
How Does Social Media Influence Financial Fraud in Publicly Listed Companies?
Wei ZHANG, Yi LI
China Journal of Econometrics ›› 2024, Vol. 4 ›› Issue (4) : 899-923.
How Does Social Media Influence Financial Fraud in Publicly Listed Companies?
With the rise of social media, its impact on the financial transparency of publicly listed companies has received increasing attention. This study investigates how social media, particularly posting activity on East Money's stock message boards, affects the financial fraud behavior of listed companies. Utilizing data from East Money's stock message boards and a bivariate probit regression model, the study finds that the number of posts on the message boards is inversely related to the probability of fraud occurrence and positively related to the probability of fraud detection. This finding indicates that social media may play a dual role in both deterring financial fraud and uncovering it. To address endogeneity issues, the study employs an instrumental variable approach. Additionally, based on the "fraud triangle" theory, the paper proposes and validates two mechanisms through which message board posting activity reduces the likelihood of financial fraud: By decreasing potential opportunities for fraud and increasing the difficulty of rationalizing fraud. Heterogeneity analysis reveals that negative posts and posts by senior users are more effective in curbing financial fraud. This research not only enhances the understanding of how social media can function in corporate governance but also provides insights for regulatory authorities on leveraging social media for financial supervision.
social media / stock message boards / financial fraud / corporate governance {{custom_keyword}} /
表1 样本分布 |
2010 | 2011 | 2012 | 2013 | 2014 | 2015 | 2016 | 2017 | 2018 | 合计 | |
制造业 | 3 | 15 | 29 | 38 | 30 | 42 | 37 | 30 | 28 | 252 |
交通运输、仓储和邮政业 | 0 | 6 | 13 | 9 | 7 | 12 | 15 | 13 | 11 | 86 |
住宿和餐饮业 | 0 | 3 | 0 | 7 | 7 | 11 | 2 | 4 | 2 | 36 |
房地产业 | 0 | 1 | 2 | 4 | 7 | 2 | 3 | 1 | 4 | 24 |
金融业 | 2 | 3 | 3 | 1 | 4 | 4 | 1 | 1 | 0 | 19 |
建筑业 | 0 | 2 | 0 | 1 | 1 | 1 | 3 | 6 | 5 | 19 |
电力 | 1 | 2 | 3 | 2 | 2 | 2 | 4 | 1 | 1 | 18 |
采矿业 | 2 | 2 | 1 | 2 | 1 | 3 | 3 | 0 | 2 | 16 |
农、林、牧、渔业 | 0 | 1 | 2 | 1 | 2 | 1 | 3 | 1 | 0 | 11 |
批发和零售业 | 0 | 1 | 1 | 2 | 2 | 2 | 0 | 0 | 1 | 9 |
信息传输、软件和信息技术服务业 | 0 | 1 | 0 | 0 | 1 | 0 | 2 | 1 | 1 | 6 |
科学研究和技术服务业 | 2 | 0 | 0 | 1 | 0 | 0 | 0 | 0 | 1 | 4 |
