Sustaina bility 2019 , 11, x doi: FOR PEER REVIEW www.mdp i.comjournal sustainab ility [623014]

Sustaina bility 2019 , 11, x; doi: FOR PEER REVIEW www.mdp i.com/journal/ sustainab ility
Article 1
Analyst Following, Group -Affiliation, and Labor 2
Investment Efficiency: Evidence from Korea 3
Kyoungwon Mo 1 and Kyung Yun (Kailey) Lee2,* 4
1 School of Business Administration, Chung -Ang University; [anonimizat] 5
2 College of Business Administration, Hankuk University of Foreign St udies; [anonimizat] 6
* Correspondence: [anonimizat].k r; Tel.: +82 -2-820-5533 7
Received: date; Accepted: date; Published: date 8
Abstract: This paper studies how analysts’ group affiliation affects firms’ labor investment 9
efficiency. Using a 2001 -2017 sample of Korean public companies, we find that labor investment 10
efficiency increases when there are more unaffiliated analysts following business group (chaebol) 11
firms. An increase in labor investment efficiency is attributed to a reduction in firms’ over -firin g 12
problem. However, affiliated analysts are not found to influence firms’ labor investment efficiency. 13
We further document that the positive influence of unaffiliated analysts on labor investment 14
efficiency holds when firms have high cash holdings. 15
Keywords: analyst following; group -affiliated analysts; Chaebol; labor investment efficiency 16
17
1. Introduction 18
It is well known that ana lysts actively communicate with the senior managers of companies and 19
serve an important role in corporate policy decisions. In a survey of over 300 analysts, more than half 20
answered that they have direct contact with CEOs or CFOs at least five times a year [1]. Analysts 21
mentioned that conversations with senior managers are a very useful source wh en they make stock 22
recommendations or earnings forecasts. On the other hand, Graham, Harvey, and Rajgopal showed 23
that many CFOs view analysts as the most important investors in terms of setting the stock price for 24
their companies [2]. Senior managers repor ted that analysts affect their decisions on corporate 25
policies, as meeting analyst benchmarks is an important consideration for them. Therefore, analysts 26
and corpor ate managers are talking to each other in the hope that they can obtain or deliver valuable 27
information about the firm, which can affect corporate policy as well. 28
Researchers in prior studies have focused on analysts’ governance role. For example, Yu showe d 29
that managers are less likely to manage earnings when there are more analysts following [3 ]. Also, 30
Chen, Harford, and Lin argued that analyst following increases cash holdings and decreases CEOs’ 31
excess compensation, value -destroying acquisitions, and ea rnings management [4]. In practice, 32
analysts not only forecast companies’ earnings but they also voice their opinions on corporate policies 33
such as employment. For instance, when retail industries were cutting labor costs due to higher 34
market competition, analysts predicted the negative effect of low staffing. They specifically said that 35
reducing labor costs could compound sales problems over time [5]. In other cases, analysts can also 36
support companies’ layoff decisions. For example, Cisco Systems Inc. cut thousands of jobs in 2011. 37
When they announced this plan, analysts mentioned that “they wer e pleased to see Cisco taking 38
quick and decisive action on restructuring,” while also mentioning: “We all love the billion dollars in 39
cost savings, but you never ch eer people losing their jobs” [6]. 40
Based on analysts’ influence on firms’ employment polici es, in this study, we investigate 41
whether analysts’ governance role affects companies’ labor investment efficiency. Specifically, the 42
purpose of this study is to ex amine whether analysts’ group affiliation affects the labor investment 43
efficiency of the com panies they follow. There exist two competing theories on whether affiliated 44
analysts will increase labor investment efficiency: information -sharing and conflict -of-interest. If an 45

Sustainabili ty 2019 , 11, x FOR PEER REVIEW 2 of 22
analyst is affiliated with a particular firm, he or she will have a higher chance of contacting senior 46
management through conference calls and thus be able to provide greater monitoring with inside 47
information. If this is the case, compare d to unaffiliated analysts, affiliated analysts are more likely 48
to exert a higher governance role, which can increase the labor investment efficiency of companies 49
within the group. On the other hand, analysts may have a bias towards firms in the same group 50
affiliation. If they do, rather than providing a monitoring effect, they may produce biased earnings 51
forecasts and form public opinion as opinion leaders (conflict -of-interest theory). For instance, 52
Huyghebaert and Xu showed that affiliated analysts issue more positively biased forecasts, target 53
prices, and investment recommendations compared to independent analysts [7]. They noted that 54
affiliated analysts might be genuinely more optimistic about the firms they follow, or they might 55
strategically issue the ir reports due to improper incentives. These two conflicting theories indicate 56
that whether affiliated analysts positively affect labor investment efficiency is an empirical question. 57
We group analysts depending on their affiliation with business groups (c haebols). In Korea, the 58
top ten chaebols own more than 27 percent of all business assets [8] . Given the importance of chaebols 59
in the economy, Korea provides a unique setting where non -financial companies can own financial — 60
securities or insurance —firms, as long as the debt -to-equity ratio of the group is below 200 percent. 61
Prior studies have exam ined the quality of earnings forecasts and stock recommendations by analysts 62
who work for financial firms that belong to one of the chaebols [9]. Although the virtu es of analysts 63
are generally independence and fairness, the situation in Korea is a bit diff erent. In April 2016, the 64
heads of 32 local securities firms gathered to create a joint statement to oppose firms that pressured 65
analysts not to release unfavorable reports about them [10]. These firms often pressure analysts to 66
report earnings estimates o ptimistically by threatening not to give private information to the analysts. 67
By understanding the different institutional environment of analysts, we believe that examining 68
whether analysts’ group affiliation affects firms’ labor investment efficiency in the Korean setting will 69
shed light on the factors that influence corporate policy choices. 70
Overall, we find that analysts’ affiliation with chaebols affects labor e fficiency, supporting the 71
conflict -of-interest hypothesis. Specifically, using data on Korea n public companies from 2001 to 72
2017, we test the effect of analyst following on abnormal net hiring and show that there is a positive 73
association between non -affiliated analysts and labor efficiency in chaebol firms. Further tests 74
indicate that an increas e in labor efficiency comes from preventing companies from firing too many 75
of their employees. We find that this positive effect of non -affiliated analysts on labor efficiency 76
becomes stronger when there is more inside funding. Since analyst following is a ssociated with many 77
factors, a two -stage least squares regression (2SLS) model is used to mitigate endogeneity concerns. 78
While previous studies have shown analysts’ positive influence on firms, this study adds to the 79
literature by showing that not all anal ysts are the same: they may have different influence on firms 80
they follow depending on whether they are independent of the firms. Specifically, we show that 81
analysts’ governance role positively affects firms’ labor investment efficiency only when analysts are 82
unaffiliated. We test this relation using Korean public firms, as Korea provides a unique setting where 83
non-financial firms can own financial firms, and analyst s from these financial firms provide reports 84
on firms in their group (chaebol). Several pape rs have discussed the lower quality of information 85
produced by group -affiliated analysts via their earnings forecasts and stock recommendations [9, 11, 86
12]. Yet, to the best of our knowledge, there is no research regarding analysts’ influence on corporate 87
decisions, other than their role in the stock market. This paper contributes to the literature by 88
confirming that conflict -of-interest theory also applies to the re lationship between analysts’ group 89
affiliation and the labor market. 90
Section 2 outlines the research design, including explanations of the definitions of the key 91
variables. Section 3 presents the empirical results on the relationship between analysts’ grou p 92
affiliation and labor investment efficiency. Section 4 discusses additional tests, and Sec tion 5 outlines 93
the conclusions of the paper. 94
2. Key Variable Definition and Research Design 95

