Top

Empirical Economics

Published in:

11-05-2023

Public subsidies and innovation: a doubly robust machine learning approach leveraging deep neural networks

Authors: Kerda Varaku, Robin Sickles

Published in: Empirical Economics | Issue 6/2023

Activate our intelligent search to find suitable subject content or patents.

search-config

AI-assisted search

Off

Abstract

Economic growth is crucial to improve standards of living, prosperity, and welfare. R &D and knowledge spillovers can offset the diminishing returns to physical capital (machines and labor) and drive long-run growth. Market imperfections can bring R &D below the socially desired level; thus, many governments intervene to increase the stock of knowledge, and knowledge spillovers, via subsidies for R &D. We use European firm-level data to explore the effects of public subsidies on firms’ R &D input and output. Average treatment effects are estimated by controlling for both observable and unobserved heterogeneity. Possible endogeneity in subsidy assignment is addressed, and the local instrumental variable (LIV) curve is identified via double machine learning methods. Results indicate that public subsidies increase both R &D intensity and output with more pronounced effects on the R &D intensity of high-technology and knowledge-intensive firms. The effects of public support remain positive and significant even after accounting for treatment endogeneity.

previous article Predicting binary outcomes based on the pair-copula construction

next article DS-HECK: double-lasso estimation of Heckman selection model

Dont have a licence yet? Then find out more about our products and how to get one now:

Springer Professional "Wirtschaft"

Online-Abonnement

Mit Springer Professional "Wirtschaft" erhalten Sie Zugriff auf:

über 67.000 Bücher
über 340 Zeitschriften

aus folgenden Fachgebieten:

Bauwesen + Immobilien
Business IT + Informatik
Finance + Banking
Management + Führung
Marketing + Vertrieb
Versicherung + Risiko

Jetzt Wissensvorsprung sichern!

inform now

Springer Professional "Wirtschaft+Technik"

Online-Abonnement

Mit Springer Professional "Wirtschaft+Technik" erhalten Sie Zugriff auf:

über 102.000 Bücher
über 537 Zeitschriften

aus folgenden Fachgebieten:

Automobil + Motoren
Bauwesen + Immobilien
Business IT + Informatik
Elektrotechnik + Elektronik
Energie + Nachhaltigkeit
Finance + Banking
Management + Führung
Marketing + Vertrieb
Maschinenbau + Werkstoffe
Versicherung + Risiko

Jetzt Wissensvorsprung sichern!

inform now

Available only for authorised users

Given that it is impossible to construct a balanced panel dataset without losing a large number of firms, in this paper we use a pooled regression approach.

Data availability as scientific-use files (SUF) and as secure-use files in the Safe Centre (SC) in Luxembourg is reported for each country in Appendix B for the last three waves of the survey.

For example, a multinational company or a holding company.

For details about technology-based and knowledge-based classification of industries see Appendix C.

In fully parametric models it is assumed that the errors follow a normal distribution and the LIV curve shape is determined by simply the inverse of the standard normal distribution multiplied by a constant.

The threefold cross-validation is used as an alternative to the leave-one-out cross-validation since the latter is more computationally expensive.

See Appendix C for a classification of industries based on technology and knowledge intensity.

They analyze the effects of subsidies on innovation for the Italian firms in the Emilia-Romagna region via a regression discontinuity design and use as a proxy for innovation whether or not the firm has applied for at least one patent.

The MTE estimation is used as an alternative and more informative way of exploiting the continuous instrument. But it is possible to use LATE instead by dichotomizing the continuous instrument.

We focus on the endogeneity in the output model motivated by Bloom et al. (2009), Nallari and Bayraktar (2010), Ali et al. (2017), and others. This literature shows evidence that management characteristics are important determinants of the probability to generate R &D output. Given that this information is not observed in our dataset it is possible that the conditional treatments are no longer randomly assigned and causing endogeneity in our output model. Endogeneity in the input model is not considered in this paper.

These numbers are based on the estimation results using (deep) neural networks since the out-of-sample (test set) mean squared errors were lower compared to the other estimation methods. Nevertheless, all methods give comparable results.

