Outsourcing of research and development and efficiency: a DEA non-parametric analysis of the contract research organisations industry

Figures & data

Figure 1. Evolution of selected variables over time (2012–2020).

Notes: Averages across DMUs. Average real turnover, assets and income in thousand dollars, deflated with the Consumer Index Price for G20. Source: own elaboration.

Table 1. Summary statistics, selected variables (2012–2020).

	N. obs.	Type of variable	Mean	Std. Dev.	Min	Max	Growth 2012–2020
Real turnover	275	output	313,131.91	729,817.22	1,607.45	4,305,538.5	76%
Employees	249	Input	2,021.309	4,516.834	15	24,310	114%
Real assets	249	Input	415,852.76	1,083,350.1	2,637.919	7,225,184.5	159%
Real income	270	output	14,901.538	61,726.613	−25,2171.16	344,370.16	186%

Note: The type of variable details if the variable is regarded as an input or output in the DEA estimations. Source: Orbis and own elaboration.

Table 2. Computation of efficiency, inputs and outputs.

Model	Inputs	Outputs	Sample size	Effective parametric sample size
1	# of employees Total assets	Turnover	249	249
2	# of employees Total assets	Turnover Net income	245	81
3	Total employee cost Total assets	Turnover	184	184
4	Total employee cost Total assets	Turnover Net income	184	64
5	Total employee cost** Total assets	Turnover Net income	184	64

Notes: **: Model with 1,000 bootstrap replications. Source: own elaboration.

Table 3. Efficiency scores, descriptive statistics.

Variables	Obs.	Mean	Std. Dev.	Min	Max	Median
Model 1
$\widehat{{\theta }^1}$	249	.783	.232	.17	1	.846
$\widehat{{\theta }_b^1}$	249	.665	.186	.134	.887	.732
$\widehat{{\theta }_L^1}$	249	.568	.156	.106	.791	.618
$\widehat{{\theta }_U^1}$	249	.771	.228	.167	.997	.834
Model 2
$\widehat{{\theta }^2}$	245	.803	.228	.168	1	.915
$\widehat{{\theta }_b^2}$	245	.686	.185	.14	.907	.776
$\widehat{{\theta }_L^2}$	245	.579	.152	.121	.828	.617
$\widehat{{\theta }_U^2}$	245	.794	.225	.165	.997	.910
Model 3
$\widehat{{\theta }^3}$	184	.757	.262	.167	1	.826
$\widehat{{\theta }_b^3}$	184	.624	.205	.131	.898	.707
$\widehat{{\theta }_L^3}$	184	.527	.173	.101	.827	.57
$\widehat{{\theta }_U^3}$	184	.743	.257	.165	.994	.817
Model 4
$\widehat{{\theta }^4}$	184	.779	.254	.167	1	.870
$\widehat{{\theta }_b^4}$	184	.637	.197	.133	.882	.718
$\widehat{{\theta }_L^4}$	184	.532	.163	.101	.792	.572
$\widehat{{\theta }_U^4}$	184	.764	.249	.164	.995	.857
Model 5
$\widehat{{\theta }^5}$	184	.779	.254	.167	1	.870
$\widehat{{\theta }_b^5}$	184	.638	.197	.132	.881	.72
$\widehat{{\theta }_L^5}$	184	.533	.163	.104	.798	.576
$\widehat{{\theta }_U^5}$	184	.765	.249	.164	.993	.858

Source: own elaboration

Table 4. Efficiency estimators, correlations.

A. Matrix of correlations, mean bootstrap efficiency estimators across models (Models 1–5).
Variables	Model 1	Model 2	Model 3	Model 4	Model 5
Model 1	1.000
Model 2	0.899	1.000
Model 3	0.726	0.714	1.000
Model 4	0.709	0.744	0.968	1.000
Model 5	0.711	0.745	0.969	0.999	1.000

Table

B. Matrix of correlations, mean efficiency estimators computed by Model 1
	$\widehat{{\theta }^1}$	$\widehat{{\theta }_b^1}$	$\widehat{{\theta }_L^1}$	$\widehat{{\theta }_U^1}$
$\widehat{{\theta }^1}$	1.000
$\widehat{{\theta }_b^1}$	0.986	1.000
$\widehat{{\theta }_L^1}$	0.954	0.987	1.000
$\widehat{{\theta }_U^1}$	0.999	0.987	0.955	1.000

Note: All correlations are significant at 99%. Source: own elaboration.

