Economic Conditions for Innovation: Private vs. Public Sector
IInduced Innovation and Economic Environment
Tomáš Evan
Czech Technical University in PragueThákurova 2077/7, 160 00 Prague 6, Czechiatomas.evan@fit.cvut.cz
Vladimír Holý
University of Economics, PragueWinston Churchill Square 1938/4, 130 67 Prague 3, [email protected]
April 20, 2020
Abstract:
The Hicks induced innovation hypothesis states that a price increase of a productionfactor is a spur to invention. We propose an alternative hypothesis restating that a spur to inventionrequire not only an increase of one factor but also a decrease of at least one other factor to offsetthe companies’ cost. We illustrate the need for our alternative hypthesis in a historical example ofthe industrial revolution in the United Kingdom. Furthermore, we econometrically evaluate bothhypotheses in a case study of research and development (R&D) in 29 OECD countries from 2003to 2017. Specifically, we investigate dependence of investments to R&D on economic environmentrepresented by average wages and oil prices using panel regression. We find that our alternativehypothesis is supported for R&D funded and/or performed by business enterprises while the originalHicks hypothesis holds for R&D funded by the government and R&D performed by universities.
Keywords:
Research and Development, Induced Innovation, Hicks’ Theory, Price Changes of Pro-duction Factors.
JEL Codes:
C33, E22, O31, O33.
The goal of this paper is to analyse and verify the induced innovation hypothesis of J.R. Hicks firstpublished in his Theory of Wages in 1932, which attributes a spur to invention to a price increaseof a production factor (Hicks, 1963). This hypothesis was mainly tested on wages and their impacton labour saving technologies and, more recently, the impact of high energy prices on environmentaltechnology innovations and energy savings. In our preliminary study Bolotov and Evan (2017), wehave attempted to falsify, in the sense of Popper, the Hicks hypothesis by means of orthodox modelling.Using regression model for intellectual property protection as the proxy variable for innovation, wehave found that an increase in relative price of one factor mandates a relatively low price level of theother factor(s) to offset the companies’ cost, for the innovations to take place. While the low cost ofother factor(s) of production does not diminish the motivation to substitute the high-priced factor ofproduction, it also gives companies the necessary capital for the innovation process.We follow our preliminary work and investigate this alternative hypothesis from two perspectives.First, in sections 2–3, we review Hicks hypothesis and demonstrate the need for our alternativehypothesis in historical examples with focus on the industrial revolution in the United Kingdom.Second, in sections 4–7, we analyze investments to research and development (R&D) in 29 OECDcountries from 2003 to 2017 in order to assess the suitability of our alternative hypothesis for thepresent time. We discuss the findings in Section 8.1 a r X i v : . [ ec on . GN ] A p r Hicks’ Theory of Induced Innovation, Proofs and Criticism
Numerous studies of the history of economics, including those in recent years, such as Lee andKang (2007), Angelini et al. (2009), Gallouj and Savona (2009), Savona and Steinmueller (2013),Fabre (2014), and Milyaeva and Fedorkevich (2015), have consistently shown that innovations benefitcompanies, industries and economies in terms of increasing competitiveness, economic growth anddevelopment. There is however little consensus on what the main causes of innovation are. Sir J.R.Hicks (Hicks, 1963, p. 124) has stated the following hypothesis, which later became the foundationof Hicks’ widely discussed Theory of Induced Innovation.
Hypothesis 1 (Hicks Induced Innovation Hypthosis) . A change in the relative prices of the factors ofproduction is itself a spur to invention, and to invention of a particular kind – directed to economizingthe use of a factor which has become relatively expensive.