Sustainabili ty 2019 , 11, x FOR PEER REVIEW 3 of 22
2.1.Specification of labor investment efficiency 96
We measure labor investment inefficiency using abnormal net hirin g, which is defined as 97
subtracting expected net hiring from actual net hiring. According to prior studies , the expected 98
change in labor for firm i in year t is estimated using the following regression model with the firms’ 99
economic fundamentals [13-14]: 100
𝑁𝐸𝑇 _𝐻𝐼𝑅𝐸 𝑖,𝑡=𝛼0+𝛼1𝑆𝐴𝐿𝐸𝑆 _𝐺𝑅𝑂𝑊𝑇𝐻 𝑖,𝑡−1+𝛼2𝑆𝐴𝐿𝐸𝑆 _𝐺𝑅𝑂𝑊𝑇𝐻 𝑖,𝑡+𝛼3∆𝑅𝑂𝐴 𝑖,𝑡−1+𝛼4∆𝑅𝑂𝐴 𝑖,𝑡 101
+𝛼5𝑅𝑂𝐴 𝑖,𝑡+𝛼6𝑅𝐸𝑇𝑈𝑅𝑁 𝑖,𝑡+𝛼7𝑆𝐼𝑍𝐸 _𝑅𝑖,𝑡−1+𝛼8𝑄𝑢𝑖𝑐𝑘 𝑖,𝑡−1+𝛼9∆𝑄𝑢𝑖𝑐𝑘 𝑖,𝑡−1 102
+𝛼10∆𝑄𝑢𝑖𝑐𝑘 𝑖,𝑡+𝛼11𝐿𝐸𝑉 𝑖,𝑡−1+𝛼12𝐿𝑂𝑆𝑆𝐵𝐼𝑁 1𝑖,𝑡−1+𝛼13𝐿𝑂𝑆𝑆𝐵𝐼𝑁 2𝑖,𝑡−1 103
+𝛼14𝐿𝑂𝑆𝑆𝐵𝐼𝑁 3𝑖,𝑡−1+𝛼15𝐿𝑂𝑆𝑆𝐵𝐼𝑁 4𝑖,𝑡−1+𝛼16𝐿𝑂𝑆𝑆𝐵𝐼𝑁 5𝑖,𝑡−1+Industry Fixed Effects 104
+𝜀𝑖,𝑡 105
（1） 106
Where 𝑁𝐸𝑇 _𝐻𝐼𝑅𝐸 𝑖,𝑡 is the percentage chang e in number of employees at year t; 107
𝑆𝐴𝐿𝐸𝑆 _𝐺𝑅𝑂𝑊𝑇𝐻 𝑖,𝑡−1(𝑆𝐴𝐿𝐸𝑆 _𝐺𝑅𝑂𝑊𝑇𝐻 𝑖,𝑡) is the percentage change in sales revenue at year t-1 (year 108
t); 𝑅𝑂𝐴 𝑖,𝑡 is return on assets, which is net income at year t scaled by total assets at year t-1; 109
∆𝑅𝑂𝐴 𝑖,𝑡−1(∆𝑅𝑂𝐴 𝑖,𝑡) is the change in ROA at year t-1 (year t); 𝑅𝐸𝑇𝑈𝑅𝑁 𝑖,𝑡 is the total annual stock 110
return for year t; 𝑆𝐼𝑍𝐸 _𝑅𝑖,𝑡−1 is the natural logarithm of the market value of equity at the beginning 111
of the year, ranked into percentiles; 𝑄𝑢𝑖𝑐𝑘 𝑖,𝑡−1 is the quick ratio at year t-1, which is calculated by the 112
sum of cash and cash equivalents, short -term investments, and receivables, divided by current 113
liabilities ; ∆𝑄𝑢𝑖𝑐𝑘 𝑖,𝑡−1(∆𝑄𝑢𝑖𝑐𝑘 𝑖,𝑡) is the change in the quick ratio at year t-1 (year t); 𝐿𝐸𝑉 𝑖,𝑡−1 is th e 114
ratio of long -term liabilities to to tal assets at year t-1; and five 𝐿𝑂𝑆𝑆𝐵𝐼𝑁 𝑖,𝑡−1 variables are used as a n 115
indicator of whether a firm’s ROA is classified into the five small loss bi ns with an interval of 0.005 116
at year t-1: 𝐿𝑂𝑆𝑆𝐵𝐼𝑁 1𝑖,𝑡−1 equals one if a firm’s ROA at year t-1 is between -0.005 and 0, 117
𝐿𝑂𝑆𝑆𝐵𝐼𝑁 2𝑖,𝑡−1 equals one if a firm’s ROA at year t-1 is between -0.01 and -0.005, 𝐿𝑂𝑆𝑆𝐵𝐼𝑁 3𝑖,𝑡−1 118
equals one if a firm’s ROA at year t-1 is between -0.015 and -0.01, 𝐿𝑂𝑆𝑆𝐵𝐼𝑁 4𝑖,𝑡−1 equals one if a firm’s 119
ROA at year t-1 is between -0.02 and -0.015, and 𝐿𝑂𝑆𝑆𝐵𝐼𝑁 5𝑖,𝑡−1 equals one if a firm’s ROA at year t- 120
1 is between -0.025 and -0.02. 121
If a firm has a positive (negative) abnormal net hiring, it is considered to have hired more ( fewer ) 122
employees than the predicted level . For both cases, when abnormal net hiring (𝐴𝐵_𝑁𝐸𝑇 _𝐻𝐼𝑅𝐸 𝑖,𝑡) is 123
either high or low, firms are defined as being inefficient in labor investment. The absolute value of 124
abnormal net hiring, |𝐴𝐵_𝑁𝐸𝑇_𝐻𝐼𝑅𝐸 |𝑖,𝑡, is a proxy for labor investment inefficiency. 125
The d escriptive statistics for the variables used in regression Model (1) are illustrated in Table 1 , 126
Panel A. The percentage change in employee s, on average (𝑁𝐸𝑇 _𝐻𝐼𝑅𝐸 𝑖,𝑡), is 2.6% , indicating that 127
sample firms generally over -hire rather than over -fire employ ees. Table 1 , Panel B presents results 128
from regression Model (1). The model ’s R-square equal s 8% and the F-statistics are 27.48. The 129
coefficients of independent variables indicate that a n increase in sales growth rate , the current year’s 130
performance, stock returns and size, and last year’s quick ratio and change in quick ratio increase 131
firms’ employment . On the other hand, employment dec reases as change s in performance, change s 132
in quick ratio, and last year’s debt ratio increase . Firms’ net hiring decreases when the firm’s ROA is 133
slightly below zero , but the rest of the 𝐿𝑂𝑆𝑆𝐵𝐼𝑁 𝑖,𝑡−1 variables are not statistically s ignificant. 134
135

Sustainabili ty 2019 , 11, x FOR PEER REVIEW 4 of 22
Table 1. Estimating the expected level of net hiring and abnormal hiring . 136
Panel A: Descriptive statistics for variables in Model (1 )
Variables N Mean Median Std. Q1 Q3
NET_HIRE it 25,122 0.0261 0.0067 0.2757 -0.0612 0.0860
SALES_GROWTH it-1 25,122 0.1957 0.0597 0.7303 -0.1242 0.2884
SALES_GROWTH it 25,122 0.1734 0.0475 0.7381 -0.1354 0.2603
∆ROA it-1 25,122 -0.4377 -0.2573 4.4229 -0.8725 0.2989
∆ROA it 25,122 -0.4552 -0.2445 4.9157 -0.8843 0.3415
ROA it 25,122 -0.0122 0.0253 0.1751 -0.0272 0.0672
RETURN it 25,122 0.2158 -0.0022 0.9261 -0.2689 0.3737
SIZE_R it-1 25,122 0.5020 0.4900 0.2747 0.2700 0.7400
Quick it-1 25,122 1.1683 0.7734 1.6862 0.4571 1.3192
∆Quick it-1 25,122 0.1562 0.0002 0.8008 -0.2192 0.2690
∆Quickt it 25,122 0.1324 -0.0072 0.7760 -0.2232 0.2468
LEV it-1 25,122 0.0338 0.0000 0.0684 0.0000 0.0377
Panel B: Regression re sults (dependent variable = NET_HIRE )
Independent variables Coeff. (t-value)
Intercept 0.122 (1.90) *
SALES_GROWTH it-1 0.044 (18.64) ***
SALES_GROWTH it 0.078 (33.11) ***
∆ROA it-1 0.000 (0.32)
∆ROA it -0.001 (-2.00) **
ROA it 0.105 (9.99) ***
RETUR Nit 0.019 (9.86) ***
SIZE_R it-1 0.049 (7.20) ***
Quick it-1 0.007 (6.31) ***
∆Quick it-1 0.005 (2.16) **
∆Quickt it -0.021 (-9.14) ***
LEV it-1 -0.074 (-2.86) ***
LOSSBIN1 it-1 -0.027 (-1.89) *
LOSSBIN2 it-1 -0.007 (-0.43)
LOSSBIN3 it-1 -0.002 (-0.14)
LOSSBIN4 it-1 0.000 (-0.02)
LOSSBIN5 it-1 0.012 (0.72)
Industry fixed effects Yes
[F-value] [27.48] ***
R2 0.0779
N 25,122
Notes : Panel A summarizes the descriptive statistics of the variables in the Model (1). Panel B reports 137
the regressio n results of NET_HIRE on various control variables. T-statistics are calculated based on 138
robust s tandard errors clustered at the firm level and are reported in parentheses. Statistical 139

