	6,237	0.564	0.396	0.234	0.468	0.842
INT_COV	6,237	14.008	40.777	4.599	7.800	13.797
BTM	6,237	0.421	0.393	0.189	0.362	0.575
CFVOL	6,237	0.028	0.026	0.013	0.022	0.036
CUMRET	6,237	0.147	0.349	−0.064	0.129	0.326
Panel B: Correlation matrix
	RATING	WFH	LOGSIZE	LEV	ROA	LOSS	ZSCORE	SUBORD	CAP_INTEN	INT_COV	BTM	CFVOL	CUMRET
RATING	1.000
WFH	0.100	1.000
LOGSIZE	0.621	0.060	1.000	n
LEV	−0.380	−0.113	−0.138	1.000
ROA	0.368	0.017	0.082	−0.181	1.000
LOSS	−0.327	−0.009	−0.122	0.180	−0.644	1.000
ZSCORE	0.331	−0.129	−0.007	−0.380	0.503	−0.335	1.000
SUBORD	0.021	0.003	0.084	0.096	−0.110	−0.005	−0.180	1.000
CAP_INTEN	−0.046	−0.273	0.006	0.104	−0.134	0.110	−0.135	0.206	1.000
INT_COV	0.183	0.050	0.056	−0.256	0.187	−0.089	0.185	−0.061	−0.060	1.000
BTM	−0.150	−0.022	−0.040	−0.272	−0.280	0.153	−0.080	0.103	0.171	−0.050	1.000
CFVOL	−0.213	−0.021	−0.220	0.029	0.014	0.112	0.022	−0.124	−0.006	0.012	−0.027	1.000
CUMRET	0.016	0.036	−0.007	−0.031	0.174	−0.143	0.068	0.006	−0.090	0.017	−0.277	−0.005	1.000

The table reports the descriptive statistics (Panel A) and correlation matrix (Panel B) for the credit rating sample covering 6,237 firm–year observations from 2010 to 2020. RATING is the S&P credit rating measure from Compustat and Capital IQ databases. WFH and WFH_IND_ADJ are the firm-level working from home suitability scores and the industry-adjusted WFH score, respectively. HIGHWFH is a dummy variable that takes the value of 1 if a firm’s WFH suitability scores fall into the top quintile of the sample distribution and zero otherwise. Appendix A provides detailed descriptions of the variables.

Figure 1. Path analysis.

Table 2. Validation of firm-level WFH suitability.

Panel A: WFH suitability and firm determinants
	Dependent variable: WFH Suitability
	Predicted signs	coeff.	t.stat.	coeff.	t.stat.
LOGSIZE	?	0.002	(0.87)	−0.002	(−0.60)
LEV	?	−0.022	(−0.83)	−0.026	(−1.08)
BTM	?	0.001	(1.43)	0.001	(1.24)
INVENTORY	?	−0.037*	(−1.84)	−0.037	(−1.57)
ROA	?	−0.036	(−0.80)	−0.060	(−1.37)
RETVOL	?	0.290	(0.60)	−0.099	(−0.21)
IND_TELE	+	0.259***	(5.32)	0.197***	(3.57)
METRO_TELE	+	0.030**	(2.40)	0.024*	(1.89)
FSALES	+	0.021**	(2.22)	0.021**	(2.07)
HIGHTECH	+	0.059***	(3.59)	0.042***	(2.60)
ROUTINE	–			−0.238***	(−2.88)
LABOR_INTENSE	–			−0.052***	(−6.76)
Industry fixed effects		Yes		Yes
Year fixed effects		Yes		Yes
Adjusted R-squared		0.327		0.354
Observations		15,127		11,947
Panel B: WFH Suitability surrounding COVID-19 pandemic
	Pre-COVID-19	COVID-19	Difference	Difference
	(1)	(2)	(2) - (1)	(t-stat)
WFH Suitability	0.600	0.615	0.015**	(2.02)

Panel A reports the results when we regress scores of work-from-home suitability on firm fundamentals. The independent variable is a firm-level working-from-home suitability score (WFH) in Bai et al. (2020). Firm determinants include firm size (LOGSIZE), book-to-market ratio (BTM), inventory-to-sale ratio (INVENTORY), return-on-assets (ROA), return volatility (RETVOL), an indicator indicating where a firm’s foreign sale ratio is above the sample median (FSALES), an indicator indicating where a firm belongs to high tech industries (HIGHTECH), the industry-level share of jobs (IND_TELE) that can be done from home from Dingel and Neiman (2020), an indicator (METRO_TELE) indicating top 5 metropolitan areas with the highest share of jobs that can be done at home from Dingel and Neiman (2020), labor intensity (LABOR_INTENSITY) defined as the proportion of the number of employees over total sales, and a firm’s share of routine-task labor (ROUTINE) from Zhang (2019). The t-statistics are computed using standard errors robust to heteroscedasticity and clustered simultaneously on the firm and year dimensions. Panel B compares the average of WFH suitability before (2018-2019 period) and during Covid-19 (first quarter of 2020). *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

Table 3. The effect of work-from-home suitability on credit ratings: Univariate analysis.