The relative straightforwardness and immense implications of the Induced Innovation Theorytogether with the name of the well-known British economist and Nobel Prize laureate has caused thetheory to be widely discussed from the moment it was formulated. The theory had both proponents(Fellner, 1961, 1971, Samuelson, 1965, Kennedy, 1967, i.a.) as well as opponents. The latter includeNobelist W.D. Nordhaus who criticised it for necessitating very strong and limiting preconditions,among other flaws. Nordhaus thinks “the model is too defective to be used in serious economicanalysis” (Nordhaus, 1973, p. 208). Others criticised the lack of economic foundations which areimplicit in the theory and tried to establish them (Funk, 2002).The crucial fact that the production factor shares stay relatively constant in the productionfunction remains a widely accepted stylized fact. Yet, whether this is caused by instant “spur toinvention” and thus serves as proof of the theory, remains as controversial as it was back in Hicks’time. Perhaps even more controversially, there is still generally accepted mechanism by which changesin factor prices affect inventive or innovative activity (Salter and Reddaway, 1966, pp. 43–44, Ahmad,1966, Hayami and Ruttan, 1971, Ruttan and Hayami, 1984, i.a.).Not even the theory’s critics can deny, however, that it has been used and in quite a few casesproven empirically. Attempted from the start and followed by a ground building paper of WilliamFellner “Empirical Support for the Theory of Induced Innovation” (Fellner, 1971) there have beenseveral fields in which endorsement and application could be found. The line of research empiricallyconfirming the main hypothesis included in the theory was originally centred on high wages spurringlabour-saving innovation, and later agricultural development. More recently the emphasis has shiftedtowards energy prices and induced innovation in energy-saving technologies.Newell et al. (1999), i.a. found the rate of overall innovation independent of energy prices andregulation, however, the direction of innovation was responsive to energy price changes for severalproducts tested by the authors. Popp (2002), using patent citations as a measure of supply ofknowledge, found that both energy prices and the quality of existing knowledge have significantlystrong positive effects on innovation. Also using patent counts and citation data, Jang and Du (2013)confirm that demand and supply factors – including knowledge stocks and crude-oil price – havepositive and statistically significant effects on technological biofuel innovations in the United Statesof America.There is, however, also a relatively large number of other correlates to innovation such as inwardforeign direct investment, outward foreign direct investment, imports, state guarantees and incentivesamong many other, as stressed by Lin and Lin (2008). Explaining the causes of innovation is thereforea long-standing problem in social science, while the large body of existing literature has not beenconclusive. This paper adds to this ongoing discussion an attempt to alter significantly the existingHicks hypothesis to the point of its negation, among others ways through evidence from economichistory, specifically, from the example of industrialization in the U.K. (the initial one) and in theworld. 2
Historical Examples of Hicks’ Theory and the Need for an Alter-native Hypothesis
One indication that the increase in relative price of one factor mandates a relatively low price levelof other factor(s) of production would be to look at historical examples of eras of rapid innovation .A near-perfect example of such rapid and continuous innovation is the industrial revolution. Therewere many causes and necessary conditions for this long and complicated process of introduction ofmechanized production. For Rostow’s daring attempt to assign dates to various countries’ “take-offs”please see Table 1 (Rostow, 1991, Chapters 3–4 and Baldwin and Martin, 1999). These included po-litical stability, sufficient capital accumulation, relatively mature banking sector, etc. (for a somewhatmore comprehensive list see the works of Landes, 1998, Chapters 13–16, Pomeranz, 2001, Chapters5–6, Maddison, 2007, Chapters 2, 3, 6, Evan, 2014, Chapters 1–2, as well as others).Apart from these uncontroversial conditions which were sooner or later fulfilled in most countriesof the northern hemisphere at least, there is still the issue of the cause of this far-reaching economicand social change. Particularly the question, to quote Allen (2009, 2011, 2015), “why the industrialrevolution was British?” In other words, why was the innovation in Great Britain spurred fifty ormore years earlier (end of the 18th century) than in nearby countries with similar socio-economiccharacteristics? There is hardly any discussion about the prime reason for businessmen trying toreplace human labour with machines. The reason is high wages, for calculations see Botham and Hunt(1987) and Kelly et al. (2014). While not uniquely high, since both GDP and income per capita werehigher still in the Netherlands, British wages were the prime incentive for the innovation of labour-saving techniques. This, however, would not be enough, as it was not enough in the Netherlandswhich became industrialized much later than Great Britain, as shown by Landes (1998, Chapters15–16). A relatively low price of the second factor of production had to be present. The second factorin the case of British industrialization was cheap energy in both the textile and iron industries. Suchenergy was ensured by fast running streams for textile mills at first, and soon replaced by accessible,abundant coal of good enough quality. Thus, the high-wage economy of London together with the useof cheap coal shipped in from Newcastle led to the early industrialization of Britain, as it motivatedmechanical production and allowed this innovation by savings made from cheap energy.Had there not been both factors of production in this fortunate price combination, the firstindustrialization would not have been British, but Dutch or German perhaps, see Allen (2011, p. 366),also (Landes, 1998, Chapters 15–16) and Allen (2015) .Energy prices were dominant for the iron and steel industry in Britain. For textiles the costs ofinputs of raw materials were even more important. For many centuries Britain exported wool andwoollens and to a lesser extend linen. While making a good export product and enriching both landowners and merchants there was no way how to accumulate enough capital to mechanize production,nor was there a reason to do so. It was only after cotton became available, the price of which couldbe forced down to levels constituting a significant advantage, that the combination of high wages ofspinners and weavers together with cheap cotton from the West Indies and later plantations in theAmerican South allowed for induced innovation, as shown by Broadberry et al. (2013), and Tomory(2016). This combination was so powerful that British industrialists overcame a number of obstaclesincluding strong competition from high quality Indian cotton products as well as the fact that all rawcotton had to be imported from faraway and often unstable locations.Once started, the innovative process reinforced itself in several ways. It allowed for tremendouseconomies of scale and profits trumping all other non-mechanized productions around the world.This, in turn, put pressure on wages to keep rising, thus motivating the implementation of furtherlabour-saving techniques. The industrial revolution created so called Big Divergence (see Pritchett,1997) increasing incomes in industrialized countries (Europe, USA, Japan) and creating “innovativecentres” in these countries while de-industrializing everyone else (see Landes, 1998, Maddison, 2007, This will allow us to determine cause and effect, whether innovations are spurred by changes in relative prices ofinputs with the motive to economize or not. To sum up, a high price of one factor offset by a low price of another input could be sufficient to spur innovationin the U.K., which is opposite logic to the effect described by Hicks.