Sustainabili ty 2019 , 11, x FOR PEER REVIEW 5 of 22
significance at the 1%, 5%, and 10% levels is denoted by ***, **, and * , respectively. See the Appendix 140
for variable definitions. 141
142
2.2.Specification of analysts’ group affiliation 143
In this paper, we categorize analysts into four groups depending on whether they are from 144
group -affiliated securities firms and whether they follow and issue forecasts for affiliated firms 145
within the same group . Firms are defined as group -affiliated when they belong to the list of chaebols 146
provided by the Korea Fair Trade Commission (KFTC) . We follow the process of Lim and Jung (2012) 147
to determine th e number of affiliated and non -affiliated analysts following a firm . The number of 148
affiliated analysts following within -group affiliated firms (nongroup or unaffiliated firms) is defined 149
as GAGF (GANGF ), and the number of unaffiliated analysts following gr oup firms (nongroup firms) 150
is defined as NGAGF (NGANGF ). Using these four items as our independent variables of interest in 151
regression models, we aim to examine whether analysts’ group affiliatio n affects the efficiency of 152
firms’ labor investment. 153
154
2.3.Control variables 155
Eleven variables are used in the regression to control for other factors that might affect firms ’ 156
employment decision s [15]. First, a firm’s size ( SIZE ) and its financials , such as gr owth options ( MTB ), 157
liquidity ( Quick ), and dividend payout ratio ( DIVDUM ), are included in the regression model. 158
Second, we include control variables regarding a firm’s financial risk : leverage ( LEV), having losses 159
(LOSS ), a tangible asset ratio ( TANGIBLE ), and volatility in cash flows and sales revenue ( STD_CFO , 160
STD_SALE ). Lastly, the model includes institutional ownership ( INSTI ) to control for corporate 161
governance , and volatility in net hiring ( STD_NET_HIRE ) to ensure that our results are not simply 162
driven by investment volatility . 163
2.3. Main regression model 164
|𝐴𝐵_𝑁𝐸𝑇 _𝐻𝐼𝑅𝐸 |𝑖,𝑡 165
=𝛼0+𝛽1𝐺𝐴𝐺𝐹 𝑖,𝑡−1+𝛽2𝐺𝐴𝑁𝐺𝐹 𝑖,𝑡−1+𝛽3𝑁𝐺𝐴𝐺𝐹 𝑖,𝑡−1+𝛽4𝑁𝐺𝐴𝑁𝐺𝐹 𝑖,𝑡−1+𝛾1𝑀𝑇𝐵 𝑖,𝑡−1+𝛾2𝑆𝐼𝑍𝐸 𝑖,𝑡−1 166
+𝛾3𝑄𝑢𝑖𝑐𝑘 𝑖,𝑡−1+𝛾4𝐿𝐸𝑉 𝑖,𝑡−1+𝛾5𝐷𝐼𝑉𝐷𝑈𝑀 𝑖,𝑡−1+𝛾6𝑆𝑇𝐷 _𝐶𝐹𝑂 𝑖,𝑡−1+𝛾7𝑆𝑇𝐷 _𝑆𝐴𝐿𝐸 𝑖,𝑡−1 167
+𝛾8𝑇𝐴𝑁𝐺𝐼𝐵𝐿𝐸 𝑖,𝑡−1+𝛾9𝐿𝑂𝑆𝑆 𝑖,𝑡−1+𝛾10𝐼𝑁𝑆𝑇𝐼 𝑖,𝑡−1+𝛾11𝑆𝑇𝐷 _𝑁𝐸𝑇 _𝐻𝐼𝑅𝐸 𝑖,𝑡−1 168
+Year Fixed Effects +Industry Fixed Effects +𝜀𝑖,𝑡 169
（2） 170
Where |𝐴𝐵_𝑁𝐸𝑇 _𝐻𝐼𝑅𝐸 |𝑖,𝑡 is abnormal net hiring , derived from Model (1); 𝐺𝐴𝐺𝐹 𝑖,𝑡−1 and 171
𝐺𝐴𝑁𝐺𝐹 𝑖,𝑡−1 are the number of affiliated analyst s following within -group firm s and unaffiliated firms, 172
respectively ; 𝑁𝐺𝐴𝐺𝐹 𝑖,𝑡−1 and 𝑁𝐺𝐴𝑁𝐺𝐹 𝑖,𝑡−1 are the numb er of nongroup analyst following group 173
firms and nongroup firms, respectively ; 𝑀𝑇𝐵 𝑖,𝑡−1 is the market to book ratio , which is calculated by 174
dividing the market value of equity by the book value of equity ; 𝑆𝐼𝑍𝐸 𝑖,𝑡−1 is the natural logarithm of 175
the market value of equity; 𝑄𝑢𝑖𝑐𝑘 𝑖,𝑡−1 is the sum of cash and cash equivalents, short -term 176
investments, and receiva bles, divided by current li abilities; 𝐿𝐸𝑉 𝑖,𝑡−1 is the ratio of long -term 177
liabilities to the beginning balance of total assets ; 𝐷𝐼𝑉𝐷𝑈𝑀 𝑖,𝑡−1 is an indicator variable that equals 178
one if a firm pays dividends and zero otherwise; 𝑆𝑇𝐷 _𝐶𝐹𝑂 𝑖,𝑡−1 is the standard deviation of the cash 179
flows from operations over the most recent five years; 𝑆𝑇𝐷 _𝑆𝐴𝐿𝐸 𝑖,𝑡−1 is the standard deviation of 180
sales r evenue ov er the most recent five years; 𝑇𝐴𝑁𝐺𝐼𝐵𝐿𝐸 𝑖,𝑡−1 is the ratio of long -term assets 181
(property, plant , and equipment) to the beginning balance of total assets; 𝐿𝑂𝑆𝑆 𝑖,𝑡−1 is an indicator 182
variable that equals one if a firm has a net loss and zero otherwise ; 𝐼𝑁𝑆𝑇𝐼 𝑖,𝑡−1 is the number of shares 183
owned by institutional investors scaled by the number of total outstanding shares; 184

Sustainabili ty 2019 , 11, x FOR PEER REVIEW 6 of 22
𝑆𝑇𝐷 _𝑁𝐸𝑇 _𝐻𝐼𝑅𝐸 𝑖,𝑡−1 is the standar d deviation of net hiring. Industry and year fixed effects are 185
included in all regressions, and standard error s are clustered at the firm level. 186
3. Empirical Results 187
3.1. Samples and data 188
The sample includes all firms whose stocks are publicly traded either on the Korea Composite 189
Stock Price Index ( KOSPI ) or the Korea Securities Dealers Automated Quotations ( KOSD AQ) from 190
2001 to 2017. The sample period starts in 2001, which is the year the KFTC started to release data 191
regarding chaebol group affiliation. Firms in financial industries and observations with missing firm 192
characteristics are excluded from the analyses . Three databases are used in this study: the a nalyst 193
following and ac counting data are from DataguidePro, the institutional ownership data is from 194
TS2000, and the firms’ group affiliation data is from the KFTC. The final sample consists of 7, 745 firm – 195
year observations. 196
197
3.2. Descriptive statistics and correlations 198
Descriptive s tatistics of the variables we used in this study are presented in Table 2. The mean 199
and median values of |𝐴𝐵_𝑁𝐸𝑇 _𝐻𝐼𝑅𝐸 |𝑖,𝑡, the dependent variable, are 0.1147 and 0.0615, respectively , 200
which is comparable to those found in Jung, Lee , and Weber [13]. On average, 7 -8 analysts are 201
following a particular firm. Moreover, a bout 3.68 analysts are group -affiliated analysts following 202
nongroup or unaffiliated firms , 2.29 are nongroup analysts follo wing nongroup firms, 1.45 are 203
nongroup analysts following group firms, and 0.05 are affiliated analysts following within -group 204
firms. 205
Table 2. Descriptive statistics for va riables in the abnormal net hiring model ( Model (2)). 206
Variables N Mean Median Std. Q1 Q3
|AB_NET_HIRE| 7,745 0.1147 0.0615 0.1688 0.0276 0.1282
Analyst coverage 7,745 7.4878 3.0000 8.5643 1.0000 11.0000
GAGF 7,745 0.0513 0.0000 0.2960 0.0000 0.0000
GANGF 7,745 3.6820 2.0000 4.5888 1.0000 5.0000
NGAGF 7,745 1.4546 0.0000 3.8266 0.0000 0.0000
NGANGF 7,745 2.2999 1.0000 3.3781 0.0000 3.0000
MTB 7,745 0.0015 0.0011 0.0056 0.0006 0.0018
SIZE 7,745 12.2492 11.9876 1.6809 11.0441 13.1880
Quick 7,745 1.0674 0.7648 1.3079 0.4869 1.2317
LEV 7,745 0.0386 0.0000 0.0667 0.0000 0.060 2
DIVDUM 7,745 0.7584 1.0000 0.4281 1.0000 1.0000
STD_CFO 7,745 108,945,56
1 13,449,56
2 436,717,500 5,350,999 44,152,740
STD_SALE 7,745 484,220,14
0 48,324,90
6 2,144,862,02
8 17,216,40
5 171,754,76
9
TANGIBLE 7,745 0.3335 0.3278 0.1812 0.1991 0.4556
LOSS 7,745 0.1434 0.0000 0.3506 0.0000 0.0000
INSTI 7,745 0.0499 0.0000 0.1404 0.0000 0.0000
STD_NET_HIRE 7,745 0.1922 0.1000 0.3346 0.0527 0.1925

Sustainabili ty 2019 , 11, x FOR PEER REVIEW 7 of 22
Notes : Table 2 presents the descriptive statistics of the variables in our main regression mode l 207
(Model (2)). See the Appendix for variable definitions. 208
209
Table 3 illustrates the Pearson correlations among variables , including an absolute value of 210
abnormal net hiring and four variables of analysts’ gr oup affiliation . The table shows that there is a 211
negative correlation between the total number of analysts and inefficacy in labor investm ent, 212
indicating that there exists a n analyst governance role . However, the correlations between the 213
absolute value of abnormal net hiring and analysts’ group affiliatio n are negative and significant 214
(except for NGAGF ), suggesting that a negative relation a pplies to almost all types of analysts. We, 215
therefore , conclude that the effect of different group affiliation of analysts on firm s’ labor efficiency 216
is not clear throu gh correlations. A firm’s labor investment efficiency has a positive correlation with 217
firm size ( SIZE ), dividend payout ratio ( DIVDUM ), tangible asset ratio ( TANGIBLE ), and volatility of 218
cash flow from operations and sales revenues ( STD_CFO , STD_SALE ). On the other hand, it has a 219
negative correlation with market -to-book ratio ( MTB ), quick rat io (Quick ), being a loss firm ( LOSS ), 220
and the standard deviation of net hiring ( STD_NET_HIRE ). 221