租赁和商务服务业 | 0 | 0 | 1 | 0 | 0 | 0 | 0 | 0 | 0 | 1 |
合计 | 10 | 37 | 55 | 68 | 64 | 80 | 73 | 58 | 56 | 501 |
表2 主要变量描述性统计 |
均值 | 标准差 | 5%分位数 | 中值 | 95%分位数 | 造假样本 | 非造假样本 | 差别 | ||
POST | 8.592 | 0.922 | 7.308 | 8.602 | 9.984 | 8.559 | 8.624 | ||
CEO | 0.286 | 0.452 | 0.000 | 0.000 | 1.000 | 0.279 | 0.293 | ||
OWN | 0.068 | 0.141 | 0.000 | 0.000 | 0.428 | 0.064 | 0.072 | ||
INDEPEN | 0.375 | 0.060 | 0.333 | 0.364 | 0.500 | 0.375 | 0.376 | ||
STATE | 0.132 | 0.338 | 0.000 | 0.000 | 1.000 | 0.124 | 0.140 | ||
FAUD | 0.014 | 0.117 | 0.000 | 0.000 | 0.000 | 0.016 | 0.012 | 0.004 | 0.680 |
FLIST | 0.013 | 0.113 | 0.000 | 0.000 | 0.000 | 0.016 | 0.010 | 0.006 | 1.403 |
ANALYST | 1.424 | 1.209 | 0.000 | 1.386 | 3.434 | 1.334 | 1.514 | ||
SIZE | 22.307 | 0.880 | 20.967 | 22.265 | 23.941 | 22.290 | 22.324 | ||
AGE | 2.282 | 0.714 | 1.099 | 2.398 | 3.178 | 2.286 | 2.277 | 0.009 | 0.947 |
ROA | 0.632 | 0.027 | 0.115 | 0.011 | |||||
TOBINQ | 2.366 | 3.866 | 0.000 | 1.634 | 5.799 | 2.223 | 2.509 | ||
GROWTH | 0.345 | 2.592 | 0.079 | 0.939 | 0.341 | 0.348 | |||
DERATIO | 1.763 | 8.021 | 0.100 | 0.788 | 5.168 | 1.535 | 1.991 | ||
R&D | 0.032 | 0.046 | 0.000 | 0.023 | 0.108 | 0.031 | 0.033 | ||
LITIG | 28.758 | 1.864 | 25.130 | 28.969 | 30.912 | 28.772 | 28.744 | 0.028 | 0.114 |
ABVOLAT | 0.070 | 0.034 | 0.036 | 0.060 | 0.139 | 0.070 | 0.070 | 0.000 | |
ABTURN | 0.550 | 0.571 | 0.000 | 0.009 | |||||
CEOTURN | 0.257 | 0.437 | 0.000 | 0.000 | 1.000 | 0.263 | 0.251 | 0.012 | 0.430 |
表3 回归结果 |
造假 | 检测|造假 | 造假 | 检测|造假 | |||
POST | 0.399*** | TOBINQ | ||||
( | (3.677) | ( | ||||
CEO | 0.020 | GROWTH | 0.029 | |||
(0.200) | ( | (0.634) | ||||
OWN | DERATIO | |||||
( | ( | ( | ||||
INDEPEN | 0.764 | R&D | ||||
(1.570) | ( | |||||
STATE | 0.307* | LITIG | ||||
(1.813) | ( | ( | ||||
FAUD | 0.177 | ABVOLAT | 1.147* | |||
(0.491) | ( | (1.915) | ||||
FLIST | 0.548 | 0.053 | ABTURN | 0.186* | ||
(1.329) | (0.131) | (1.943) | ||||
ANALYST | 0.184*** | CEOTURN | 0.606*** | |||
( | (3.097) | (3.993) | ||||
SIZE | 0.254 | 截距项 | 2.033 | 3.676 | ||
( | (0.155) | (1.458) | (0.726) | |||
AGE | 0.080 | 年份固定效应 | Yes | Yes | ||
(0.778) | ( | 行业固定效应 | Yes | Yes | ||
ROA | 0.030 | Log likelihood | ||||
( | (0.333) |
注: *、**、***分别表示10%、5%和1%的显著性水平.下表同. |
表4 工具变量法回归结果 |
第一阶段回归 | 第二阶段回归 | ||||
(1) | (2) | (3) | (4) | ||
POST | POST | 造假 | 检测|造假 | ||
金融结构变量 | |||||
POST_P | 0.479*** | ||||
( | (3.486) | ||||