Low resistance refers to firms that are very likely to be assigned the treatment even when the instrument values are very low (government revenue in this case). These are firms that are considered to be more “deserving”/“in need” of the treatment compared to firms that would get the treatment only in cases when the instrument value is very large (a larger government revenue). The latter firms would be considered “high resistance.”

These are measured in EUR thus, they could also be affected by the depreciation of EUR over time.

GERD is measured in USD constant prices using 2010 base year and Purchasing Power Parities and as % of GDP. Source: OECD estimates based on OECD Main Science and Technology Indicators Database, 2019.

Source: Eurostat.

Source: European Innovation Scoreboard 2014–2020.

This includes only funding by the government and excludes other sources of public funding such as higher education and EU funding.

For this paper, the dimension of the input is around 100.

Acs ZJ, Audretsch DB (2003) Innovation and technological change. In: Handbook of entrepreneurship research, pp 55–79. Springer

Aghion P, Howitt P (1990) A model of growth through creative destruction. Technical report, National Bureau of Economic Research

Ahn SC, Lee YH, Schmidt P (2013) Panel data models with multiple time-varying individual effects. J Econom 174(1):1–14

Aigner D, Lovell CK, Schmidt P (1977) Formulation and estimation of stochastic frontier production function models. J Econom 6(1):21–37

Ali Z, Sun H, Ali M (2017) The impact of managerial and adaptive capabilities to stimulate organizational innovation in SMEs: a complementary PLS-SEM approach. Sustainability 9(12):2157

Almus M, Czarnitzki D (2003) The effects of public R &D subsidies on firms’ innovation activities: The case of Eastern Germany. J Bus Econ Stat 21(2):226–236

Angrist JD, Pischke J-S (2008) Mostly harmless econometrics: an Empiricist’s Companion. Princeton University Press

Angrist JD, Imbens GW (1995) Identification and estimation of local average treatment effects. Technical report, National Bureau of Economic Research

Arrow KJ (1962) Economic welfare and the allocation of resources for invention, pp 609–626. Princeton University Press

Arrow, K. J. (1971). The economic implications of learning by doing. In: Readings in the theory of growth, pp 131–149. Springer

Audretsch DB (1995) Innovation and industry evolution. MIT Press

Audretsch DB, Link AN, Scott JT (2002) Public/private technology partnerships: evaluating SBIR-supported research. Res Policy 31(1):145–158

Austin PC (2007) The performance of different propensity score methods for estimating marginal odds ratios. Stat Med 26(16):3078–3094

Austin PC, Grootendorst P, Normand S-LT, Anderson GM (2007) Conditioning on the propensity score can result in biased estimation of common measures of treatment effect: a Monte Carlo study. Stat Med 26(4):754–768

Balestriero R et al. (2018) A spline theory of deep learning. In: international conference on machine learning, pp 374–383

Bang H, Robins JM (2005) Doubly robust estimation in missing data and causal inference models. Biometrics 61(4):962–973

Bloom N, Kretschmer T, Van Reenan J (2009) Work-life balance, management practices and productivity. In: International differences in the business practices and productivity of firms, pp 15–54. University of Chicago Press

Breiman L et al (2001) Statistical modeling: the two cultures (with comments and a rejoinder by the author). Stat Sci 16(3):199–231

Bronzini R, Piselli P (2016) The impact of R &D subsidies on firm innovation. Res Policy 45(2):442–457

Busom I (2000) An empirical evaluation of the effects of R &D subsidies. Econ Innov New Technol 9(2):111–148

Carmichael J (1981) The effects of mission-oriented public R &D spending on private industry. J Financ 36(3):617–627

Chen X, White H (1999) Improved rates and asymptotic normality for nonparametric neural network estimators. IEEE Trans Inf Theory 45(2):682–691

Chernozhukov V, Chetverikov D, Demirer M, Duflo E, Hansen C, Newey W, Robins J (2018) Double/debiased machine learning for treatment and structural parameters. Econ J 21(1):C1–C68

Cybenko G (1989) Approximation by superpositions of a sigmoidal function. Math Control Signals Syst 2(4):303–314

Czarnitzki D, Fier A (2001) Do R &D subsidies matter? Evidence for the German service sector, Technical report, ZEW Discussion Papers

