189targeted firms that is not captured by these estimates. For example, the effects of the support programs on employment growth and demand for highly educated employees can, say, take off after some time. This means that significant results for individual years might be wiped out when aggregating the results over time. As a robustness check, we therefore also estimate yearly post-treatment effects.
The results from our year-by-year post-treatment analysis are presented in Fig. 2 (labor demand) and Fig. 3 (demand for highly educated employees). We choose to present the results from our preferred model specification, i.e., our DiD-estimations using a control group of matched firms. All point estimates and their corresponding 95% confidence intervals are displayed up to 5 years after the support programs have ended. We find no tendencies of a positive post-treatment effect on demand for labor (Fig. 2) and, if anything, the results are on the negative side. Thus, our year-overyear analysis confirms the finding that R&D grants had no post-treatment effects on the targeted SMEs. However, when investigating the relative demand for skilled labor, the estimate is positive and significant at the 5% level during the first
5Results are available on request.
190S.-O. Daunfeldt et al.
Note: *, **, ***, corresponds to levels of significance of 10, 5, 1%, with robust standard errors in parenthesis. Estimation 2 and 5 weighted with CEM-matched 9.6e-09(1.2e-07) 4.4e-05(2.3e-04) 0.001**(5.5e-04) 9.3e-06**(4.4e-06)
0.006***(4.6e-04)
0.057**(0.028)
5.4e-08(1.5e-07)
0.016***(0.003)
0.006**(0.003)
Treated vs. all
0.01 (6)
1.7e-07(2.0e-07) 8.4e-05(4.4.e05) 1.1e-07(1.5e-07)
0.0174**(0.007)
0.002*(0.001)
0.005***(0.001)
Treated vs. ctrl. 0.041(0.032) 0.004(0.003)
3.3e-05**
(1.5e-05) 0.01 (5)
1.9e-08(8.8e-08) 1.7e-07(1.3e-07) Interaction models
0.076***(0.027)
5.4e-05*(2.9e-05) 0.0039(0.004) 0.007*(0.004)
0.008***(0.003)
0.0037(0.006) 0.003(0.019) Treated only
0.02 (4)
1.4e-06(7.2e-08)***
0.003***(6.9e-04)
0.0005*(2.7e-04) 1.4e-04**(6.2e-06) 0.007*(7.8e-04)
0.010**(0.004)
Treated vs. all 0.009(0.006) Table 5 Estimation results, relative demand for high human capital workers 0.02 (3)
0.001*(4.1e-04) 2.4e-05(1.6e-05) 1.1e-05(8.1e-06)
0.006***(0.001)
Treated vs. ctrl. 0.002(0.001) 0.014*(0.007) weights. Firm- and period fixed effects included in all models 0.009(0.007)
0.03 (2)
0.016***(0.006)
1.4e-05(1.4e-05) 2.1e-05(1.4e-05) 0.001(0.075) Basic models 0.002(0.007) 0.005(0.009) 0.004(0.005) Treated only
0.07 (1)
ln(Y)t*(Post-support)t ln(wage premium)t ln(Y)t*(Grant/Y)t
Δ(R&D-int.)t-1
(Post-support)t R2-within (Grant)t (Profit)t
ln(K) t
ln(Y)t
191Do Targeted R&D Grants toward SMEs Increase Employment and Demand for High. . .
Marginal effect 95% CI
0.05
0
t+1
t+2
t+3
t+4
t+5
-0.05
-0.1
-0.15
-0.2
Fig. 2 Post-treatment effects of R&D grant on demand for labor, year-by-year, DiD-estimations
0.05
Marginal effect 95% CI
0.04
0.03
0.02
0.01
0
t+1
t+2
t+3
t+4
t+5
-0.01
-0.02
-0.03
Fig. 3 Post-treatment effects of R&D grant on demand for high human capital labor, year-by-year, DiD-estimations
post-treatment year and remains positive at the 10% significance level during the following year. There are thus some indications that the targeted R&D grants under study increased relative demand for high human capital labor during the first two posttreatment years. This implies that firms that receive targeted R&D grants increase their share of workers with higher education even though no effects on total S.-O. Daunfeldt et al.
192employment are detectable. However, the positive significance size ceases to exist after 3 years.
6
Discussion and Conclusions
Government support programs targeted toward innovative SMEs have become more common in recent years, and these programs are generally considered to be important in increasing innovative activities, and consequently employment growth, among growth-oriented SMEs (Bradley et al., 2021).
A challenge when evaluating these targeted R&D grant programs is how to estimate the counterfactual effect, i.e., the development of firms that were supported in the absence of receiving any government R&D grant. SMEs are not randomly selected by the programs; rather, R&D grants are often awarded to the most promising growth-oriented firms based on a combination of criteria. Hence, assessments might conclude that government support programs have been highly effective in increasing firms’ labor demand, even though targeted SMEs would have increased their number of employees and workers with higher education regardless of whether they received the R&D grant or not. This selection problem is often handled using a matching technique, thereby comparing firms that received support with similar firms that did not receive any targeted R&D grants. We rely on Coarsened Exact Matching to investigate the effects of two growth-oriented support programs in Sweden targeted toward innovative SMEs, making it possible to provide a more robust approach to matching. Our analyses are made possible due to access to a unique micro database on government firm support programs, compiled by the Swedish Government Agency for Growth Policy Analysis. This database alleviates the previous data acces-based concerns by finding appropriate matching firms. The most striking result of our analyses is the absence of statistically significant effects. We find no robust evidence that the government support programs had any positive and statistically significant effects on the number of employees brought into these growth-oriented SMEs. Additionally, there is not any robust evidence of an impact of the grants on the skill composition of the labor force.
The lack of statistically significant findings is troublesome considering that government support programs require a positive impact to cover the administrative costs that are associated with these programs. When the expected return of engaging in nonproductive entrepreneurship is high, entrepreneurs might also use time and resources to apply for government firm support programs instead of developing their businesses (Baumol, 1990). Firm support programs can thus crowd out more productive investments.