Sustaina bility 2019 , 11, x; doi: FOR PEER REVIEW www.mdp i.com/journal/ sustainab ility
Table 3. Correlations . 222
223
# Variables 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
1 |AB_NET_HIRE| 1.00 -0.05 -0.03 -0.04 -0.08 0.01 0.03 -0.07 0.05 0.02 -0.16 -0.05 -0.03 -0.09 0.10 -0.02 0.17
(0.00) (0.00) (0.00) (0.00) (0.32) (0.02) (0.00) (0.00) (0.11) (0.00) (0.00) (0.00) (0.00) (0.00) (0.12) (0.00)
2 Analyst following -0.05 1.00 0.32 0.94 0.70 0.43 0.03 0.68 -0.12 0.25 0.13 0.41 0.33 0.13 -0.08 0.18 -0.06
(0.00) (0.00) (0.00) (0.00) (0.00) (0.01) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
3 GAGF -0.03 0.32 1.00 0.28 0.42 -0.12 0.00 0.26 -0.06 0.10 0.04 0.30 0.24 0.02 -0.02 0.11 -0.04
(0.00) (0.00) (0.00) (0.00) (0.00) (0.84) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.08) (0.13) (0.00) (0.00)
4 GANGF -0.04 0.94 0.28 1.00 0.63 0.29 0.02 0.59 -0.11 0.23 0.12 0.35 0.28 0.13 -0.07 0.22 -0.05
(0.00) (0.00) (0.00) (0.00) (0.00) (0.08) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
5 NGAGF -0.08 0.70 0.42 0.63 1.00 -0.26 -0.01 0.61 -0.13 0.28 0.08 0.49 0.40 0.12 -0.02 0.07 -0.07
(0.00) (0.00) (0.00) (0.00) (0.00 ) (0.54) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.18) (0.00) (0.00)
6 NGANGF 0.01 0.43 -0.12 0.29 -0.26 1.00 0.05 0.20 0.00 -0.01 0.06 -0.01 -0.03 0.00 -0.10 0.06 0.00
(0.32) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.67) (0.29) (0.00) (0.64) (0.01) (0.81) (0.00) (0.00) (0.79)
7 MTB 0.03 0.03 0.00 0.02 -0.01 0.05 1.00 0.07 0.01 0.00 -0.06 -0.01 -0.01 -0.06 0.02 -0.05 0.03
(0.02) (0.01) (0.84) (0.08) (0.54) (0.00) (0.00) (0.31) (0.77) (0.00) (0.22) (0.26) (0.00) (0.04) (0.00) (0.01)
8 SIZE -0.07 0.68 0.26 0.59 0.61 0.20 0.07 1.00 -0.13 0.26 0.15 0.48 0.41 0.10 -0.11 -0.10 -0.07
(0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
9 Quick 0.05 -0.12 -0.06 -0.11 -0.13 0.00 0.01 -0.13 1.00 -0.12 0.01 -0.09 -0.07 -0.27 -0.05 -0.07 0.02
(0.00) (0.00) (0.00) (0.00) (0.00) (0.67) (0.31) (0.00) (0.00) (0.60) (0.00) (0.00) (0.00) (0.00) (0.00) (0.13)
10 LEV 0.02 0.25 0.10 0.23 0.28 -0.01 0.00 0.26 -0.12 1.00 -0.09 0.21 0.18 0.18 0.14 0.11 0.01
(0.11) (0.00 ) (0.00) (0.00) (0.00) (0.29) (0.77) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.44)
11 DIVDUM -0.16 0.13 0.04 0.12 0.08 0.06 -0.06 0.15 0.01 -0.09 1.00 0.04 0.06 0.06 -0.48 0.02 -0.18
(0.00) (0.00) (0.00 ) (0.00) (0.00) (0.00) (0.00) (0.00) (0.60) (0.00) (0.00) (0.00) (0.00) (0.00) (0.14) (0.00)

Sustainabili ty 2019 , 11, x FOR PEER REVIEW 2 of 22
12 STD_CFO -0.05 0.41 0.30 0.35 0.49 -0.01 -0.01 0.48 -0.09 0.21 0.04 1.00 0.75 0.06 0.00 0.00 -0.05
(0.00) (0.00) (0.00) (0.00) (0.00) (0.64) (0.22) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.78) (0.77) (0.00)
13 STD_SALE -0.03 0.33 0.24 0.28 0.40 -0.03 -0.01 0.41 -0.07 0.18 0.06 0.75 1.00 0.01 -0.02 -0.01 -0.03
(0.00) (0.00) (0.00) (0.00) (0.00) (0.01) (0.26) (0.00) (0.00) (0.00) (0.00) (0.00) (0.32) (0.14) (0.61) (0.03)
14 TANGIBL E -0.09 0.13 0.02 0.13 0.12 0.00 -0.06 0.10 -0.27 0.18 0.06 0.06 0.01 1.00 0.02 0.15 -0.09
(0.00) (0.00) (0.08) (0.00) (0.00) (0.81) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.32) (0.07) (0.00) (0.00)
15 LOSS 0.10 -0.08 -0.02 -0.07 -0.02 -0.10 0.02 -0.11 -0.05 0.14 -0.48 0.00 -0.02 0.02 1.00 0.03 0.08
(0.00) (0.00) (0.13) (0.00) (0.18) (0.00) (0.04) (0.00) (0.00) (0.00) (0.00) (0.78) (0.14) (0.07) (0.02) (0.00)
16 INSTI -0.02 0.18 0.11 0.22 0.07 0.06 -0.05 -0.10 -0.07 0.11 0.02 0.00 -0.01 0.15 0.03 1.00 -0.02
(0.12) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.14) (0.77) (0.61) (0.00) (0.02) (0.12)
17 STD_NET_HIRE 0.17 -0.06 -0.04 -0.05 -0.07 0.00 0.03 -0.07 0.02 0.01 -0.18 -0.05 -0.03 -0.09 0.08 -0.02 1.00
(0.00) (0.00) (0.00) (0.00) (0.00) (0.79) (0.01) (0.00) (0.13) (0.44) (0.00) (0.00) (0.03) (0.00) (0.00) (0.12)
Notes : Table 3 presents the Pearson correlation matrix of the variables in our main regression model ( Model (2)). p-values are presented in parentheses. See the 224
Appendix for variable definitions. 225

Sustaina bility 2019 , 11, x; doi: FOR PEER REVIEW www.mdp i.com/journal/ sustainab ility
226
3.3. Regression results 227
Using Model (2), we examine d the association between analysts’ group -affiliation and a 228
firm’s abnormal net hiring . The results are report ed in Table 4 . The regression result using a full 229
sample is shown in column (1) . The coefficient of NGAGF is negative and significant ( -0.0019 ; p-value 230
< 0.05 ), while the coefficients o f GAGF , GANGF , and NGANGF are insignificant. This result indicate s 231
that la bor invest ment efficiency increases in group firms (chaebols) as there are more nongroup 232
analysts following , but it is not affected by group -affiliated analysts. Coefficients on control variables 233
suggest that abnormal net hiring decreases when a firm paid dividends (DIVDUM ) or had fewer 234
tangible assets (TANGIBLE ) in the prior year . On the other hand , abnormal net hiring increases when 235
a firm had a loss (LOSS ), more quick assets (Quick ), or higher volatil ity in net hiring (STD_NET_HIRE ) 236
in a prior year. 237
Table 4. The effect of analyst following on abnormal net hiring . 238
Dependent variable: |AB_NET_HIRE|
Full sample AB_NET_HIRE > 0 AB_NET_HIRE < 0
Independent variables (1) (2) (3)
Intercept 0.3483 *** 0.9784 *** 0.1393 ***
(8.76) (13.98) (3.26)
GAGF -0.0033 -0.0013 -0.0086
(-0.61) (-0.17) (-1.35)
GANGF 0.0006 -0.0001 0.0003
(0.67) (-0.06) (0.27)
NGAGF -0.0019 ** 0.0002 -0.0029 ***
(-1.98) (0.12) (-2.94)
NGANGF -0.0008 -0.0008 -0.0011
(-0.98) (-0.59) (-1.27)
MTB 0.3882 9.7148 *** 0.1766 *
(0.98) (3.33) (1.75)
SIZE -0.0011 -0.0113 ** 0.0056 **
(-0.44) (-2.30) (2.05)
Quick 0.0041 * 0.0019 0.0064 ***
(1.87) (0.44) (3.10)
LEV 0.0895 * 0.1259 0.0562
(1.87) (1.56) (1.12)
DIVDUM -0.0421 *** -0.0375 *** -0.0393 ***
(-7.12) (-3.79) (-6.29)
STD_CFO 0.0000 0.0000 0.0000
(-0.98) (1.17) (-1.49)
STD_SALE 0.0000 0.0000 ** 0.0000
(-0.76) (-2.14) (0.87)
TANGIBLE -0.0346 ** -0.0453 * -0.0169
(-2.10) (-1.73) (-1.05)