Czarnitzki D, Fier A (2002) Do innovation subsidies crowd out private investment? Evidence from the German service sector, Technical report, ZEW Discussion Papers

David PA, Hall BH, Toole AA (2000) Is public R &D a complement or substitute for private R &D? A review of the econometric evidence. Res Policy 29(4–5):497–529

European Commission et. al (2014). Taking stock of the europe 2020 strategy for smart, sustainable and inclusive growth

Farrell, M. H., Liang, T., and Misra, S. (2018). Deep neural networks for estimation and inference. arXiv preprint arXiv:1809.09953

Fier, A. (2002). Staatliche Förderung industrieller Forschung in Deutschland: eine empirische Wirkungsanalyse der direkten Projektförderung des Bundes. Nomos-Verlag-Ges

Friedman JH, Roosen CB (1995) An introduction to multivariate adaptive regression splines. Stat Methods Med Res 4(3):197–217

Frölich M (2007) Nonparametric iv estimation of local average treatment effects with covariates. J Econom 139(1):35–75

Gilmer J, Schoenholz SS, Riley PF, Vinyals O, Dahl GE (2017) Neural message passing for quantum chemistry. arXiv preprint arXiv:1704.01212

Glorot X, Bengio Y (2010) Understanding the difficulty of training deep feedforward neural networks. In: Proceedings of the thirteenth international conference on artificial intelligence and statistics, pp 249–256

Görg H, Strobl E (2007) The effect of R &D subsidies on private R &D. Economica 74(294):215–234

Grifell-Tatjé E, Lovell CK, Sickles RC (2018) The oxford handbook of productivity analysis. Oxford University Press

Griliches, Z. (1979). Issues in assessing the contribution of research and development to productivity growth. Bell J Econ pp 92–116

Griliches, Z. (1998). Patent statistics as economic indicators: a survey. In: R &D and productivity: the econometric evidence, pp 287–343. University of Chicago Press

Gulshan V, Peng L, Coram M, Stumpe MC, Wu D, Narayanaswamy A, Venugopalan S, Widner K, Madams T, Cuadros J et al (2016) Development and validation of a deep learning algorithm for detection of diabetic retinopathy in retinal fundus photographs. JAMA 316(22):2402–2410

Hamberg D (1966) R &D: Essays on the Economics of Research and Development. Random house, New York, NY

Harhoff D (1998) R &D and productivity in German manufacturing firms. Econ Innov New Technol 6(1):29–50

Hartford J, Lewis G, Leyton-Brown K, Taddy M (2017) Deep iv: a flexible approach for counterfactual prediction. In: Proceedings of the 34th international conference on machine learning Vol 70, pp 1414–1423. JMLR. org

Hastie T, Tibshirani R, Friedman J, Hastie T, Tibshirani R, Friedman J (2009) Random forests. Data mining, inference, and prediction, The elements of statistical learning, pp 587–604

Heaton J, Polson N, Witte JH (2017) Deep learning for finance: deep portfolios. Appl Stoch Models Bus Ind 33(1):3–12

Heckman JJ, Vytlacil E (2005) Structural equations, treatment effects, and econometric policy evaluation 1. Econometrica 73(3):669–738

Heckman JJ, Ichimura H, Smith J, Todd P (1996) Sources of selection bias in evaluating social programs: an interpretation of conventional measures and evidence on the effectiveness of matching as a program evaluation method. Proc Natl Acad Sci 93(23):13416–13420

Heckman J, Ichimura H, Smith J, Todd P (1998) Characterizing selection bias using experimental data. Technical report, National Bureau of Economic Research

Helpman E (2009) The mystery of economic growth. Harvard University Press

Heshmati A, Loof H et al. (2005) The impact of public funds on private R &D investment: new evidence from a firm level innovation study. Technical report

Hinton G, Deng L, Yu D, Dahl GE, Mohamed A-R, Jaitly N, Senior A, Vanhoucke V, Nguyen P, Sainath TN et al (2012) Deep neural networks for acoustic modeling in speech recognition: the shared views of four research groups. IEEE Signal Proc Magaz 29(6):82–97