	RATING
WFH Suitability	WFH	WFH_IND_ADJ
1 (Low)	11.5354	11.5511
2	11.9395	12.2754
3	12.4009	12.1303
4	12.7551	12.6354
5	12.7674	12.8369
6	12.8247	12.5780
7	12.8561	13.0744
8	12.6620	12.7684
9	12.7092	12.5556
10 (High)	12.5625	12.6077
10 - 1	1.0271***	1.0566***
t-statistic	(5.46)	(5.66)

The table reports the distribution of credit rating by work-from-home suitability decile groups. RATING is the S&P credit rating measure from Compustat and Capital IQ databases. WFH and WFH_IND_ADJ are the firm-level working from home suitability scores and the industry-adjusted WFH score, respectively. In each year t, we sort stocks into decile portfolios based on their WFH (WFH_IND_ADJ) scores and then compute the portfolio credit rating by taking the average of the ratings across all stocks in the portfolio. The 10–1 measure row reports the average rating difference between the low and high WFH suitability stocks. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

Table 4. The effect of work-from-home suitability on credit ratings: Regression analysis.

	Dependent variable: RATING
	OLS	OLS	Ordered Logit
	(1)	(2)	(3)
HIGHWFH	0.3803***	0.1024**	0.3926***
	(2.70)	(2.05)	(5.23)
LOGSIZE	1.2037***	0.8358***	1.2471***
	(18.02)	(8.27)	(53.02)
LEV	−3.6854***	−2.0518***	−3.7638***
	(−7.81)	(−6.21)	(−21.06)
ROA	5.4000***	0.2811	5.1691***
	(2.65)	(0.48)	(8.53)
LOSS	−0.3513**	−0.1749**	−0.2291**
	(−2.21)	(−2.31)	(−2.57)
ZSCORE	0.5459***	0.2130	0.7669***
	(3.26)	(1.41)	(22.15)
SUBORD	−0.5031**	−0.1181	−0.4355***
	(−2.22)	(−0.93)	(−4.96)
CAP_INTEN	0.5974***	1.4519***	0.5527***
	(2.81)	(4.68)	(6.75)
INT_COV	0.0019	−0.0000	0.0015**
	(1.12)	(−0.01)	(2.09)
BTM	−1.1249***	−0.4048***	−1.1423***
	(−5.94)	(−3.25)	(−14.86)
CFVOL	−7.4445***	−2.9657**	−9.3809***
	(−3.43)	(−2.57)	(−9.07)
CUMRET	−0.7324***	−0.4179***	−0.7082***
	(−4.42)	(−4.18)	(−9.41)
Year fixed effects	Yes	Yes	Yes
Industry fixed effects	Yes	No	Yes
Firm fixed effects	No	Yes	No
R-squared/Pseudo R-squared	0.6752	0.9404	0.1751
Observations	6,237	6,237	6,237

The table reports the results of the regression of work-from-home suitability on credit risk assessment. The dependent variables of interest are the S&P credit rating measures (RATING) from Compustat and Capital IQ databases. The independent variable of interest (HIGHWFH) is a dummy variable that takes the value of 1 if a firm’s WFH suitability scores fall into the top quintile of the sample distribution and zero otherwise. The t-statistics (z-statistics) are computed using standard errors robust to heteroscedasticity and clustered simultaneously on the firm and year dimensions (Gow et al., 2010; Petersen, 2009). *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively. Appendix A provides detailed descriptions of the variables.

Table 5. Propensity-score-matched sample analysis.

Panel A: Summary statistics for the matched sample
	Treat	Control	Difference	p-value
	(1)	(2)	(1) – (2)
RATING	12.755	12.199	0.556***	(0.002)
LOGSIZE	8.788	8.808	−0.020	(0.780)
LEVERAGE	0.346	0.341	0.005	(0.569)
ROA	0.053	0.053	0.000	(0.969)
LOSS	0.134	0.159	−0.025	(0.171)
ZSCORE	1.304	1.296	0.008	(0.914)
SUBORD	0.091	0.094	−0.003	(0.860)
CAP_INTEN	0.392	0.406	−0.013	(0.490)
INT_COV	17.960	18.252	−0.292	(0.911)
BTM	0.365	0.373	−0.008	(0.685)
CFVOL	0.030	0.030	0.000	(0.956)
CUMRET	0.178	0.186	−0.008	(0.658)
Panel B: Regression analysis based on the matched sample
	Dependent Variable: RATING
	Model (1)	Model (2)
	OLS	Ordered Logit
HIGHWFH	0.5474***	0.5110***
	(3.21)	(5.00)
LOGSIZE	1.3471***	1.3931***
	(14.14)	(27.99)
LEV	−3.6828***	−3.7512***
	(−4.95)	(−10.45)
ROA	6.0576***	6.5240***
	(2.95)	(5.81)
LOSS	−0.3981	−0.3185*
	(−1.63)	(−1.75)
ZSCORE	0.2774*	0.5152***
	(1.71)	(8.01)
SUBORD	−0.5939	−0.6387***
	(−1.60)	(−3.14)
CAP_INTEN	0.6694**	0.5316***
	(2.03)	(2.93)
INT_COV	0.0021	0.0013
	(0.87)	(1.18)
BTM	−1.4652***	−1.4672***
	(−4.83)	(−9.19)
CFVOL	−3.7616	−6.1821***
	(−1.35)	(−3.13)
CUMRET	−0.5710***	−0.7105***
	(−4.38)	(−4.56)
Year fixed effects	Yes	Yes
Industry fixed effects	Yes	Yes
R-squared/Pseudo R-squared	0.6980	0.1951
Observations	1,534	1,534