Sustainabili ty 2019 , 11, x FOR PEER REVIEW 2 of 22
LOSS 0.0162 ** 0.0204 0.0217 ***
(2.29) (1.57) (2.91)
INSTI -0.0341 -0.0205 -0.0298
(-1.45) (-0.47) (-1.27)
STD_NET_HIRE 0.0582 *** 0.0742 *** 0.0390 ***
(6.25) (4.68) (3.88)
Year -fixed e ffect Yes Yes Yes
Industry -fixed effect Yes Yes Yes
[F-value] [21.47] *** [10.09] *** [16.63] ***
R2 0.092 0.113 0.134
N 7,745 3,447 4,298
Notes : Table 4 presents the regression results of abnormal net hiring ( |AB_NET_HIRE| or AB_NET _HIRE ) on 239
analysts ’ group -affiliation (GAGF , GANGF , NGAGF , NGANGF ) and control variables. Column (1) reports 240
regression results when an absolute value of abnormal net hiring ( |AB_NET_HIRE| ) is used as the dependent 241
variable , and Column s (2) and (3) report results using a signed abnormal net hiring (AB_NET_HIRE ) as the 242
dependent variable. Model (1) test s the full sample and Model s (2) and (3) test two different subsamples having 243
positive and negative abnormal net hiring, respectively . T-statistics are calcul ated based on robust standard 244
errors clustered at the firm level and are reported in parentheses below the coefficient estimat es. The F-value is 245
reported in square bracket s. Statistical significance at the 1%, 5%, and 10% levels is denoted by ***, **, and *, 246
respectively. See the Appendix for variable definitions. 247
248
Columns (2) a nd (3) report regression results using two different subsamples . In the full 249
sample of 7,745 firm -year observations, 3,447 observations have positive abnormal net hiring , while 250
4,298 observations have negative abnormal net hi ring. Depending on whether 𝐴𝐵_𝑁𝐸𝑇 _𝐻𝐼𝑅𝐸 𝑖,𝑡 is 251
positive or negative , we separate d the observations into two subsamples and defined overinvestment 252
(underinvestment) group s as firms whose abnormal net hiring is above (below) zero. After running 253
Model (2) for each of two subgroups , we find th at the coefficients o f GAGF , GANGF , and NGANGF 254
are insignificant for both subsamples , suggesting that group -affiliated analysts and nongroup 255
analysts in nongroup firms do not affect abnormal net hiring. However, while the coefficie nt of 256
NGAGF is insignific ant using the overinvestment group, it is negative and significant for the 257
underinve stment group . Thus, w e argue that unaffiliated analysts are the ones who are enhancing 258
the labor investment efficiency of group firms by resolving underinvestment problem s. Overall , our 259
results support the conflict -of-interest theory rather than the information sharing theory. 260
A firm’s net hiring captures changes in employees , which is the number of hired employees 261
minus the number of fired employee s. Following Jung, Lee , and Weber [13], we form ed four groups 262
of firms based on the expected difference between the number of hired employees and the number 263
of fired employees : (1) The o ver-hiring group consists of firms who are overinvesting while positiv e 264
net hiring is expected (abnormal net hiring and expected net hiring are positive); (2) The u nder -firing 265
group consists of firms who are overinvesting while negative net hiring is expected (abnormal net 266
hiring is positive and expected net hiring is negati ve); (3) The u nder -hiring group consists of firms 267
who are underinvesting while positive net hiring is expected (abnormal net hiring is negative and 268
expected net hiring is positive); (4) The o ver-firing group consists of firms who are underinvesting 269
while negative net hiring i s exp ected (abnormal net hiring and expected net hiring is negative). Out 270
of the total sample, the proportions of these four groups are 42%, 2%, 19%, and 37%, respectively. 271
Using four subsamples of companies with different types of labor investment ineffi cienc ies, 272
we test ed the effect of analysts’ group affiliation. As presented in Table 5, the coefficient of NGAGF 273

Sustainabili ty 2019 , 11, x FOR PEER REVIEW 3 of 22
is negative and significant only in the over -firing sample. This indicates that our results reg arding a 274
positive effect of nongroup analysts on group firms’ labor investment efficiency hold in the over – 275
firing sample. Thus, we conclude that unaffiliated analysts improve nongroup firms’ labor 276
investment efficiency by decreasing the over -firing problem, while analysts who work fo r group – 277
affiliated firms do not positively affect labor investment efficiency . 278
Table 5. The effect of analyst following on over – and under -hiring (and firing) . 279
Dependent variable: |AB_NET_HIRE|
Over -hiring Under -firing Under -hiring Over -firing
Independent variables (1) (2) (3) (4)
Intercept 0.9689 *** 0.0422 * 0.2125 0.1232 *
(13.47) (1.70) (8.87) (1.93)
GAGF -0.0028 0.0055 0.0036 -0.0109
(-0.34) (0.49) (0.79) (-1.16)
GANGF -0.0002 0.0003 0.0007 0.0003
(-0.16) (0.54) (1.20) (0.21)
NGAGF -0.0002 0.0002 -0.0005 -0.0034 **
(-0.13) (0.23) (-0.83) (-2.24)
NGANGF -0.0010 0.0009 0.0002 -0.0008
(-0.77) (1.30) (0.47) (-0.59)
MTB 9.2717 *** 3.2007 * 2.6103 0.1570
(3.15) (1.80) (2.30) (1.10)
SIZE -0.0104 ** -0.0034 * -0.0017 0.0083 **
(-2.08) (-1.96) (-1.05) (2.17)
Quick 0.0013 -0.0045 0.0050 0.0100 **
(0.30) (-0.92) (2.87) (2.45)
LEV 0.1494 * 0.0048 0.0581 0.0389
(1.75) (0.27) (1.60) (0.61)
DIVDUM -0.0386 *** 0.0040 -0.0188 -0.0387 ***
(-3.74) (0.91) (-3.73) (-4.94)
STD_CFO 0.0000 0.0000 0.0000 0.0000
(1.40) (0.61) (-0.63) (-0.96)
STD_SALE 0.0000 ** 0.0000 0.0000 0.0000
(-2.31) (-0.06) (0.13) (0.29)
TANGIBLE -0.0441 0.0003 -0.0211 -0.0229
(-1.61) (0.02) (-2.02) (-1.02)
LOSS 0.0302 ** 0.0036 -0.0056 0.0119
(2.14) (0.75) (-0.79) (1.33)
INSTI -0.0052 -0.0052 0.0119 -0.0313
(-0.11) (-0.41) (0.88) (-1.06)
STD_NET_HIRE 0.0742 *** -0.0002 0.0055 0.0409 ***
(4.60) (-0.07) (0.82) (3.09)
Year -fixed effect Yes Yes Yes Yes
Industry -fixed effect Yes Yes Yes Yes

Sustainabili ty 2019 , 11, x FOR PEER REVIEW 4 of 22
[F-value] [10.55] *** [27.18] *** [342.62] *** [33.35] ***
R2 0.114 0.575 0.252 0.146
N 3,316 151 1,527 2,891
Notes: Table 5 reports the results of estimating Model (2) on four different subsamples. Over -hiring is the 280
sample where both abnormal net hiring and expected net hiring are positive. Under -firing is the sample where 281
abnormal ne t hiring is positive and expected net hiring is negative. Under -hiring is the sample where abnormal 282
net hiring is negative and expected net hiring is positive. Over -hiring is the sample where abnormal net hiring 283
is negative and expected net hiring is negat ive. T-statistics are calculated based on robust standard errors 284
clustered at the firm level and are reported in parentheses below the coefficient estimates . The F-value is 285
reported in square bracket s. Statistical significance at the 1%, 5%, and 10% levels is denoted by ***, **, and *, 286
respectively. See the Appendix for variable definitions. 287
288
As shown in column (4), o ther factors such as firm size ( SIZE ), quick asset ratio ( Quick ), and a 289
volatility in net hiring ( STD_NET_HIRE ) increase over -firing, but dividend payme nts (DIVDUM ) 290
decreases ov er-firing. On the other hand, regarding the overinvestment problem, larger firms ( SIZE ), 291
and firms with fewer growth opportunities ( MTB ), leverage ( LEV), volatility in sales revenue 292
(STD_SALE ), and net hire ( STD_NET_HIRE ) decrease over -hiring. Also, loss f irms (LOSS ) or firms 293
who paid dividends (DIVDUM ) have less of an over -hiring problem . Lastly, firms’ under -firing 294
decreases for larger firms ( SIZE ) with less market -to-book ratio ( MTB ). 295
296
4. Additional Tests 297
4.1. The impact of inside fund o n labor investmen t efficiency 298
We document a positive association between nongroup , unaffiliated analysts and labor 299
investment efficiency in group firms. Although analysts can influence firms’ employment decisions, 300
it would not be possible for firms to cha nge their employme nt policy without sufficient fund s. Given 301
that the cost of financing increases in the order of internal funds, debt, and equity ( pecking order 302
theory), financing through internal fund s can induce more efficient investment. Therefore , we e xpect 303
to see our main result hold when a firm has sufficient internal fund s, which is measured by the level 304
of cash and cash equivalent s. A firm is considered to have a high (low) level of cash when its cash 305
and cash equivalents scaled by total assets is a bove (below) the yearly median. We then re-estimate d 306
regression Model (2) by using these two subsample s. 307
Table 6 presents the regression results from this test of the effect of funds within the company 308
on the relation ship between nongroup analysts follow ing and labor inves tment efficiency. As we 309
predicted, the positive effect of nongroup analysts on group firms’ labor investment efficiency holds 310
only when firms have sufficient funds : the coefficient of NGAGF is negative and significant at the 311
10% level . The coefficient of NGAGF is negative but insignificant when using a subsample of firms 312
with low cash level s. Thus, the result implies that the nongroup analysts’ effect on the increase in 313
labor investment efficiency becomes stronger when a firm has high cas h holding s. 314
Table 6. The effect of inside funding on our hypothesized relationship . 315
Dependent variable: |AB_NET_HIRE|
High Fund Low Fund
Independent variables (1) (2)
Intercept 0.3563 *** 0.3027 ***

