Title: “The impact of R&D expenditure on technology spillover and industrial growth”
Research Questions
How has the expenditures on R&D impacted technological spillover between the sectors and geographical areas?
Are there any margin of differences between the expenditures on R&D in Europe in comparison to other parts of the world?
Has the margin of difference suggest a different level of technological spillover between geographical areas?
Can we conclude that the spillover in one geographical area is better than the other?
Methodology
Data Collection
Importing the data set
The data set consist of 1629 observations with 7 variables.
Checking for the structure of the data
## Classes 'tbl_df', 'tbl' and 'data.frame': 1629 obs. of 7 variables:
## $ year : num 1983 1983 1983 1983 1983 ...
## $ fi : num 1 2 3 4 5 6 7 8 9 10 ...
## $ sector: num 4 5 2 2 11 5 1 11 3 2 ...
## $ geo : num 3 3 3 1 4 1 3 3 3 3 ...
## $ patent: num 18 4 29 45 1 0 1 0 0 47 ...
## $ rdexp : num 5.29 4.31 3.76 5.87 4.21 ...
## $ spil : num 8.98 10.42 9.65 9.63 8.7 ...Extracts of the first 5 observations
## # A tibble: 5 x 7
## year fi sector geo patent rdexp spil
## <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 1983 1 4 3 18 5.29 8.98
## 2 1983 2 5 3 4 4.31 10.4
## 3 1983 3 2 3 29 3.76 9.65
## 4 1983 4 2 1 45 5.87 9.63
## 5 1983 5 11 4 1 4.21 8.70Checking for missing number
## [1] 0This shows that there is no missing values
Note: fi , sector and geo are categorical variables but they encoded into numeric for the purpose of analysis.
Summary of the data
## year fi sector geo
## Min. :1983 Min. : 1 Min. : 1.00 Min. :1.000
## 1st Qu.:1985 1st Qu.: 46 1st Qu.: 3.00 1st Qu.:3.000
## Median :1987 Median : 91 Median : 5.00 Median :3.000
## Mean :1987 Mean : 91 Mean : 6.32 Mean :2.641
## 3rd Qu.:1989 3rd Qu.:136 3rd Qu.: 9.00 3rd Qu.:3.000
## Max. :1991 Max. :181 Max. :15.00 Max. :4.000
## patent rdexp spil
## Min. : 0.00 Min. :0.8651 Min. : 6.825
## 1st Qu.: 3.00 1st Qu.:4.1617 1st Qu.: 8.864
## Median : 18.00 Median :5.0752 Median : 9.619
## Mean : 60.79 Mean :5.2013 Mean : 9.399
## 3rd Qu.: 57.00 3rd Qu.:6.0911 3rd Qu.: 9.980
## Max. :925.00 Max. :8.7000 Max. :10.759## vars n mean sd median trimmed mad min max
## year 1 1629 1987.00 2.58 1987.00 1987.00 2.97 1983.00 1991.00
## fi 2 1629 91.00 52.27 91.00 91.00 66.72 1.00 181.00
## sector 3 1629 6.32 4.23 5.00 5.89 4.45 1.00 15.00
## geo 4 1629 2.64 0.73 3.00 2.79 0.00 1.00 4.00
## patent 5 1629 60.79 121.56 18.00 31.13 25.20 0.00 925.00
## rdexp 6 1629 5.20 1.26 5.08 5.12 1.43 0.87 8.70
## spil 7 1629 9.40 0.93 9.62 9.46 0.89 6.82 10.76
## range skew kurtosis se
## year 8.00 0.00 -1.23 0.06
## fi 180.00 0.00 -1.20 1.29
## sector 14.00 0.71 -0.65 0.10
## geo 3.00 -1.58 0.86 0.02
## patent 925.00 3.84 17.36 3.01
## rdexp 7.83 0.46 -0.46 0.03
## spil 3.93 -0.65 -0.29 0.02Exploratory Data Analysis
Checking the distribution of the dataset
Boxplot: To check for outliers
Correlation Coefficients
From the correlation table, its obvious there is a strong relationship of about 55% between patent and rdexp which implies that there is a multicollinearity between the model if i will have to consider all the variables for model building. Rd exp on the other hand has 35% correlation coefficient with spil, the highest among other predictors.