Figure 2. Average bootstrap efficiency over time, Model 1 (2012–2020).

Source: own elaboration.

Table 5. Average efficiency scores of best performers, Model 1 (2012–2020).

DMU	Obs.	Mean	St. Dev.	Min	Max
C. RECHERCHES BIOLOGIQUES	8	.817	.040	.737	.871
CMIC	9	.790	.088	.656	.887
GKM GESELLSCHAFT	4	.798	.106	.639	.865
ICON ESPAÑA	9	.755	.067	.662	.845
IMM RECHERCHE	5	.783	.046	.708	.831
MEDPACE HOLDINGS	5	.76	.128	.534	.837
OY MEDFILES	9	.810	.051	.732	.866
PPD AUSTRALIA	7	.82	.012	.805	.841
SOIKEN HOLDINGS	9	.77	.114	.496	.879
SYNEOS	9	.78	.048	.665	.823
TFS	7	.76	.091	.621	.834

Note: Top performers in bold. Source: own elaboration.

Figure 3. Efficiency over time of the best performers.

Note: bootstrap efficiency, computed according to Model 1. Source: own elaboration.

Figure 4. Persistence of efficiency.

Note: Efficiency computed according to Model 1. Source: own elaboration.

Table 6. Mean efficiency by size.

	Mean	Standard deviation	P value of test
Very big	0.676	0.214	0.536
Big	0.619	0.204	0.0416**
Small	0.629	0.155	0.01*
Very small	0.723	0.154	0***
Whole sample	0.665	0.186

Notes: *: significant at 90%. **: significant at 95%. ***: significant at 99%. Last column: p value of the Krishnamoorthy and Yu (2004) test of equality of means. Source: own elaboration.

Table

DMU	Obs.	Mean	St. dev.	Min	Max
ACLARIS	4	.494	.256	.159	.713
BIOCLINICA INC	1	.274	.	.274	.274
C. RECHERCHES BIOLOGIQUES	8	.817	.040	.737	.871
CHILTERN INT.	7	.615	.237	.259	.852
CMED	9	.456	.088	.364	.630
CMIC	9	.790	.088	.656	.887
COVANCE INC	2	.747	.037	.721	.774
CROMSOURCE S.R.L	9	.641	.131	.485	.854
CROWN CRO OY	3	.836	.014	.824	.852
GENSCRIPT	6	.283	.031	.230	.318
GEORGE CLINICAL	6	.680	.208	.383	.877
GKM GESELLSCHAFT	4	.798	.106	.639	.865
ICON UK	8	.382	.095	.208	.490
ICON ESPAÑA	9	.755	.067	.662	.845
IMM RECHERCHE	5	.783	.046	.708	.831
INVITES BIOCORE	8	.628	.132	.499	.857
IQVIA SPAIN	8	.754	.058	.691	.837
KCR S.A.	6	.618	.127	.479	.785
LABCORP SARL	8	.742	.120	.543	.846
LABCORP UK	8	.357	.167	.133	.712
LINICAL CO	9	.469	.078	.324	.568
MEDPACE HOLDINGS	5	.760	.128	.534	.837
NAMSA	5	.583	.114	.485	.772
NOVOTECH	2	.454	.143	.352	.555
OY MEDFILES	9	.810	.051	.732	.866
PIVOTAL SL	9	.65	.110	.494	.814
PPD AUSTRALIA	7	.82	.012	.805	.841
PPD BULGARIA	9	.603	.245	.198	.838
PPD GERMANY	8	.731	.231	.178	.859
PPD POLAND	5	.668	.099	.519	.739
PRA	8	.752	.156	.373	.857
QPS AUSTRIA	6	.753	.115	.557	.876
SCANTOX	5	.604	.095	.519	.753
SOIKEN HOLDINGS	9	.77	.114	.496	.879
SYNEOS	9	.78	.048	.665	.823
TFS	7	.76	.091	.621	.834

Source: own elaboration.

Krishnamoorthy, K., & Yu, J. (2004). Modified Nel and Van der Merwe test for the multivariate Behrens–Fisher problem. Statistics & Probability Letters, 66(2), 161–169. https://doi.org/10.1016/j.spl.2003.10.012

 Web of Science ®Google Scholar