The table presents the results of the regression of work-from-home suitability on credit risk assessment for a propensity-score-matched sample over the period from 2010 to 2020. The treatment (control) group includes firms with (without) WFH suitability scores that fall into the top quintile of the sample distribution. We present the post-match diagnostic analysis in Panel A. Panel B reports the results for the regression analysis conducted on the matched sample. The dependent variables of interest are the S&P credit rating measures (RATING) from Compustat and Capital IQ databases. The independent variable of interest (HIGHWFH) is a dummy variable that takes the value of 1 if a firm’s WFH suitability scores fall into the top quintile of the sample distribution and zero otherwise. The t-statistics (z-statistics) are computed using standard errors robust to heteroscedasticity and clustered simultaneously on the firm and year dimensions (Gow et al., 2010; Petersen, 2009). *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively. Appendix A provides detailed descriptions of the variables.

Table 6. Path analysis.

	Path = TFP		Path = Cashflow Volatility
	(1)	(2)	(3)	(4)
	Coeff.	p-Value	Coeff.	p-Value
Direct Path
P (WFH, RATING)	0.0390***	(0.00)	0.0390***	(0.00)
Indirect Path
P (WFH, Path)	0.2020***	(0.00)	−0.0765***	(0.00)
P (Path, RATING)	0.0866***	(0.00)	−0.0347***	(0.00)
P (WFH, Path) × P (Path, RATING)	0.0175***	(0.00)	0.0027***	(0.00)
Total effect	0.0565***	(0.00)	0.00417***	(0.00)
Mediated % in Total	30.97%		6.37%
Observations	4,348		4,348

The table reports the results from a path analysis that examines the effect of work-from-home suitability on credit ratings through ex ante firm productivity and cash flow volatility. We estimate the following models in the path analysis: (a) $\begin{array}{rl}RATING& ={\text{β}}_0+{\text{β}}_1\times WFH+\text{β}2\times TFP+\text{β}3\times CashflowVolatility+Controls+{ϵ}_t\end{array}$(a) (b) $\begin{array}{rl}TFP& ={\alpha }_0+{\alpha }_1\times WFH+{ϵ}_t\end{array}$(b) (c) $\begin{array}{rl}CashflowVolatility& ={\mu }_0+{\mu }_1\times WFH+{ϵ}_t\end{array}$(c) The independent variable of interest is the firm-level work-from-home suitability score (WFH); RATING refers to a credit rating measure from Compustat and Capital IQ databases, as discussed in Model (1). To capture Firm Productivity, we employ the firm-level total factor productivity measure (denoted TFP) as proposed in İmrohoroğlu and Tüzel (2014). To capture Cash flow Volatility, we use the volatility of future cash flows, measured as the standard deviation of operating cash flows scaled by total assets from year t + 1 to year t + 4. The term Controls represents the relevant control variables from the baseline regression in Model (1). The path coefficient β₁ is the magnitude of the direct path from work-from-home suitability to credit ratings. The path coefficient β₂ (β₃) is the magnitude of the path from firm productivity (cashflow volatility) to credit ratings. The path coefficient ${\hat{\alpha }}_1\times \widehat{{\text{β}}_2}({\hat{\mu }}_1\times \widehat{{\text{β}}_3})$ is the magnitude of the indirect path from work-from-home suitability to credit ratings mediated through firm productivity (cashflow volatility). The significance of the indirect effect is estimated using Sobel (1982) test statistics. The table reports the path coefficients of interest. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively

Table 7. Cross-sectional analysis.

