建立Fama-French 3 factor时间序列[R code]

是 否 +2 论坛币

k人 参与回答

经管之家送您一份

应届毕业生专属福利!

求职就业群

赵安豆老师微信：zhaoandou666

经管之家联合CDA

送您一个全额奖学金名额~ !

立即领取

感谢您参与论坛问题回答

经管之家送您两个论坛币！

+2 论坛币

参照姜国华对中国市场3因子的定义，建立了因子时间序列的数据。用到的数据全部来自国泰安，有：
    --TRD_Mnth
    --FS_Combas



主要是用dplyr包写的，概念上理解起来比较简单：

1. ## Fama French size and bm factor return
   
2. 
   
3. require(dplyr)
   
4. require(reshape2)
   
5. require(ggplot2)
   
6. 
   
7. setwd("/Users/wang/Documents/Undergraduate/ps1")
   
8. 
   
9. ## import dataset
   
10. # monthly stock trading file
    
11. msf1 <- data.table::fread("TRD_Mnth1.csv")   
    
12. msf2 <- data.table::fread("TRD_Mnth2.csv")
    
13. # banlance sheet
    
14. bs <- data.table::fread("FS_Combas.csv")
    
15. msf<-rbind(msf1, msf2)
    
16. str(msf)
    
17. 
    
18. ## alphanumeric transfomation
    
19. mon_encode <- data.frame(mon_name=month.abb,
    
20.                          mon=1:12,
    
21.                          stringsAsFactors = F)
    
22. 
    
23. msf <- msf %>%
    
24.     filter(Markettype %in% c(1L,4L,16L)) %>%
    
25.     mutate(yr = as.integer(substr(Trdmnt, 5,6)),  # year var
    
26.            mon_name = substr(Trdmnt, 1,3), # month var
    
27.            mcap = Msmvttl*1000   # market cap
    
28.            ) %>%
    
29.     select(Stkcd, yr, mon_name, ret=Mretwd, mcap)
    
30. 
    
31. msf <- msf[complete.cases(msf),] %>%  # delete missing values
    
32.     left_join(mon_encode) %>%
    
33.     mutate(yr=yr+1900L*(yr>=90)+2000L*(yr<90)) %>%
    
34.     select(Stkcd, yr, mon, ret, mcap)
    
35. 
    
36. ## check if key var is unique
    
37. if(nrow(msf)-nrow(distinct(msf, Stkcd, yr, mon))==0)
    
38. message('No dupkey')
    
39. 
    
40. ## merge b/s with mkt cap in DEC
    
41. bs <- bs %>%
    
42.     mutate(Accper = as.Date(Accper),
    
43.            yr = as.integer(substr(Accper,1,4)),
    
44.            mon = as.integer(substr(Accper,6,7))) %>%
    
45.     # only select yearly consolidated statement
    
46.     filter(Typrep == 'A', months(Accper, abbreviate=T) == '12') %>%
    
47.     select(Stkcd, yr, mon, Accper, be = A003000000)
    
48.    
    
49. bs <- bs[complete.cases(bs),]
    
50. if(nrow(bs)-nrow(distinct(bs, Stkcd, yr, mon))==0)
    
51.     message('No dupkey')
    
52. 
    
53. # book/market ratio
    
54. bm <- msf %>%
    
55.     inner_join(bs, by = c("Stkcd", "yr", "mon")) %>%
    
56.     mutate(bm = be/mcap) %>%
    
57.     select(Stkcd, yr, bm)
    
58. 
    
59. ## merge size and BM factor with msf(monthly stock trading file)
    
60. size <- msf %>% filter(mon == 4L) %>%
    
61.     select(Stkcd, yr, size = mcap) # use mcap at Apr. 30
    
62. 
    
63. ff <- msf %>% mutate(yr4bm = yr-1,
    
64.                       yr4size = ifelse(mon<=4, yr-1, yr)) %>%
    
65.     inner_join(bm, by=c("Stkcd"="Stkcd", "yr4bm"="yr")) %>%
    
66.     inner_join(size, by=c("Stkcd"="Stkcd", "yr4size"="yr")) %>%
    
67.     select(Stkcd, yr, mon, ret, bm, size)
    
68.    
    
69. ## each month, construct 2*3 ports based on size/bm
    
70. ff <- ff %>%
    
71.     group_by(yr, mon) %>%
    
72.     mutate(size_rank = ntile(size, 2)) %>%
    
73.     group_by(size_rank, add = T) %>%
    
74.     mutate(
    
75.         bm_rank = ifelse(
    
76.             percent_rank(bm) < 0.3, 1,
    
77.                   ifelse(percent_rank(bm) <0.7, 2, 3)
    
78.                    )
    
79.         )
    
80. 
    
81. ## calculate equal weighted port ret
    
82. port_ret <- ff %>%
    
83.     group_by(yr, mon, size_rank, bm_rank) %>%
    
84.     summarise(ret = mean(ret),
    
85.               obs = n())
    
86. # convert from long to wide to calculate ff factor
    
87. port_ret_l <- melt(port_ret,
    
88.                  id.vars = c("yr", "mon", "size_rank", "bm_rank"),
    
89.                  measure.vars = "ret")
    
90. port_ret_w <- dcast(port_ret_l, yr+mon ~ size_rank + bm_rank)
    
91. names(port_ret_w)[3:8] <- paste0("p", names(port_ret_w)[3:8])
    
92. 
    
93. ff <- port_ret_w %>%
    
94.     mutate(smb=(p1_1+p1_2+p1_3-p2_1-p2_2-p2_3)/3,
    
95.            hml=(p1_3+p2_3-p1_1-p2_1)/2) %>%
    
96.     select(yr, mon, smb, hml) %>%
    
97.     arrange(yr, mon)
    
98. 
    
99. ## plot size and bm premium
    
100. ff <- ff %>%
     
101.     mutate(date = as.Date(paste(yr, mon, '01', sep='-'))) %>%
     
102.     filter(date>=as.Date("2000-01-01")) %>%
     
103.     select(date, smb, hml)
     
104. 
     
105. ff <- melt(ff, id.vars = 'date',
     
106.            variable.name = 'ff.factor',
     
107.            value.name = 'ff.premium')
     
108. 
     
109. ggplot(data=ff, aes(x=date, y=ff.premium, colour=ff.factor)) + geom_line()  
     
110. ggsave("ffpremium.pdf")
     
111. 
     
112. ## house cleaning
     
113. rm(list=ls())

复制代码

扫码加我 拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

分享0 收藏6 回帖

关键词：FRENCH factor Facto fact 时间序列

这个相当是建立了规模和因子的市场月度的时序数列是吗？

鎏金铄石 发表于 2017-8-16 11:23
这个相当是建立了规模和因子的市场月度的时序数列是吗？

是的   

先马，慢慢看

谢谢分享，非常有用

楼主大度