Sustainabili ty 2019 , 11, x FOR PEER REVIEW 5 of 22
(6.05) (6.39)
GAGF -0.0069 0.0008
(-0.67) (0.13)
GANGF 0.0003 0.0008
(0.22) (0.87)
NGAGF -0.0026 * -0.0013
(-1.76) (-1.00)
NGANGF -0.0016 -0.0015
(-1.38) (-1.36)
MTB 6.8580 *** 0.1147
(2.77) (0.69)
SIZE -0.0013 -0.0033
-0.31) (-1.10)
Quick 0.0030 0.0053
(1.04) (0.87)
LEV 0.1744 ** 0.0522
(2.14) (0.96)
DIVDUM -0.0440 *** -0.0301 ***
(-4.72) (-3.75)
STD_CFO 0.0000 0.0000
(0.12) (-1.07)
STD_ SALE 0.0000 0.0000
(-0.44) (-0.13)
TANGIBLE -0.0450 * 0.0154
(-1.80) (0.72)
LOSS 0.0469 *** -0.0039
(3.73) (-0.57)
INSTI 0.0044 -0.0524 **
(0.09) (-2.41)
STD_NET_HIRE 0.0659 *** 0.0401 ***
(4.36) (3.97)
Year -fixed effect Yes Yes
Industry -fixed effect Yes Yes
[F-value] [13.35] *** [153.16] ***
R2 0.122 0.088
N 3,873 3,872
Notes: Table 6 reports the results of estimating Model (2) on two different subsamples. We separated 316
firms into two groups: one with greater inside funding and the other with lower inside funding. A 317
firm is considered to have a greater (lower) cash holding if a firm’s c ash and cash equivalents are 318
greater (less) than the sample yearly median . T-statistics are calculated based on robust standard 319
errors clustered at the firm level and are reported in parentheses below the coefficient estimates. Th e 320
F-value is reported in square bracket s. Statistical significance at the 1%, 5%, and 10% levels is denoted 321
by ***, **, and *, respectively. See the Appendix for variable definitions. 322

Sustainabili ty 2019 , 11, x FOR PEER REVIEW 6 of 22
323
324
4.2. Sensitivity test – endogeneity 325
We also conducted a test to ensure that our main finding is causal. O ur main specification 326
instruments analyst s following using five different factors ( firm size, performance, growth rate, 327
external financing, an d volatility of cash flow s), and then estimates the relationship between analyst s 328
following and labor investment efficiency using a 2SLS regression model . Following Yu (2008), f ive 329
factors are selected to control for other factors that affect analyst cover age. In the fir st stage, we 330
model ed the estimate for the number of analysts following as: 331
𝐺𝐴𝐺𝐹 𝑖,𝑡 𝑜𝑟 𝐺𝐴𝑁𝐺𝐹 𝑖,𝑡 𝑜𝑟 𝑁𝐺𝐴𝐺𝐹 𝑖,𝑡 𝑜𝑟 𝑁𝐺𝐴𝑁𝐺𝐹 𝑖,𝑡 =𝛼0+𝛽1𝑀𝐾𝑉 𝑖,𝑡−1+
𝛽2𝑅𝑂𝐴 𝑖,𝑡−1+𝛽3𝐺𝑅𝑂𝑊𝑇𝐻 𝑖,𝑡−1+𝛽4𝐸𝑋𝐹𝐼𝑁𝐴𝐶𝑇 𝑖,𝑡−1+𝛾1𝐶𝐹𝑉𝑜𝑙𝑎𝑡𝑖𝑙𝑖𝑡𝑦 𝑖,𝑡−1+
Year Fixed Effects +𝜀𝑖,𝑡 (3)
Where the dependent variable is one of the analysts’ group affiliation variables ; MKV is the 332
market value; ROA is return on assets, which is measured by net inco me divided by total assets; 333
GROWTH is the growth rate of total assets, calculated by dividing change in assets by last year’s 334
assets; EXFINACT is net cash proceeds from external financing scaled by total assets; and CFVolatility 335
is a standard deviation of cash flow scaled by last year’s total assets. 336
The residuals from regression Model (3) are used as independent variables of interest in the next 337
step (regression Model (2)) to obtain the effect of analyst following that is uncorrelat ed with firm size, 338
prof itability, growth, external financing, and cash flow volatility. The results of the first and second 339
step regressions are illustrated in Table 7. The results of the first stage regression indicate that firm 340
size ( MKV ) is positively a ssociated with analyst following. For group firms, cash flow volatility 341
(CFVolatility ) is negatively correlated with analyst following , and f or nongroup firms, past 342
profitability ( ROA ) increases analyst following while external financing activities ( EXFINA CT) 343
decreases analyst f ollowing. Our main interest is in the second stage regression in which we estimate 344
the effect of analyst following on abnormal net hiring . Consistent with the res ults in Table 4, none of 345
the coefficients o f group -affiliated analyst f ollowing ( residual(GAGF ), residual(GANGF) ) are 346
significant , indicating that affiliated analysts do not influence firms’ labor investment efficiency. 347
However, we find that nongroup unaffiliated analysts following decrease firms’ abnormal net hiring 348
regardle ss of whether following firms are group or nongroup . Specifically, b oth residual(NGAGF) and 349
residual(NGANGF) have negative and significant coefficients. After addressing the potential 350
endogeneity problem, we confirm that our main finding that nongroup anal ysts following increase s 351
firms’ labor investment efficiency holds. 352
Table 7. 2SLS regression . 353
First stage : Regression to estimate expected level of analyst coverage
Dependent variable:
Independent variables GAGF
(1) GANGF
(2) NGAGF
(3) NGANG F
(4)
Intercept -0.027 * -0.809 *** -0.455 ** 5.120 ***
(-1.74) (-3.57) (-2.38) (29.68)
MKV 0.000 *** 0.000 *** 0.000 *** 0.000 ***
(24.15) (26.42) (33.97) (3.89)
ROA -0.020 2.653 *** -0.770 ** 3.424 ***
(-0.75) (6.90) (-2.37) (11.69)

Sustainabili ty 2019 , 11, x FOR PEER REVIEW 7 of 22
GROWTH 0.002 0.018 0.051 0.002
(0.31) (0.16) (0.54) (0.02)
EXFINACT 0.015 -3.211 *** -0.485 -2.877 ***
(0.42) (-6.04) -1.08) -7.11)
CFVolatility -0.058 ** -1.504 *** -2.051 *** 0.581 **
(-2.36) (-4.14) (-6.69) (2.10)
Year fixed effects Yes Yes Yes Yes
[F-value] [31.85] *** [71.38] *** [64.75] *** [41.54] ***
R2 0.0818 0.1664 0.1533 0.104
N 7,533 7,533 7,533 7,533
Second stage : Regression using residual analyst coverage
Dependent variable: |AB_NET_HIRE|
Full sample AB_NET_HIRE > 0 AB_NET_HIRE < 0
Independent variables (1) (2) (3)
Intercept 0.3500 *** 0.3500 *** 0.1518 ***
(9.07 ) (9.07) (3.59)
residual(GAGF) -0.0037 -0.0037 -0.0081
(-0.69) (-0.69) (-1.27)
residual(GANGF) 0.0007 0.0007 0.0007
(0.85) (0.85) (0.76)
residual(NGAGF) -0.0017 * -0.0017 * -0.0029 ***
(-1.75) (-1.75) (-3.12)
residual(NGANGF) -0.0015 * -0.0015 * -0.0015 *
(-1.96) (-1.96) (-1.80)
MTB 0.3766 0.3766 0.1657 *
(0.98) (0.98) (1.78)
SIZE -0.0014 -0.0014 0.0044 *
(-0.60) (-0.60) (1.70)
Quick 0.0041 * 0.0041 * 0.0062 **
(1.68) (1.68) (2.26)
LEV 0.0964 * 0.0964 * 0.0631
(1.95) (1.95) (1.21)
DIVDUM -0.0434 *** -0.0434 *** -0.0395 ***
(-7.10) (-7.10) (-6.07)
STD_CFO 0.0000 * 0.0000 * 0.0000 ***
(-1.78) -1.78) (-2.67)
STD_SALE 0.0000 0.0000 0.0000
(-0.95) (-0.95) (0.80)
TANGIBLE -0.0355 ** -0.0355 ** -0.0206
(-2.11) (-2.11) (-1.25)
LOSS 0.0152 ** 0.0152 ** 0.0224 ***
(2.10) (2.10) (2.91)
INSTI -0.0358 -0.0358 -0.0309