	Dependent variable: RATING
	Past Variability		Grade		Competition		High-tech and others
	High	Low	Speculative	Investment	High	Low	High	Others
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
HIGHWFH	0.2008*	−0.0205	0.2504***	0.0103	0.7637***	0.2759	0.0024	0.3878***
	(1.88)	(−0.27)	(3.21)	(0.11)	(4.07)	(1.42)	(0.01)	(2.87)
LOGSIZE	0.6150***	1.1683***	0.9737***	0.4445***	1.2744***	1.0523***	1.4115***	1.1660***
	(3.68)	(3.28)	(45.22)	(15.03)	(13.72)	(9.58)	(12.86)	(15.84)
LEV	−1.9437***	−1.9328**	−3.4679***	−0.9567***	−3.9996***	−3.6803***	−3.4439***	−3.9050***
	(−3.96)	(−2.41)	(−16.89)	(−4.86)	(−5.73)	(−4.27)	(−3.85)	(−7.77)
ROA	2.5224**	−1.5268	5.1340***	0.2189	7.5738**	−3.8671***	−2.7794	6.4660***
	(3.03)	(−1.14)	(8.25)	(0.36)	(2.41)	(−3.49)	(−1.08)	(3.09)
LOSS	−0.1362	−0.2147	0.2715**	−0.6924***	0.2435	−0.8516***	−0.1972	−0.3907**
	(−1.41)	(−1.42)	(2.28)	(−7.62)	(0.76)	(−3.62)	(−0.69)	(−2.22)
ZSCORE	−0.1615	0.6083***	0.2626***	0.3135***	0.2547	0.8671***	1.0632***	0.4899***
	(−1.05)	(4.10)	(9.03)	(8.97)	(1.35)	(6.78)	(7.35)	(2.86)
SUBORD	0.0441	−0.6054**	−0.2301**	0.0949	−0.8008**	0.1638	−0.9599*	−0.4377*
	(0.19)	(−2.80)	(−2.00)	(1.03)	(−2.01)	(0.49)	(−1.73)	(−1.77)
CAP_INTEN	0.9740*	2.8970***	1.0194***	0.0391	1.0846***	0.3384	−0.4004	0.7751***
	(1.99)	(3.21)	(10.44)	(0.44)	(2.62)	(1.02)	(−0.74)	(3.47)
INT_COV	−0.0028***	0.0075***	0.0022***	−0.0006	0.0001	−0.0003	0.0036	0.0001
	(−3.63)	(3.66)	(4.36)	(−0.54)	(0.13)	(−0.25)	(1.31)	(0.07)
BTM	−0.5661***	−0.2661	−2.0857***	−0.1565**	−1.9916***	−0.4570**	−1.7260***	−1.1202***
	(−3.45)	(−1.35)	(−17.67)	(−2.22)	(−5.17)	(−2.13)	(−3.72)	(−5.70)
CFVOL	−2.9865	−8.0422	−4.2898***	−4.0047***	−19.6685***	−3.2253	−6.1565	−7.2233***
	(−1.52)	(−0.81)	(−3.65)	(−3.90)	(−4.77)	(−0.97)	(−1.14)	(−2.91)
CUMRET	−0.4931***	−0.5056***	−0.7187***	−0.3240***	−0.9516***	−0.3665**	−0.7403***	−0.7117***
	(−3.79)	(−4.13)	(−6.83)	(−4.14)	(−4.33)	(−2.58)	(−3.63)	(−3.72)
Year fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Industry fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Adjusted R-squared	0.939	0.962	0.566	0.269	0.739	0.684	0.782	0.663
Observations	1,445	1,418	3,624	2,613	1,558	1,559	956	5,281
SUR & Wald Test for differences in coefficients:
χ^2 Test	2.92* (0.08)		3.07* (0.07)		7.37*** (0.01)		7.71*** (0.01)
p-value	(0.08)		(0.07)		(0.01)		(0.01)

The table reports the results from the regressions of S&P credit ratings on work-from-home suitability for subsamples of firms. The dependent variables of interest are the S&P credit rating measures (RATING) from Compustat and Capital IQ databases. The independent variable of interest (HIGHWFH) is a dummy variable that takes the value of 1 if a firm’s WFH suitability scores fall into the top quintile of the sample distribution and zero otherwise. Columns (1)-(2) report the results for high and low levels of past variability based on the top (bottom) quartile of the cash flow volatility, respectively. Columns (3)-(4) report results separately for subsamples of firms with investment-grade ratings and speculative-grade ratings, respectively. Columns (5)-(6) report the results for high and low levels of market competition based on the top (bottom) quartile of Lerner index (Lerner, 1934), respectively. Columns (7)-(8) report the results for high tech and other industries, respectively. We follow Decker et al. (2017) and define high-tech industries using their 4-digit NAICS levels, including computer and communications equipment manufacturing (NAICS 3341, 3342); semiconductor, navigational, and other electronic component manufacturing (NAICS 3344, 3345); pharmaceutical and medicine manufacturing (NAICS 3254); aerospace product (NAICS 3364); software publishers, internet and broadcasting (NAICS 5112, 5161); other telecommunications (NAICS 5179); internet service provider and data processing service (NAICS 5181, 5182); architectural, engineering and related services (NAICS 5413); computer systems design (NAICS 5415); and scientific research and development (NAICS 5417). The t-statistics are computed using standard errors robust to heteroscedasticity and clustered simultaneously on the firm and year dimensions. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

Table 8. Work-from-home suitability and bond yield spread.