Hollanders H, Es-Sadki N, Merkelbach I, Khalilova A (2020) European innovation scoreboard 2020. Publications Office of the European Union, European Commission

Hussinger K (2008) R &D and subsidies at the firm level: an application of parametric and semiparametric two-step selection models. J Appl Econom 23(6):729–747

Johansson, F., Shalit, U., and Sontag, D. (2016). Learning representations for counterfactual inference. In: International conference on machine learning, pages 3020–3029

Kang JD, Schafer JL et al (2007) Demystifying double robustness: a comparison of alternative strategies for estimating a population mean from incomplete data. Stat Sci 22(4):523–539

Kennedy EH, Lorch S, Small DS (2019) Robust causal inference with continuous instruments using the local instrumental variable curve. J R Stat Soc Ser B (Stat Methodol) 81(1):121–143

Kingma DP, Ba J (2014) Adam: a method for stochastic optimization. arXiv preprint arXiv:1412.6980

King G, Nielsen R (2019) Forthcoming. Why propensity scores should not be used for matching. Political Anal, 27(4)

Klette TJ, Møen J, Griliches Z (2000) Do subsidies to commercial R &D reduce market failures? Microeconometric evaluation studies. Res Policy 29(4–5):471–495

Krizhevsky A, Sutskever I, Hinton GE (2012) Imagenet classification with deep convolutional neural networks. In: Advances in neural information processing systems, pp 1097–1105

Lach S (2002) Do R &D subsidies stimulate or displace private R &D? Evidence from Israel. J Ind Econ 50(4):369–390

Levin RC, Cohen WM, Mowery DC (1985) R &D appropriability, opportunity, and market structure: new evidence on some Schumpeterian hypotheses. Am Econ Rev 75(2):20–24

Lichtenberg FR (1987) The effect of government funding on private industrial research and development: a re-assessment. J Ind Econ 30(1):97–104

Lichtenberg FR (1988) The privateR &D investment response to federal design and technical competitions. Am Econ Rev 78(3):550–559

Lipsitz SR, Ibrahim JG, Zhao LP (1999) A weighted estimating equation for missing covariate data with properties similar to maximum likelihood. J Am Stat Assoc 94(448):1147–1160

Lucas RE (1998) On the mechanics of economic development. Econ Soc Monogr 29:61–70

Lunceford JK, Davidian M (2004) Stratification and weighting via the propensity score in estimation of causal treatment effects: a comparative study. Stat Med 23(19):2937–2960

Marino M, Lhuillery S, Parrotta P, Sala D (2016) Additionality or crowding-out? An overall evaluation of public R &D subsidy on private R &D expenditure. Res Policy 45(9):1715–1730

McClelland JL, Rumelhart DE, Group PR, et al. (1986) Parallel distributed processing. Explor Microstruct Cogn. 2:216–271

Montufar GF, Pascanu R, Cho K, Bengio Y (2014) On the number of linear regions of deep neural networks. In: Advances in neural information processing systems, pp 2924–2932

Mullainathan S, Spiess J (2017) Machine learning: an applied econometric approach. J Econ Perspect 31(2):87–106

Nallari R, Bayraktar N (2010) Micro efficiency and macro growth. The World Bank

Nelson RR (1959) The simple economics of basic scientific research. J Political Econ 67(3):297–306

Neugebauer R, van der Laan M (2005) Why prefer double robust estimators in causal inference? J Stat Plann Inference 129(1–2):405–426

Newey, K. and McFadden, D. (1994). Large sample estimation and hypothesis. Handbook of Econometrics, IV, Edited by RF Engle and DL McFadden, pp 2112–2245

Oecd E (2005) Oslo manual: guidelines for collecting and interpreting innovation data. Paris 2005, Sp, 46

Ortega-Argilés R, Potters L, Vivarelli M (2011) R &D and productivity: testing sectoral peculiarities using micro data. Empir Econ 41(3):817–839

Pati, Y. C., Rezaiifar, R., and Krishnaprasad, P. S. (1993). Orthogonal matching pursuit: Recursive function approximation with applications to wavelet decomposition. In Proceedings of 27th asilomar conference on signals, systems and computers, pages 40–44. IEEE