Data in tables
##
## Attaching package: 'dplyr'## The following objects are masked from 'package:stats':
##
## filter, lag## The following objects are masked from 'package:base':
##
## intersect, setdiff, setequal, union## # A tibble: 4 x 3
## geo sum_rdexp sum_spil
## <dbl> <dbl> <dbl>
## 1 1 1460. 2378.
## 2 2 707. 1083.
## 3 3 6264. 11771.
## 4 4 41.3 78.8## # A tibble: 15 x 3
## sector sum_rdexp sum_spil
## <dbl> <dbl> <dbl>
## 1 1 573. 1105.
## 2 2 1295. 2477.
## 3 3 983. 1724.
## 4 4 628. 1080.
## 5 5 1514. 2764.
## 6 6 334. 638.
## 7 7 518. 787.
## 8 8 122. 210.
## 9 9 511. 1038.
## 10 10 391. 826.
## 11 11 124. 236.
## 12 12 342. 599.
## 13 13 173. 324.
## 14 14 67.5 173.
## 15 15 897. 1331.## # A tibble: 9 x 3
## year sum_rdexp sum_spil
## <dbl> <dbl> <dbl>
## 1 1983 892. 1664.
## 2 1984 912. 1678.
## 3 1985 929. 1692.
## 4 1986 937. 1696.
## 5 1987 946. 1701.
## 6 1988 957. 1711.
## 7 1989 962. 1715.
## 8 1990 969. 1725.
## 9 1991 969. 1731.Bar Chart
How has the expenditures on R&D impacted technological spillover between the sectors and geographical areas?
It is evident from the chart that the more a sector spend on r/d the highest the technological spillover effect on the sector. Take for an instance, Sector_1 with about 600m pounds expenditure on r/d further improve the sector as a result technological spillover by about 1200m pounds. This further improve the sector and the industry.
Scatter Plot of R/D exp and Spillover.
##
## Attaching package: 'ggplot2'## The following objects are masked from 'package:psych':
##
## %+%, alpha
Scatter Plot of R/D exp and Spillover over the years under review.
library(ggplot2)
#Scatter_plot
scatter <- ggplot(Yr, aes(year, sum_spil))
scatter + geom_point() + labs(x ="Year", y = "sum_spil") +
geom_smooth(method = "lm", colour = "Red", alpha = 0.1, fill = "Blue")
scatter <- ggplot(Yr, aes(year, sum_rdexp))
scatter + geom_point() + labs(x ="Year", y = "sum_rdexp") +
geom_smooth(method = "lm", colour = "Red", alpha = 0.1, fill = "Blue")
Linear Regression
##
## Call:
## lm(formula = spil ~ ., data = data)
##
## Residuals:
## Min 1Q Median 3Q Max
## -2.48192 -0.43267 0.05886 0.46771 1.83142
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -6.174e+01 1.467e+01 -4.209 2.71e-05 ***
## year 3.586e-02 7.389e-03 4.853 1.33e-06 ***
## fi -4.228e-04 3.649e-04 -1.159 0.247
## sector -9.342e-02 4.621e-03 -20.214 < 2e-16 ***
## geo -2.777e-01 2.858e-02 -9.714 < 2e-16 ***
## patent -1.335e-03 1.934e-04 -6.904 7.24e-12 ***
## rdexp 2.568e-01 1.853e-02 13.855 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.7639 on 1622 degrees of freedom
## Multiple R-squared: 0.3206, Adjusted R-squared: 0.3181
## F-statistic: 127.6 on 6 and 1622 DF, p-value: < 2.2e-16
New Model
regressor = lm(formula = spil ~ year + sector + geo + patent + rdexp,
data = data)
summary(regressor)##
## Call:
## lm(formula = spil ~ year + sector + geo + patent + rdexp, data = data)
##
## Residuals:
## Min 1Q Median 3Q Max
## -2.46087 -0.43013 0.06179 0.47028 1.83642
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -6.160e+01 1.467e+01 -4.199 2.83e-05 ***
## year 3.576e-02 7.389e-03 4.839 1.43e-06 ***
## sector -9.343e-02 4.622e-03 -20.214 < 2e-16 ***
## geo -2.743e-01 2.844e-02 -9.645 < 2e-16 ***
## patent -1.331e-03 1.934e-04 -6.881 8.48e-12 ***
## rdexp 2.583e-01 1.849e-02 13.972 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.7639 on 1623 degrees of freedom
## Multiple R-squared: 0.32, Adjusted R-squared: 0.3179
## F-statistic: 152.8 on 5 and 1623 DF, p-value: < 2.2e-16plot(regressor)
This model is significant as p-value is less than 0.05 (p-value: < 2.2e-16). With Adjusted R-squared of roughly 31.8% (Adjusted R-squared: 0.3179) it means the model can explain only 31.8% of the variation in technological spillover while 68% of the variation in spillover is being explained by other factors outside the model. Each of the variables are significant because they have all got P values that is far less than 0.05.
Conclusion
- Based on this model, Expenditure on Research and Development is a major factor for the reasons for variation in technological spillover. This explains the fact that R/D has got a huge impact on spillover.