	Dependent variable: LnSPREAD
	(1)	(2)	(3)	(4)
WFH	−0.1311**	−0.0875***
	(−2.54)	(−2.78)
WFH_IND_ADJ			−0.1303**	−0.0828**
			(−2.50)	(−2.40)
LnOFFER		0.1039***		0.1040***
		(4.13)		(4.14)
MATURITY		0.0174***		0.0174***
		(9.42)		(9.43)
LOGSIZE		−0.1269***		−0.1270***
		(−6.68)		(−6.71)
LEV		0.3452**		0.3455**
		(1.96)		(1.97)
ROA		0.0860		0.0836
		(0.21)		(0.20)
LOSS		0.0680*		0.0669*
		(1.67)		(1.65)
ZSCORE		−0.0633***		−0.0634***
		(−3.00)		(−3.01)
SUBORD		−0.0431		−0.0432
		(−0.68)		(−0.68)
CAP_INTEN		0.0937*		0.0948*
		(1.74)		(1.76)
INT_COV		−0.0007***		−0.0007***
		(−4.31)		(−4.28)
BTM		0.3110***		0.3117***
		(3.55)		(3.56)
CFVOL		1.9266***		1.9202***
		(5.00)		(4.96)
CUMRET		0.0335		0.0332
		(0.46)		(0.45)
Bond rating fixed effects	Yes	Yes	Yes	Yes
Year fixed effects	Yes	Yes	Yes	Yes
Industry fixed effects	Yes	Yes	Yes	Yes
Adjusted R-squared	0.6374	0.7358	0.6373	0.7357
Observations	1,501	1,501	1,501	1,501

The table presents the results of the regression of work-from-home suitability on new bonds’ offering credit spreads. WFH and WFH_IND_ADJ are the firm-level working from home suitability scores and the industry-adjusted WFH score, respectively. The dependent variable of interest is the offering credit spreads, measured as the logarithm of the offering yield to maturity in excess of similar duration treasuries (LnSPREAD). The t-statistics are computed using standard errors robust to heteroscedasticity and clustered simultaneously on the firm and year dimensions. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

Table 9. The impact of COVID-19 on corporate default risk.

Panel A: Change in Corporate Default Risk from 2019 to 2020
	2019 (Quarters 3 and 4)	2020(Quarters 1 and 2)	Diff (p-value)
EDF	0.0519	0.0712	<0.001
Panel B: Parallel Trend Tests
	Dependent variable: Default risk
	EDF	EDF_IND_ADJ
	(1)	(2)
High_Suitability × TimeTrend	−0.0009	−0.0013
	(−0.59)	(−0.86)
TimeTrend	0.0009	0.0020
	(0.96)	(1.56)
Controls	Yes	Yes
Industry fixed effects	Yes	Yes
Adjusted R-squared	0.3087	0.2681
Observations	4,000	4,000
Panel C: Difference-in-differences regression analysis
	EDF	EDF	EDF	EDF	EDF	EDF_IND_ADJ	EDF_IND_ADJ	EDF_IND_ADJ	EDF_IND_ADJ
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
High_Suitability × COVID19	−0.0170**	−0.019***	−0.018**	−0.017***	−0.019***	−0.016***	−0.013*	−0.012***	−0.013***
	(−2.16)	(−7.64)	(−2.46)	(−6.46)	(−9.53)	(−10.33)	(−1.87)	(−4.35)	(−5.76)
COVID19				0.034***	0.031***			0.003	0.000
				(16.38)	(24.76)			(1.33)	(0.12)
High_Suitability				−0.011**	−0.000			−0.014***	−0.003
				(−2.21)	(−0.01)			(−2.94)	(−0.78)
LOGSIZE		−0.005***	−0.071**		−0.005***	−0.006***	−0.052*		−0.006***
		(−2.98)	(−2.32)		(−2.98)	(−3.27)	(−1.86)		(−3.28)
LEV		0.157***	0.207**		0.157***	0.155***	0.178**		0.155***
		(4.89)	(2.57)		(4.87)	(4.86)	(2.51)		(4.85)
BTM		−0.006***	−0.015***		−0.006***	−0.006***	−0.010*		−0.006***
		(−3.25)	(−2.97)		(−3.12)	(−4.02)	(−1.88)		(−3.79)
ASSET_LIQUIDITY		−0.323***	−0.282**		−0.323***	−0.266***	−0.136		−0.266***
		(−4.15)	(−2.09)		(−4.14)	(−5.06)	(−1.03)		(−5.11)
INCOME/ASSET		−0.002	0.007*		−0.002	−0.002	0.002		−0.002
		(−0.90)	(1.77)		(−0.91)	(−0.95)	(0.57)		(−0.97)
1/SIGMA		−0.020***	−0.005		−0.020***	−0.019***	−0.000		−0.019***
		(−11.68)	(−1.49)		(−11.78)	(−11.64)	(−0.09)		(−11.70)
EXRET		−0.293***	−0.247***		−0.293***	−0.290***	−0.237***		−0.290***
		(−15.57)	(−7.16)		(−15.88)	(−13.61)	(−7.29)		(−13.88)
Industry fixed effects	No	Yes	No	Yes	Yes	Yes	No	Yes	Yes
Firm fixed effects	Yes	No	Yes	No	No	No	Yes	No	No
Quarter fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Adjusted R-squared	0.529	0.361	0.624	0.101	0.361	0.298	0.603	0.028	0.298
Observations	2,461	2,461	2,461	2,461	2,461	2,461	2,461	2,461	2,461