Robins JM (2000) Marginal structural models versus structural nested models as tools for causal inference. In: Statistical models in epidemiology, the environment, and clinical trials, pp 95–133. Springer

Robins J, Rotnitzky A (2001) Comment on the Bickel and Kwon article, On double robustness. Statistica Sinica 11(4):920–936

Robins JM, Rotnitzky A, Zhao LP (1995) Analysis of semiparametric regression models for repeated outcomes in the presence of missing data. J Am Stat Assoc 90(429):106–121

Romer PM (1990) Endogenous technological change. J Political Econ, 98(5, Part 2):S71–S102

Romer PM (1987) Growth based on increasing returns due to specialization. Am Econ Rev 77(2):56–62

Rosenbaum PR, Rubin DB (1983) The central role of the propensity score in observational studies for causal effects. Biometrika 70(1):41–55

Scharfstein DO, Rotnitzky A, Robins JM (1999) Adjusting for nonignorable drop-out using semiparametric nonresponse models. J Am Stat Assoc 94(448):1096–1120

Shalit U, Johansson FD, Sontag D (2017) Estimating individual treatment effect: generalization bounds and algorithms. In: Proceedings of the 34th international conference on machine learning Vol 70, pp 3076–3085. JMLR. org

Sickles RC (2022) The ET interview: peter schmidt interviewed by Robin C. Sickles, Rice University. Econ Theory, 1–19

Simonyan K, Zisserman A (2014) Very deep convolutional networks for large-scale image recognition. arXiv preprint arXiv:1409.1556

Solow RM (1956) A contribution to the theory of economic growth. Q J Econ 70(1):65–94

Stiglitz JE (1988) Econ Public Sector. WW Norton

Toivanen O, Niininen P (2000) Investment, R &D, subsidies, and credit constraints. Working Paper No. 264, Helsinki School of Economics

Van der Laan, M. J., Laan, M., and Robins, J. M. (2003). Unified methods for censored longitudinal data and causality. Springer Science & Business Media

Varaku, K. (2020). Stock price forecasting and hypothesis testing using neural networks. Available at SSRN: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3597684

Vytlacil E (2002) Independence, monotonicity, and latent index models: an equivalence result. Econometrica 70(1):331–341

Wager S, Athey S (2018) Estimation and inference of heterogeneous treatment effects using random forests. J Am Stat Assoc 113(523):1228–1242

Wallsten SJ (2000) The effects of government-industry R &D programs on private R &D: the case of the small business innovation research program. RAND J Econ 31(1):82–100

Wang Z, Scott DW (2019) Nonparametric density estimation for high-dimensional data-algorithms and applications. Wiley Interdiscip Rev Comput Stat 11(4):e1461

Wang N, Yeung DY (2013) Learning a deep compact image representation for visual tracking. In: Advances in neural information processing systems, pp 809–817

Westreich D, Lessler J, Funk MJ (2010) Propensity score estimation: neural networks, support vector machines, decision trees (cart), and meta-classifiers as alternatives to logistic regression. J Clin Epidemiol 63(8):826–833

Wooldridge JM (2010) Econometric analysis of cross section and panel data. MIT press

Title: Public subsidies and innovation: a doubly robust machine learning approach leveraging deep neural networks
Authors: Kerda Varaku
Robin Sickles
Publication date: 11-05-2023
Publisher: Springer Berlin Heidelberg
Published in: Empirical Economics / Issue 6/2023
Print ISSN: 0377-7332
Electronic ISSN: 1435-8921
DOI: https://doi.org/10.1007/s00181-023-02398-7

Springer Professional

Abstract

Please log in to get access to your license.

Dont have a licence yet? Then find out more about our products and how to get one now:

Springer Professional "Wirtschaft"

Springer Professional "Wirtschaft+Technik"

Other articles of this Issue 6/2023

Forecasting in the presence of in-sample and out-of-sample breaks

Information loss in volatility measurement with flat price trading

Approximating long-memory processes with low-order autoregressions: Implications for modeling realized volatility

Unbiased estimation of the OLS covariance matrix when the errors are clustered

Refined GMM estimators for simultaneous equations models with network interactions

Internal adjustment costs of firm-specific factors and the neoclassical theory of the firm

Premium Partner