Sustainabili ty 2019 , 11, x FOR PEER REVIEW 8 of 22
(-1.49) (-1.49) (-1.28)
STD_NET_HIRE 0.0589 *** 0.0589 *** 0.0404 ***
(6.14) (6.14) (3.83)
Year -fixed effect Yes Yes Yes
Industry -fixed effect Yes Yes Yes
[F-value] [17.08] *** [17.08] *** [16.65] ***
R2 0.093 0.093 0.132
N 7,533 7,533 7,533
Notes: Table 7 reports the results from 2SLS regressio ns to test the effect of analysts ’ group affiliation 354
on labor investment efficiency . Panel A presents the results of the first stage , where analysts’ group 355
affiliation variables ( GAGF , GANGF , NGAGF , NGANGF ) are instru mented with five variables (MKV , 356
ROA , GROWTH , EXFINACT , CFVolatility ). Panel B present s the estimates from the second -stage 357
regressions. The dependent variables are the absolute value of abnormal net hiring 358
(|AB_NET_HIRE| ) for the full sample (Column (1)) , and a signed abnormal net hiring 359
(AB_NET_HIRE ) for subsample analyses (Column s (2) and (3)) . T-statistics are calculated based on 360
robust standard errors clustered at the firm level and are reported in parentheses below the coefficient 361
estimates. The F -value is reported in square bracket s. Statistical significance at the 1%, 5 %, and 10% 362
levels is denoted by ***, **, and *, respectively. See the Appendix for variable definitions. 363
364
365
4.3. Sensitivity test – matching firm size 366
Lastly, a s a robustness test, we use size ma tching to reduce potential measurement error in the 367
main regression model. In particular, we divided firm s into two groups , one with firms that have a 368
positive number of group analysts following nongroup or unaffiliated firms ( GANGF ) and another 369
with firms that have a zero GANGF value . Based on firm size, the top and bottom 10% of the treatment 370
sample (positi ve GANGF group) is matched with a control sample ( zero GAGNGF group) using a 371
random selection method , which is similar to what Mo and Lee suggested [16]. 372
Table 8 shows the results after using size matching. As shown in the table, after we match the 373
treatm ent sample to the co ntrol group by size , the coefficients of analyst following become 374
insignificant in the full sample (column (1)). With an overinvestment subsample, there is a negative 375
association between nongroup analysts following and labor investment efficiency of group firms , 376
indicating that nongroup analysts negatively affect group firms’ overinvestment in labor. However, 377
when it comes to the underinvestment problem, both affiliated and unaffiliated analysts enhance 378
group firms’ labor investment effi ciency. In sum, the results suggest that analyst following generally 379
helps group firms overcome labor underinvestment , but do not help them overcome overinvestment 380
(Table 8) . 381
Table 8. Size matching . 382
Dependent variable: |AB_NET_HIRE|
Full sample AB_NET_HIRE > 0 AB_NET_HIRE < 0
Independent variables (1) (2) (3)
Intercept -0.0595 0.4423 ** -0.0622
(-0.54) (2.39) (-0.62)
GAGF -0.0655 – -0.0862 *

Sustainabili ty 2019 , 11, x FOR PEER REVIEW 9 of 22
(-1.14) (-1.90)
GANGF 0.0154 0.0202 -0.0051
(1.51) (1.22) (-0.57)
NGAGF 0.0134 0.0767 * -0.0093 *
(0.72) (1.88) (-1.77)
NGANGF -0.0043 -0.0103 0.0015
(-1.14) (-1.62) (0.30)
MTB 6.3553 9.0011 1.2786
(1.37) (1.02) (0.23)
SIZE 0.0044 -0.0070 0.0144 *
(0.55) (-0.44) (1.92)
Quick 0.0004 -0.0072 0.0072
(0.07) (-0.52) (1.46)
LEV -0.0308 -0.0381 0.0243
(-0.28) (-0.20) (0.15)
DIVDUM -0.0262 0.0114 -0.0563 ***
(-1.57) (0.41) (-2.88)
STD_CFO 0.0000 0.0000 0.0000
(0.27) (0.19) (0.14)
STD_SALE 0.0000 0.0000 0.0000
(-1.39) (-1.07) (0.59)
TANGIBLE -0.0988 *** -0.1287 * -0.0597 *
(-2.65) (-1.93) (-1.67)
LOSS 0.0472 ** 0.0430 0.0398 **
(2.30) (1.21) (2.01)
INSTI -0.0909 *** -0.1063 -0.1185 **
(-2.70) (-1.54) (-2.10)
STD_NET_HIRE 0.0557 *** 0.1058 *** 0.0231
(2.93) (3.05) (1.05)
Year -fixed effect Yes Yes Yes
Industry -fixed effect Yes Yes Yes
[F-value] [59.08] *** [79.99] *** [4.72] ***
R2 0.186 0.321 0.311
N 1,710 1,054 656
Notes : Table 8 reports the results of estimating Model (2) after matching firms by their sizes . The 383
dependent variables are an absolute value of abnormal net hiring ( |AB_NET_HIRE| ) for the full 384
sample (Column (1)) , and a signed abnormal net hiring (AB_NET_HIRE ) for subsample analyses 385
(Columns (2) and (3)). The i ndependent variable of interest is one of the analysts’ group affiliation 386
variables ( GAGF , GANGF , NGAGF , NGANGF ). T-statistics are calculated based on robust standard 387
errors clustered at the firm level an d are reported in parentheses below the coefficient estimates. The 388
F-value is reported in square bracket s. Statistical significance at the 1%, 5%, and 10% levels is d enoted 389
by ***, **, and *, respectively. See the Appendix for variable definitions. 390

Sustainabili ty 2019 , 11, x FOR PEER REVIEW 10 of 22
. 5.Conclusion 391
Despite the importance of analysts’ independence, it has not been sufficiently emphas ized as 392
one of the virtue s of analysts . Prior studies have shown that affiliated analysts are more likely to issue 393
inaccurate earnings forecast s and biased recommendations . This implies that analysts’ group 394
affiliation lowers their indepen denc e, thus decre asing the quality of information they provide to the 395
market. In this paper, we examine d whether analysts’ group affiliation affects corporate decisions , 396
which go beyond earnings forecast s and stock recommendations . 397
We stud ied the relation ship between analy sts’ group affiliation and firms’ labor investment 398
decision s. Analysts respond to firms’ use of strategic staff planning , and they advance a view on 399
corporate employment decision s. Considering that companies listen to what analysts say, it was 400
expected tha t analyst following would influence the efficiency of firms’ employment decisions. In 401
this paper, we f ound that there is an increase in labor investment efficiency of group firms when 402
unaffiliated analysts ’ following increases. The result s from further tes ts show ed that an increase in 403
efficiency of labor investment is driven by resolving the underinvestment problem, especially in 404
decreasing firms’ over -firing problem . On the other hand, when analysts are affiliated with firms they 405
follow, their coverage doe s not affect firms’ labor investment efficiency. Additional analyses prove 406
that the positive association between unaffiliated analyst coverage and labor investment efficiency 407
becomes stronger when there is high cashflow within firms . Our results are robust to different model 408
specifications as well . 409
410
Author Contribut ions: 411
Conflicts of Interest: “The authors declare no conflict of interest. ” 412
413

Sustainabili ty 2019 , 11, x FOR PEER REVIEW 11 of 22
Appendix. Variable Definitions 414
Variable name Variable definition
Dependent Variables
NET_HIRE
The percentage change in the number of employees
AB_NET_HIRE Abnormal net hir ing, defined as residuals from the following model:
𝑁𝐸𝑇 _𝐻𝐼𝑅𝐸 𝑖,𝑡=𝛼0+𝛼1𝑆𝐴𝐿𝐸𝑆 _𝐺𝑅𝑂𝑊𝑇𝐻 𝑖,𝑡−1+𝛼2𝑆𝐴𝐿𝐸𝑆 _𝐺𝑅𝑂𝑊𝑇𝐻 𝑖,𝑡+𝛼3∆𝑅𝑂𝐴 𝑖,𝑡−1+
𝛼4∆𝑅𝑂𝐴 𝑖,𝑡+𝛼5𝑅𝑂𝐴 𝑖,𝑡+𝛼6𝑅𝐸𝑇𝑈𝑅𝑁 𝑖,𝑡+𝛼7𝑆𝐼𝑍𝐸 _𝑅𝑖,𝑡−1+𝛼8𝑄𝑢𝑖𝑐𝑘 𝑖,𝑡−1+
𝛼9∆𝑄𝑢𝑖𝑐𝑘 𝑖,𝑡−1+𝛼10∆𝑄𝑢𝑖𝑐𝑘 𝑖,𝑡+𝛼11𝐿𝐸𝑉 𝑖,𝑡−1+𝛼12𝐿𝑂𝑆𝑆𝐵𝐼𝑁 1𝑖,𝑡−1+𝛼13𝐿𝑂𝑆𝑆𝐵𝐼𝑁 2𝑖,𝑡−1+
𝛼14𝐿𝑂𝑆𝑆𝐵𝐼𝑁 3𝑖,𝑡−1+𝛼15𝐿𝑂𝑆𝑆𝐵𝐼𝑁 4𝑖,𝑡−1+𝛼16𝐿𝑂𝑆𝑆𝐵𝐼𝑁 5𝑖,𝑡−1+Industry Fixed Effects +
𝜀𝑖,𝑡.
|AB_ NET_HIRE|
The absolute value of a bnormal net hiring

