半導體業的 FDI 考量與地緣政治風險的關係-以臺灣公司為例

library(sjPlot)

data_path <- "C:/Users/v106K/OneDrive/桌面/國政經參考資料/R data"

data <- read.csv(
  file.path(data_path, "FDI_GPR_simple_analysis.csv"),
  fileEncoding = "UTF-8-BOM",
  check.names = FALSE
)

data$company <- factor(data$company)
data$industry_position <- factor(data$industry_position)
data$invest <- as.numeric(data$invest)
data$global_gpr_lag1 <- as.numeric(data$global_gpr_lag1)
data$host_gpr_lag1 <- as.numeric(data$host_gpr_lag1)
data$host_gpr_high <- as.factor(data$host_gpr_high)

# H1：公司年度資料，避免公司年度重複
company_year_simple <- data[!duplicated(data[c("company", "year")]), ]

model_h1 <- lm(
  fdi_total_billion ~ global_gpr_lag1 + company,
  data = company_year_simple
)

summary(model_h1)


Call:
lm(formula = fdi_total_billion ~ global_gpr_lag1 + company, data = company_year_simple)

Residuals:
    Min      1Q  Median      3Q     Max 
-169.91  -65.51  -13.28   20.79  426.00 

Coefficients:
                Estimate Std. Error t value Pr(>|t|)  
(Intercept)     -156.148    119.278  -1.309    0.205  
global_gpr_lag1    1.430      1.014   1.409    0.174  
companyMTK        11.551     64.176   0.180    0.859  
companyTSM       133.846     64.176   2.086    0.050 .
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Residual standard error: 128.4 on 20 degrees of freedom
Multiple R-squared:  0.2682,    Adjusted R-squared:  0.1584 
F-statistic: 2.443 on 3 and 20 DF,  p-value: 0.09398

tab_model(
  model_h1,
  show.se = TRUE,
  show.r2 = TRUE,
  p.style = "stars",
  digits = 3,
  digits.p = 3,
  dv.labels = "H1：Global GPR 對 FDI 總額之影響"
)

	H1：Global GPR 對 FDI 總額之影響
Predictors	Estimates	std. Error	CI
(Intercept)	-156.148	119.278	-404.957 – 92.661
global gpr lag1	1.430	1.014	-0.686 – 3.545
company [MTK]	11.551	64.176	-122.318 – 145.420
company [TSM]	133.846	64.176	-0.023 – 267.715
Observations	24
R² / R² adjusted	0.268 / 0.158
* p<0.05 p<0.01 * p<0.001

# H2：卡方檢定
h2_data <- data[!is.na(data$host_gpr_lag1), ]

table_h2 <- table(h2_data$host_gpr_high, h2_data$invest)
table_h2

chisq.test(table_h2)


    Pearson's Chi-squared test with Yates' continuity correction

data:  table_h2
X-squared = 5.7436, df = 1, p-value = 0.01655

# H2：Logit 迴歸
model_h2 <- glm(
  invest ~ host_gpr_lag1 + company,
  data = h2_data,
  family = binomial
)

summary(model_h2)


Call:
glm(formula = invest ~ host_gpr_lag1 + company, family = binomial, 
    data = h2_data)

Coefficients:
              Estimate Std. Error z value Pr(>|z|)    
(Intercept)    -2.1217     0.3112  -6.818 9.23e-12 ***
host_gpr_lag1   0.7311     0.1940   3.770 0.000164 ***
companyMTK     -0.4295     0.4183  -1.027 0.304528    
companyTSM     -0.6448     0.4384  -1.471 0.141315    
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

(Dispersion parameter for binomial family taken to be 1)

    Null deviance: 235.10  on 311  degrees of freedom
Residual deviance: 219.72  on 308  degrees of freedom
AIC: 227.72

Number of Fisher Scoring iterations: 5

tab_model(
  model_h2,
  show.se = TRUE,
  show.r2 = TRUE,
  p.style = "stars",
  transform = "exp",
  digits = 3,
  digits.p = 3,
  dv.labels = "H2：Host GPR 對是否投資某國之影響"
)

	H2：Host GPR 對是否投資某國之影響
Predictors	Odds Ratios	std. Error	CI
(Intercept)	0.120 ^***	0.037	0.063 – 0.214
host gpr lag1	2.077 ^***	0.403	1.414 – 3.048
company [MTK]	0.651	0.272	0.281 – 1.467
company [TSM]	0.525	0.230	0.215 – 1.220
Observations	312
R² Tjur	0.056
* p<0.05 p<0.01 * p<0.001

# H3：卡方檢定
table_h3 <- table(data$industry_position, data$invest)
table_h3

          
             0   1
  IC設計   100  12
  封裝測試  95  17
  晶圓製造 102  10

chisq.test(table_h3)


    Pearson's Chi-squared test

data:  table_h3
X-squared = 2.2626, df = 2, p-value = 0.3226

# H3：簡單 Logit
model_h3 <- glm(
  invest ~ industry_position,
  data = data,
  family = binomial
)

summary(model_h3)


Call:
glm(formula = invest ~ industry_position, family = binomial, 
    data = data)

Coefficients:
                          Estimate Std. Error z value Pr(>|z|)    
(Intercept)                -2.1203     0.3055  -6.940 3.92e-12 ***
industry_position封裝測試   0.3996     0.4033   0.991    0.322    
industry_position晶圓製造  -0.2021     0.4507  -0.448    0.654    
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

(Dispersion parameter for binomial family taken to be 1)

    Null deviance: 241.26  on 335  degrees of freedom
Residual deviance: 239.05  on 333  degrees of freedom
AIC: 245.05

Number of Fisher Scoring iterations: 5

tab_model(
  model_h3,
  show.se = TRUE,
  show.r2 = TRUE,
  p.style = "stars",
  transform = "exp",
  digits = 3,
  digits.p = 3,
  dv.labels = "H3：產業鏈位置對是否投資某國之影響"
)

	H3：產業鏈位置對是否投資某國之影響
Predictors	Odds Ratios	std. Error	CI
(Intercept)	0.120 ^***	0.037	0.063 – 0.209
industry position [封裝測試]	1.491	0.601	0.681 – 3.358
industry position [晶圓製造]	0.817	0.368	0.331 – 1.978
Observations	336
R² Tjur	0.007
* p<0.05 p<0.01 * p<0.001

You can add options to executable code like this

[1] 4

The echo: false option disables the printing of code (only output is displayed).