ADVER | 0.116*** | 0.123*** | |||
(3.924) | (4.136) | ||||
CEO | 0.004 | ||||
( | ( | (0.040) | ( | ||
OWN | 0.247 | 0.245 | |||
(0.978) | (0.973) | ( | ( | ||
INDEPEN | 1.413*** | 0.744 | |||
(2.733) | (1.101) | ||||
STATE | 0.043 | 0.056 | 0.194 | ||
(0.465) | (0.596) | (1.176) | ( | ||
FAUD | 0.183 | ||||
( | ( | (0.509) | ( | ||
FLIST | 0.037 | 0.543 | 0.015 | ||
( | (0.136) | (1.220) | (0.032) | ||
ANALYST | 0.006 | 0.003 | 0.201*** | ||
(0.233) | (0.114) | ( | (2.884) | ||
SIZE | 0.162*** | 3.346*** | 0.366 | ||
(4.401) | (3.987) | ( | (0.207) | ||
AGE | 0.163** | 0.201*** | 0.073 | ||
(2.220) | (2.707) | (0.698) | ( | ||
ROA | 0.051 | ||||
( | ( | ( | (0.567) | ||
TOBINQ | 0.004 | ||||
(0.534) | ( | ||||
GROWTH | 0.000 | 0.012 | |||
(0.001) | (0.271) | ||||
DERATIO | |||||
( | ( | ||||
R&D | 2.028*** | ||||
(2.937) | ( | ||||
LITIG | |||||
( | ( | ||||
ABVOLAT | 2.652*** | 0.653 | |||
(2.882) | (0.584) | ||||
ABTURN | 0.159 | ||||
( | (1.399) | ||||
CEOTURN | 0.054 | 0.721*** | |||
(0.770) | (3.170) | ||||
截距项 | 5.233*** | 1.958 | 2.755 | ||
(5.810) | ( | (1.439) | (0.532) | ||
年份固定效应 | Yes | Yes | Yes | Yes | |
行业固定效应 | Yes | Yes | Yes | Yes | |
调整后的 | 10.00% | 10.92% | |||
15.401 | 17.103 | ||||
Log likelihood | |||||
样本量 | 1002 | 1002 | 1002 | 1002 |
表5 异常发帖量 |
造假 | 检测|造假 | |
RPOST | 0.727*** | |
( | (4.148) | |
截距项 | 1.064 | 7.385*** |
(0.819) | (2.615) | |
控制变量 | Yes | Yes |
年份固定效应 | Yes | Yes |
行业固定效应 | Yes | Yes |
Log likelihood | ||
样本量 | 1002 | 1002 |
表6 剔除噪音 |
造假 | 检测|造假 | ||
Panel A: 有回复的帖子 | |||
POST | 0.103* | ||
( | ( | ||
截距项 | 0.949 | 4.859 | |
(0.606) | (1.124) | ||
控制变量 | Yes | Yes | |
年份固定效应 | Yes | Yes | |
行业固定效应 | Yes | Yes | |
Log likelihood | |||
样本量 | 1002 | 1002 | |
Panel B: 包含10个字以上的帖子 | |||
POST | 0.173** | ||
( | ( | ||
截距项 | 0.361 | 7.085 | |
(0.247) | (1.643) | ||
控制变量 | Yes | Yes | |
年份固定效应 | Yes | Yes | |
行业固定效应 | Yes | Yes | |
Log likelihood | |||
样本量 | 1002 | 1002 |
表7 社交媒体和传统媒体对财务造假影响的比较 |
造假 | 检测|造假 | 造假 | 检测|造假 | |||
POST | 0.182** | ROA | ||||
( | (2.546) | ( | ( | |||
NEWS | TOBINQ | |||||
( | ( | |||||
POST*NEWS | 0.215** | GROWTH | 0.032 | |||
(2.082) | (0.738) | |||||
CEO | 0.207 | DERATIO | ||||
(1.635) | ( | ( | ||||
OWN | 0.887 | R&D | ||||
( | (1.554) | ( | ||||
INDEPEN | 0.246 | LITIG | ||||
(0.468) | ( | |||||
STATE | 0.185 | ABVOLAT | 0.627 | |||
(1.033) | ( | (0.502) | ||||
FAUD | 0.717 | ABTURN | ||||
( | (0.949) | ( | ||||