Analyst Following Variables
Analyst Following The number of analysts covering a firm
GAGF The number of affiliated analysts following within -group firms
GANGF The number of affiliated analysts f ollowing unaffiliated firms
NGAGF The number of unaffiliated analysts following group firms
NGANGF The number of unaffiliated analysts following nongroup firms
residual( GAGF) The residual coverage of affiliated analysts on within -group firms, defined as residuals from
the following model:
𝐺𝐴𝐺𝐹 𝑖,𝑡=𝛼0+𝛽1𝑀𝐾𝑉 𝑖,𝑡−1+𝛽2𝑅𝑂𝐴 𝑖,𝑡−1+𝛽3𝐺𝑅𝑂𝑊𝑇𝐻 𝑖,𝑡−1+𝛽4𝐸𝑋𝐹𝐼𝑁𝐴𝐶𝑇 𝑖,𝑡−1+
𝛾1𝐶𝐹𝑉𝑜𝑙𝑎𝑡𝑖𝑙𝑖𝑡𝑦 𝑖,𝑡−1+Year Fixed Effects +𝜀𝑖,𝑡
residual( GANG F) The residual coverage of affiliated analysts on unaffiliated firms, defined as residuals from the
following model:
𝐺𝐴𝑁𝐺𝐹 𝑖,𝑡=𝛼0+𝛽1𝑀𝐾𝑉 𝑖,𝑡−1+𝛽2𝑅𝑂𝐴 𝑖,𝑡−1+𝛽3𝐺𝑅𝑂𝑊𝑇𝐻 𝑖,𝑡−1+𝛽4𝐸𝑋𝐹𝐼𝑁𝐴𝐶𝑇 𝑖,𝑡−1+
𝛾1𝐶𝐹𝑉𝑜𝑙𝑎𝑡𝑖𝑙𝑖𝑡𝑦 𝑖,𝑡−1+Year Fixed Effects +𝜀𝑖,𝑡
residual( NGAGF) The residual coverage of unaffiliated analysts on group firms, defined as residuals from the
following model:
𝑁𝐺𝐴𝐺𝐹 𝑖,𝑡=𝛼0+𝛽1𝑀𝐾𝑉 𝑖,𝑡−1+𝛽2𝑅𝑂𝐴 𝑖,𝑡−1+𝛽3𝐺𝑅𝑂𝑊𝑇𝐻 𝑖,𝑡−1+𝛽4𝐸𝑋𝐹𝐼𝑁𝐴𝐶𝑇 𝑖,𝑡−1+
𝛾1𝐶𝐹𝑉𝑜𝑙𝑎𝑡𝑖𝑙𝑖𝑡𝑦 𝑖,𝑡−1+Year Fixed Effects +𝜀𝑖,𝑡
residual( NGANGF) The residual coverage of unaffiliated analysts on nongroup firms, defined as residuals from the
following model:
𝑁𝐺𝐴𝑁𝐺𝐹 𝑖,𝑡=𝛼0+𝛽1𝑀𝐾𝑉 𝑖,𝑡−1+𝛽2𝑅𝑂𝐴 𝑖,𝑡−1+𝛽3𝐺𝑅𝑂𝑊𝑇𝐻 𝑖,𝑡−1+𝛽4𝐸𝑋𝐹𝐼𝑁𝐴𝐶𝑇 𝑖,𝑡−1+
𝛾1𝐶𝐹𝑉𝑜𝑙𝑎𝑡𝑖𝑙𝑖𝑡𝑦 𝑖,𝑡−1+Year Fixed Effects +𝜀𝑖,𝑡

Control Variables of Regression Model (1)
SALES_GROWTH The percentage change in sales revenue
ROA Return on assets, calculated by net income divided by the beginning balance of total assets
∆ROA The change in ROA
RETURN Total annual stock returns
SIZE_R The natural logari thm of the market value of equity at the beginning of the year
Quick Quick ratio, calculated by the sum of cash and cash equivalents, short -term investments, and
receivables divided by current liabilities

Sustainabili ty 2019 , 11, x FOR PEER REVIEW 12 of 22
∆Quick The change in the quick ratio
LEV Debt ra tio, calculated by l ong-term liabilities divided by the beginning balance of total assets
LOSSBIN1 Indicator variable that equals one if a firm ’s ROA in the prior year is in between -0.005 and 0 ,
and zero otherwise
LOSSBIN2 Indicator variable that equals one if a firm ’s ROA in the prior year is in between -0.01 and –
0.005, and zero otherwise
LOSSBIN3 Indicator variable that equals one if a firm ’s ROA in the prior year is in between -0.015 and –
0.01, and zero otherwise
LOSSBIN4 Indicator variable that eq uals one if a firm ’s ROA in the prior year is in between -0.02 and –
0.015 , and zero otherwise
LOSSBIN5 Indicator variable that equals one if a firm ’s ROA in the prior year is in between -0.025 and –
0.02, and zero otherwise

Control Variables of Regress ion Model (2)
MTB Market -to-book ratio , calculated from market value of equity divided by book value of equity
SIZE The natural log of the market value of equity .
DIVDUM Indicator variable that equals one if a firm pays a dividend, and zero otherwise
STD_CFO The s tandard deviation of cash flows from operations over the recent 5 years
STD_SALE The s tandard deviation of sales revenue from operations over the recent 5 years
TANGIBLE The ratio of long -term assets to the beginning balance of total assets
LOSS Indicator variable that equals one if a firm has a net loss , and zero otherwise
INSTI The percentage of shares owned by institutional investors
STD_NET_HIRE The standard deviation of net hiring

Variables Used in Additional Tests
High (Low) Fund A firm is considered to have a high (low) level of inside fund when its cash and cash equivalents
scaled by total assets is above (below) the yearly median.
MKV Market value
GROWTH The g rowth rate of total assets
EXFINACT Net cash proceeds from externa l financing scaled by total assets
CFV olatility The st andard deviation of cash flow scaled by last year ’s total assets

415
416
Referenc es 417
1. 1. Brown, L.D.; Call, A.C.; Clement, M.B.; Sharp, N.Y. Inside the “black box” of sell‐side financial analysts. 418
Journal of Accounting Research 2015, 53, 1, 1-47. 419
2. 2. Graham, J.R.; Harvey, C.R.; Rajgopal, S. The economic implications of corporate financial reporting. 420
Journal of Accounting and Economics 2005, 40, 1 -3, 3-73. 421
3. 3. Yu, F.F. Analyst coverage and earnings management. Journal of Financial Economics 2008, 88,2, 245 -271. 422
4. 4. Chen, T.; Harford, J.; Lin, C. Do analysts matter for governance? Evidence from natural experiments. 423
Journal of Financial Economics 2015, 115, 2, 383 -410. 424

Sustainabili ty 2019 , 11, x FOR PEER REVIEW 13 of 22
5. 5. Wells, J. Retail labor cuts are hurting in -store experience, analysts say. 2017, June 2. Available online: 425
https://www.grocerydive.com/news/grocery –retail -labor -cuts-are-hurting -in-store -experience -analysts – 426
say/535045/ (accessed May 2, 2019). 427
6. 6. Finkle, J. Cisco braces for biggest layoffs in its history. 2011, May 13. Available online: 428
https://www.reuters.com/article/us -cisco/cisco -braces -for-biggest -layoffs -in-its-history – 429
idUSTRE74A78K20110513 (accessed May 2, 2019). 430
7. 7. Huyghebaert, N.; Xu, W. Bias in the post -IPO earnings forecasts of affiliated analysts: Evidence from a 431
Chin ese natural experiment. Journal of Accounting and Economics 2016, 61, 2 -3, 486 -505. 432
8. 8. Pae, P. South Korea’s Chaebol. 2018, October 5. Available online: 433
https://www.washingtonpost.com/business/south -koreas -chaebol/2018/10/05/e27095ac -c871 -11e8 -9c0f- 434
2ffaf6 d422aa_story.html?noredirect=on&utm_term=.872e52026fe3 (accessed May 2, 2019). 435
9. 9. Lim, Y.; Jung, K. Conflict of interest or information sharing? Evidence from affiliated analyst performance 436
in Korea. Contemporary Accounting Research 2012, 29, 2, 505 -537. 437
10. 10. Lee, C. A company that is 'jumping' to an analyst. 2016, April 19. Available online: 438
https://translate.google.com/translate?hl=en&sl=ko&u=http://biz.chosun.com/site/data/html_dir/2016/04/1 439
8/2016041802246.html&prev=search (accessed May 2, 2019). 440
11. 11. Kwa k, B.; Mo, K. Group -Affiliated Analysts’ Strategic Forecasts During a Year: Evidence from Korea. 441
Emerging Markets Finance and Trade 2019, 55, 1, 59 -77. 442
12. 12. Mantecon, T.; Altintig, Z.A. Chaebol -affiliated analysts: Conflicts of interest and market response s. 443
Journal of Banking & Finance 2012, 36,2, 584 -596. 444
13. 13. Jung, B.; Lee, W.J.; Weber, D. P. Financial reporting quality and labor investment efficiency. 445
Contemporary Accounting Research 2014, 31, 4, 1047 -1076. 446
14. 14. Pinnuck, M.; Lillis, A. M. Profits versus l osses: Does reporting an accounting loss act as a heuristic 447
trigger to exercise the abandonment option and divest employees?. The Accounting Review 2007, 82, 4, 448
1031 -1053. 449
15. 15. Biddle, G.C.; Hilary, G.; Verdi, R.S. How does financial reporting quality relat e to investment efficiency? 450
Journal of Accounting and Economics 2009, 48, 2 -3, 112 -131. 451
16. 16. Mo, K.; Lee, J. IFRS Adoption and the Choice Between Public and Private Debt: Evidence from South 452
Korea. Emerging Markets Finance and Trade 2018, 54, 11, 2533 -2556. 453
© 2019 by the authors. Submitted for possible open access publication under the ter ms
and conditions of the Creative Commons Attrib ution (CC BY) license
(http://creativecommon s.org/licenses/by/4.0/).
454