Panel A reports the expected default frequency in the last two quarters of 2019 and the first two quarters of 2020. The p-value is for the t-test for the means. Panel B reports the results for the parallel trend test. EDF is expected default frequency measure, following Bharath and Shumway (2008) and Brogaard et al. (2017). EDF_IND_ADJ is the industry adjusted EDF measure. TimeTrend is a time trend variable that captures four quarters in 2019. Panel C reports the difference-in-difference analysis. High_Suitability is a dummy variable that takes the value of 1 if a firm’s average WFH suitability score during the pre-COVID-19 period (2010-2019) belongs to the top quartile of the sample distribution and zero otherwise. COVID19 is a dummy variable that takes the value of 1 for the first two quarters of 2020 and zero for the last two quarters of 2019. LOGSIZE is the natural logarithm of firm size measured by total assets; LEV is the leverage ratio measured as the ratio of total debt to book assets; BTM is the book to market ratio; INCOME/ASSET is the ratio of net income to total asset; 1/SIGMA is the inverse of annualized stock return volatility; ASSET_LIQUIDITY is the ratio of current assets over current liabilities; EXRET is the annual excess return. The t-statistics are computed using standard errors robust to heteroscedasticity and clustered simultaneously on the firm and quarter dimensions. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

Table 10. Work-from-home suitability, firm-level COVID exposure, and default risk.

	Dependent variable: Default risk
	Firm-level COVID exposure		Firm-level COVID exposure
	High	Low	High	Low
	EDF	EDF	EDF_IND_ADJ	EDF_IND_ADJ
	(1)	(2)	(3)	(4)
High_Suitability × COVID19	−0.0260**	−0.0124	−0.0257**	−0.0051
	(−2.35)	(−1.43)	(−2.43)	(−0.64)
LOGSIZE	−0.1176***	−0.0404	−0.0968***	−0.0256
	(−3.27)	(−0.79)	(−2.82)	(−0.56)
LEV	0.3743***	0.0804	0.3104***	0.0847
	(3.90)	(0.87)	(3.27)	(1.05)
BTM	−0.0131**	−0.0107	−0.0086	−0.0062
	(−1.96)	(−1.59)	(−1.17)	(−0.93)
ASSET_LIQUIDITY	−0.3131	−0.2610*	−0.2201	−0.1522
	(−1.16)	(−1.90)	(−0.83)	(−1.29)
INCOME/ASSET	0.0067	0.0068	0.0043	0.0019
	(1.23)	(1.24)	(0.73)	(0.36)
1/SIGMA	−0.0018	−0.0093***	0.0010	−0.0042
	(−0.37)	(−2.60)	(0.22)	(−1.24)
EXCESS_RETURN	−0.2202***	−0.2351***	−0.2065***	−0.2149***
	(−5.77)	(−5.05)	(−5.62)	(−5.28)
Firm fixed effects	Yes	Yes	Yes	Yes
Quarter fixed effects	Yes	Yes	Yes	Yes
Adjusted R-squared	0.6650	0.5970	0.5888	0.6041
Observations	1,006	1,216	962	644
SUR & Wald Test for differences in coefficients
differences	(1)–(2)		(3)–(4)
χ^2 Test	3.79**		3.18*
p-value	(0.05)		(0.07)

The table reports the results of the impact of the COVID-19 pandemic on corporate default risk across two subsamples: high and low COVID-exposure firms. We employ the firm-level exposure to the COVID-19 pandemic, measured by the frequency of keywords related to COVID-19 extracted from the earnings call transcripts, from Hassan et al. (2020). We define high (low) COVID-exposure firms as those above (below) the sample mean of the COVID-19 risk exposure over the first two quarters of 2020. We estimate a difference-in-differences regression model as in Table 9 for two subsamples. EDF is expected default frequency measure, following Bharath and Shumway (2008) and Brogaard et al. (2017). EDF_IND_ADJ is the industry adjusted EDF measure. High_Suitability is a dummy variable that takes the value of 1 if a firm’s average WFH suitability score during the pre-COVID-19 period (2010-2019) belongs to the top quartile of the sample distribution and zero otherwise. COVID19 is a dummy variable that takes the value of 1 for the first two quarters of 2020 and zero for the last two quarters of 2019. LOGSIZE is the natural logarithm of firm size measured by total assets; LEV is the leverage ratio measured as the ratio of total debt to book assets; BTM is the book to market ratio; INCOME/ASSET is the ratio of net income to total asset; 1/SIGMA is the inverse of annualized stock return volatility; ASSET_LIQUIDITY is the ratio of current assets over current liabilities; EXRET is the annual excess return. The t-statistics are computed using standard errors robust to heteroscedasticity and clustered simultaneously on the firm and quarter dimensions. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