FLIST | 0.691 | 0.306 | CEOTURN | 0.359*** | ||
(1.252) | (0.618) | (2.691) | ||||
ANALYST | 0.152** | 截距项 | 3.644** | 5.047 | ||
( | (2.545) | (2.055) | (1.058) | |||
SIZE | 年份固定效应 | Yes | Yes | |||
( | ( | 行业固定效应 | Yes | Yes | ||
AGE | 0.098 | Log likelihood | ||||
(0.889) | ( | |||||
样本量 | 1002 | 1002 |
表8 潜在机会减少路径检验结果 |
(1) | (2) | (3) | (4) | |
SYNCH | SYNCH | CORRE | CORRE | |
POST | ||||
( | ( | ( | ( | |
控制变量 | No | Yes | No | Yes |
年份固定效应 | Yes | Yes | Yes | Yes |
行业固定效应 | Yes | Yes | Yes | Yes |
样本量 | 21274 | 21274 | 21274 | 21274 |
调整后的 | 42.87% | 45.63% | 47.95% | 51.98% |
表9 合理化难度增加路径检验结果 |
(1) | (2) | (3) | (4) | |
CAR | CAR | CAR | CAR | |
POST | ||||
( | ( | ( | ( | |
控制变量 | No | Yes | No | Yes |
年份固定效应 | Yes | Yes | Yes | Yes |
行业固定效应 | Yes | Yes | Yes | Yes |
样本量 | 501 | 501 | 501 | 501 |
调整后的 | 2.07% | 1.15% | 2.62% | 2.35% |
表10 帖子情绪的影响 |
(1) | (2) | (3) | (4) | (5) | (6) | |
造假 | 检测|造假 | 造假 | 检测|造假 | 造假 | 检测|造假 | |
PPOST | 0.232*** | 0.106* | 0.705*** | 0.031 | ||
(3.068) | (1.657) | (2.778) | (0.179) | |||
NPOST | 0.150** | 0.085 | ||||
( | (2.301) | ( | (0.532) | |||
控制变量 | Yes | Yes | Yes | Yes | Yes | Yes |
年份固定效应 | Yes | Yes | Yes | Yes | Yes | Yes |
行业固定效应 | Yes | Yes | Yes | Yes | Yes | Yes |
Log likelihood | ||||||
样本量 | 1002 | 1002 | 1002 | 1002 | 1002 | 1002 |
表11 用户等级的影响 |
(1) | (2) | (3) | (4) | (5) | (6) | |
造假 | 检测|造假 | 造假 | 检测|造假 | 造假 | 检测|造假 | |
JPOST | 0.182*** | 0.101 | ||||
( | (2.592) | (0.652) | ( | |||
SPOST | 0.123*** | 0.191*** | ||||
( | (3.437) | ( | (4.037) | |||
控制变量 | Yes | Yes | Yes | Yes | Yes | Yes |
年份固定效应 | Yes | Yes | Yes | Yes | Yes | Yes |
行业固定效应 | Yes | Yes | Yes | Yes | Yes | Yes |
Log likelihood | ||||||
样本量 | 1002 | 1002 | 1002 | 1002 | 1002 | 1002 |
表12 其他类型的造假 |
造假 | 检测|造假 | |
POST | 0.297*** | |
( | (4.767) | |
控制变量 | Yes | Yes |
年份固定效应 | Yes | Yes |
行业固定效应 | Yes | Yes |
Log likelihood | ||
样本量 | 5962 | 5962 |
表13 变量定义 |
Panel A: 在双变量probit模型中出现的变量 | |
POST | 公司的一年之内的股吧上帖子数目加1再取自然对数 |
CEO | 虚拟变量, 公司的CEO和董事长是一个人时取1, 否则取0 |
OWN | 高管持有多少比例的公司股份 |
INDEPEN | 董事会中的独立董事数目占比 |
STATE | 虚拟变量, 当公司是国有企业时取1, 否则取0 |
FAUD | 虚拟变量, 当公司由国外审计公司(也就是所谓的四大会计师事务所)审计时取1, 否则取0 |
FLIST | 虚拟变量, 当公司在海外市场发行过股票时取1, 否则取0 |
ANALYST | 一年之内发布过针对公司报告的分析师人数加1再取自然对数 |
SIZE | 公司总资产的自然对数值 |
AGE | 公司上市的年份数 |
ROA | 公司的税后净利润除以总资产 |
TOBINQ | 公司股票的市场价值与资本重置成本的比率 |
GROWTH | 公司在某一年的销售额减去过去一年的销售额再除以过去一年的销售额 |
DERATIO | 公司的总负债除以所有者权益 |
R&D | 公司研发费用占总销售额的比率 |