Table

Variables	Descriptions
Credit rating measures
RATING	S&P’s Long-Term Domestic Issuer Credit Rating from Compustat and Capital IQ databases.
Work from home measures
WFH	A firm’s working-from-home suitability scores in Bai et al. (2020).
WFH_IND_ADJ	Industry adjusted WFH score, measured as the difference between each firm WFH score and the industry average value.
HIGHWFH	A dummy variable that takes the value of 1 if a firm’s WFH suitability scores fall into the top quintile of the sample distribution and zero otherwise
Firm characteristics
LOGSIZE	The logarithm of firm size measured by total assets.
LEV	Leverage ratio, measured as the ratio of total debt to book assets.
ROA	Net income before extraordinary items, divided by total assets.
LOSS	An indicator that takes the value of 1 if the net income before extraordinary items is negative in the current and prior fiscal year, and zero otherwise.
ZSCORE	Altman (1968)’s z-score.
SUBORD	An indicator that is equal to one if a firm has subordinated debt, and zero otherwise.
CAP_INTEN	Gross property, plant, and equipment, divided by total assets.
INT_COV	Operating income before depreciation, divided by interest expenses.
BTM	Book to market, measured as the ratio of the book value of equity over the market capitalization value.
CFVOL	Standard deviation of operating cash flows scaled by total assets from year t-4 to year t-1.
CUMRET	Cumulative monthly stock returns in a given year.
Bond data
LnSPREAD	Offering credit spreads in logarithm form.
LnOFFER	Natural logarithm of the offering amount of the new bond.
MATURITY	The number of years until the maturity of the new bond.
Other variables
TFP	The firm-level total factor productivity measure from İmrohoroğlu and Tüzel (2014).
DTD	Merton (1974)’s distance-to-default measure.
G-INDEX	Corporate governance index of Gompers et al. (2003).
LN_ANALYST	Natural logarithm of the number of analysts issuing an annual forecast for a firm during the year from I/B/E/S data.
INSTO	Institutional ownership measure from Thomson Reuters Institutional Holdings (13F) Database.
INVENTORY	The inventory-to-sale ratio.
LABOR_INTENSE	Labor intensity, measured as the proportion of the number of employees over total sales.
ROUTINE	A firm’s share of routine-task labor from Zhang (2019).
CSR	CSR rating, measured following Cronqvist and Yu (2017).
RETVOL	Stock return volatility, measured as the standard deviation of the monthly stock return over a year.
IND_TELE	the industry-level share of jobs that can be done from home from Dingel and Neiman (2020)
METRO_TELE	An indicator indicating top 5 metropolitan areas with the highest share of jobs that can be done at home from Dingel and Neiman (2020).
ROLLOVER	A measure of rollover risk, estimated following Gopalan et al. (2014).
C-SCORE	Khan and Watts (2009) C-SCORE to capture a firm-level accounting conservatism
EDF	Expected default frequency measure, following Brogaard et al. (2017) and Bharath and Shumway (2008)
ED_IND_ADJ	The industry-adjusted expected default frequency measure, measured as the difference between each firm EDF measure and the industry average value.

Table

Credit Risk	S&P Rating	Numerical Rating
Highest quality	AAA	22
High quality	AA+, AA, AA-	21, 20, 19
Upper medium	A+, A, A-	18, 17, 16
Medium	BBB+, BBB, BBB-	15, 14, 13
No investment grade	BB+, BB, BB-	12, 11, 10
Speculative lower grade	B+, B, B-	9, 8, 7
Speculative risky	CCC+, CCC, CCC-	6, 5, 4
Speculative poor standing	CCC+, CCC, CCC-	3
Highly vulnerable	C	2
In default	D, SD	1

Appendix B provides a detailed discussion of credit rating classifications.

The figure depicts the direct and indirect path through which work-from-home suitability can affect credit ratings. We estimate the following models in the path analysis: (a) $\begin{array}{rl}RATING& ={\text{β}}_0+{\text{β}}_1\times WFH+\text{β}2\times TFP\\ & +\text{β}3\times CashflowVolatility+Controls+{ϵ}_t\end{array}$(a) (b) $\begin{array}{rl}TFP& ={\alpha }_0+{\alpha }_1\times WFH+{ϵ}_t\end{array}$(b) (c) $\begin{array}{rl}CashflowVolatility& ={\mu }_0+{\mu }_1\times WFH+{ϵ}_t\end{array}$(c) The independent variable of interest is the firm-level work-from-home suitability score (WFH); RATING refers to a credit rating measure from Compustat and Capital IQ databases, as discussed in Model (1). To capture Firm Productivity, we employ the firm-level total factor productivity measure (denoted TFP) as proposed in İmrohoroğlu and Tüzel (2014). To capture Cash flow Volatility, we use the volatility of future cash flows, measured as the standard deviation of operating cash flows scaled by total assets from year t + 1 to year t + 4. The term Controls represents the relevant control variables from the baseline regression in Model (1). The path coefficient β₁ is the magnitude of the direct path from work-from-home suitability to credit ratings. The path coefficient β₂ (β₃) is the magnitude of the path from firm productivity (cashflow volatility) to credit ratings. The path coefficient ${\hat{\alpha }}_1\times \widehat{{\text{β}}_2}({\hat{\mu }}_1\times \widehat{{\text{β}}_3})$ is the magnitude of the indirect path from work-from-home suitability to credit ratings mediated through firm productivity (cashflow volatility).