LIGIT | 一年之内公司所处的行业内所有被起诉的公司市值之和的自然对数值 |
ABVOLAT | 一年之内公司股票周度收益的标准差 |
ABTURN | 一年之内公司股票月度换手率减去平均值之后的自然对数值 |
CEOTURN | 虚拟变量, 当公司在一年之内更换过一次以上的CEO时取1, 否则取0 |
Panel B: 其他变量 | |
NEWS | 一年之内针对公司的媒体报道数目加1之后取自然对数值 |
ADVER | 公司年度广告费用加1之后取自然对数值 |
POST_P | 工具变量法中, 利用第一阶段回归得到的预测的POST值 |
CAR (0, 5) | 在财务造假被披露之后的0 |
CAR (0, 10) | 在财务造假被披露之后的0 |
SYNCH | 从(一年的)市场模型回归得到的调整后的 |
CORRE | 一年之内公司股票周度收益和市场周度收益之间的皮尔森系数 |
PPOST | 公司股吧中一年之内情感正面帖子的数目加1之后取自然对数 |
NPOST | 公司股吧中一年之内情感负面帖子的数目加1之后取自然对数 |
SPOST | 公司股吧中一年之内有等级是6 |
JPOST | 公司股吧中一年之内有等级是1 |
AGROW | 公司某一年的资产总额除以过去一年的资产总额 |
PAST | 公司过去一年的累计收益率 |
RETAIL | 1减去机构投资者持有的公司流通股的比率 |
BAIDU | 公司股票代码在一年之内的百度搜索量 |
RPOST | 用POST对一系列公司变量(包括SIZE、ROA、TOBINQ、AGROW、ANALYST、NEWS、PAST、RETAIL和BAIDU)进行回归得到的残差 |
陈运森, 袁薇, 李哲, 监管型小股东行权的有效性研究: 基于投服中心的经验证据[J]. 管理世界, 2021, 37 (6): 142- 158.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
郭峰, 吕晓亮, 林致远, 龚志强, 池鱼之殃: 上市公司社交媒体联结与股价溢出效应——基于中国监管处罚的事件研究[J]. 管理科学学报, 2023, 26 (4): 111- 131.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
郝媛媛, 叶强, 李一军, 基于影评数据的在线评论有用性影响因素研究[J]. 管理科学学报, 2010, 13 (8): 78- 88.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
卢锐, 张亚楠, 蔡贵龙, 社交媒体、公司传闻与股价冲击——来自东方财富股吧论坛的经验证据[J]. 会计研究, 2023, (4): 59- 73.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
卢珊, 王惠文, 赵吉昌, 基于分步降维框架的股吧短文本数据关键信息抽取与个股收益预测[J]. 计量经济学报, 2023, 3 (3): 707- 721.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
汪昌云, 孙艳梅, 代理冲突、公司治理和上市公司财务欺诈的研究[J]. 管理世界, 2010, (7): 130- 143.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
徐黄凯, "安然事件"引发的法律风暴——萨班斯法案的诞生[J]. 法制与社会, 2011, (34): 53- 54.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
姚加权, 冯绪, 王赞钧, 纪荣嵘, 张维, 语调、情绪及市场影响: 基于金融情绪词典[J]. 管理科学学报, 2021, 24 (5): 26- 46.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
张维, 李奕, 王鹏飞, 社交媒体与资本市场研究进展与展望: 基于文献计量的方法[J]. 计量经济学报, 2022, 2 (1): 32- 57.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
张新民, 陈德球, 移动互联网时代企业商业模式、价值共创与治理风险——基于瑞幸咖啡财务造假的案例分析[J]. 管理世界, 2020, 36 (5): 74- 86.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
郑建东, 吕晓亮, 吕斌, 郭峰, 社交媒体平台信息交互与资本市场定价效率——基于股吧论坛亿级大数据的证据[J]. 数量经济技术经济研究, 2022, 39 (11): 91- 112.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
Wang E, Wong T, Zhang T, Hsu C, (2019). Do Chinese Social Media Delineate the Optimistic Bias of Traditional Media?[R]. Working Paper.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_ref.label}} |
{{custom_citation.content}}
{{custom_citation.annotation}}
|
/
〈 |
|
〉 |