Bai, J., Brynjolfsson, E., Jin, W., Steffen, S., & Wan, C. (2020). The future of work: Work from home suitability and firm resilience during the COVID-19 pandemic. Working Paper. Massachusetts Institute of Technology. National Bureau of Economic Research (NBER) Working Paper (No. w28588).

 Google Scholar

Dingel, J. I., & Neiman, B. (2020). How many jobs can be done at home?. Journal of Public Economics, 189, 104235. https://doi.org/10.1016/j.jpubeco.2020.104235

 PubMed Web of Science ®Google Scholar

Zhang, M. B. (2019). Labor-technology substitution: Implications for asset pricing. The Journal of Finance, 74(4), 1793–1839. doi:10.1111/jofi.12766

 Web of Science ®Google Scholar

Gow, I. D., Ormazabal, G., & Taylor, D. J. (2010). Correcting for cross-sectional and time-series dependence in accounting research. The Accounting Review, 85(2), 483–512. https://doi.org/10.2308/accr.2010.85.2.483

 Web of Science ®Google Scholar

Petersen, M. A. (2009). Estimating standard errors in finance panel data sets: Comparing approaches. Review of Financial Studies, 22(1), 435–480. https://doi.org/10.1093/rfs/hhn053

 Web of Science ®Google Scholar

İmrohoroğlu, A., & Tüzel, Ş. (2014). Firm-level productivity, risk, and return. Management Science, 60(8), 2073–2090. https://doi.org/10.1287/mnsc.2013.1852

 Web of Science ®Google Scholar

Sobel, M. E. (1982). Asymptotic confidence intervals for indirect effects in structural equation models. Sociological Methodology, 13, 290–312. https://doi.org/10.2307/270723

 Google Scholar

Lerner, A. P. (1934). Economic theory and socialist economy. The Review of Economic Studies, 2(1), 51–61. https://doi.org/10.2307/2967550

 Google Scholar

Decker, R. A., Haltiwanger, J., Jarmin, R. S., & Miranda, J. (2017). Declining dynamism, allocative efficiency, and the productivity slowdown. American Economic Review, 107(5), 322–326. https://doi.org/10.1257/aer.p20171020

 Web of Science ®Google Scholar

Bharath, S. T., & Shumway, T. (2008). Forecasting default with the Merton distance to default model. Review of Financial Studies, 21(3), 1339–1369. https://doi.org/10.1093/rfs/hhn044

 Web of Science ®Google Scholar

Brogaard, J., Li, D., & Xia, Y. (2017). Stock liquidity and default risk. Journal of Financial Economics, 124(3), 486–502. https://doi.org/10.1016/j.jfineco.2017.03.003

 Web of Science ®Google Scholar

Hassan, T. A., Hollander, S., Van Lent, L., Schwedeler, M., & Tahoun, A. (2020). Firm-level exposure to epidemic diseases: COVID-19, SARS, and H1N1. National Bureau of Economic Research (NBER) Working Paper (No. w26971).

 Google Scholar

Altman, E. I. (1968). Financial ratios, discriminant analysis and the prediction of corporate bankruptcy. The Journal of Finance, 23(4), 589–609. https://doi.org/10.1111/j.1540-6261.1968.tb00843.x

 Web of Science ®Google Scholar

Merton, R. C. (1974). On the pricing of corporate debt: The risk structure of interest rates. The Journal of Finance, 29(2), 449–470. https://doi.org/10.2307/2978814

 Web of Science ®Google Scholar

Gompers, P., Ishii, J., & Metrick, A. (2003). Corporate governance and equity prices. The Quarterly Journal of Economics, 118(1), 107–156. https://doi.org/10.1162/00335530360535162

 Web of Science ®Google Scholar

Cronqvist, H., & Yu, F. (2017). Shaped by their daughters: Executives, female socialization, and corporate social responsibility. Journal of Financial Economics, 126(3), 543–562. https://doi.org/10.1016/j.jfineco.2017.09.003

 Web of Science ®Google Scholar

Gopalan, R., Song, F., & Yerramilli, V. (2014). Debt maturity structure and credit quality. Journal of Financial and Quantitative Analysis, 49(4), 817–842. https://doi.org/10.1017/S0022109014000520

 Web of Science ®Google Scholar

Khan, M., & Watts, R. L. (2009). Estimation and empirical properties of a firm-year measure of accounting conservatism. Journal of Accounting and Economics, 48(2-3), 132–150. https://doi.org/10.1016/j.jacceco.2009.08.002

 Web of Science ®Google Scholar

Data Availability

The data are available from the sources identified in the paper.