Hypothesis 2 (Alternative Induced Innovation Hypothesis) . Innovation is spurred by an increasein the relative price of one factor of production compensated by a decrease in relative price of anotherfactor of production.
This paper attempts to reflect the changed conditions for innovative process in societies surroundingus today. The current economic conditions are not of market economy but mixed economy instead,while government finances and guarantees much of the research and development (R&D) in developedcountries around the world (see Table 2 and Figure 2). It is unlikely, therefore, that the conditionsmotivating for the innovation would remain the same as were in previous two centuries.Therefore, in this paper we try to identify whether the market conditions of relative prices offactors of production still hold sway as motivating factor for the spur to innovation, according toour alternative Hicks hypothesis. Or, if the motivation is more in line with the motivation thatcan be expected from the governmental sector. This would include correlation of research outputwith budgetary constraints and GDP levels, which is the base for government expenditure ratherthan any market conditions. We also include educational sector, that is, universities, as well as non-governmental research oriented institutions in our analysis. The last included actor with potentiallysignificant impact on research and development levels of a country would be foreign based businessesout of which those using foreign direct investment as a mode of entry for their investment might beof relevance for country’s R&D.In line with our previous research and concerning literature we consider impact of relative pricesof different factors of production and keep wages and price of oil as the two most relevant. Oil hasclearly replaced coal as the most important source of energy while labour costs remain the single4ost important factor of production across industries. There is large body of literature relatingwages of skilled and unskilled labour and innovation. It is clearly two-way street with high wagesmotivating innovation and innovation favouring high-skilled labour (Bogliacino et al., 2018, i.a.).In the light of our results somehow surprisingly the available literature considers energy prices asto have strong and positive effect on innovation. Popp (2002) suggests the impact is so strong itcan be advantageous for government to use market-based environmental policies to help ameliorateglobal warming with the help of price-induced technological change. In several sectors of economy(for a review see Ruttan, 2001) higher oil prices were found to lead to an increased innovation.Particularly in the automotive industry as it is an energy-intensive sector. Crabb and Johnson (2010)also suggest, however, a government intervention, this time based upon the premise of possible under-investment in private research and development given the relatively slow diffusion of knowledge inthe sector combined with high effectiveness of carbon-based taxes in encouraging innovation amongother things. Somehow opposite view can be found on European electricity industry where morederegulation namely in barriers of entry, public ownership and vertical integration seems to havepositive impact on innovation activity (Cambini et al., 2016).As a measure of innovation activities, we utilize the size of investments to research and develop-ment. The investment is of course just the beginning of the innovation process but is a clear anduniversal indicator of its magnitude. On the other hand, measurable R&D outputs such as the numberof patents, the share of high-technology exports and the number of scientific publications relate onlyto specific fields and sectors. Holý and Šafr (2018) investigate the relation between the inputs andoutputs of the R&D process and find that countries with higher GDP per capita may not necessarilyproduce more outputs in terms of the patent and citation counts, but are significantly more efficientin transforming the investments and the human capital into these outputs.
We investigate the member countries of Organisation for Economic Co-Operation and Development(OECD) from 2003 to 2017. The main analyzed variable is the annual gross domestic R&D ex-penditure per capita. The economic environment is represented by the average monthly wage andautomotive diesel oil price per litres. All variables are current prices in USD adjusted for pur-chasing power parities. The evolution of the variables over time is illustrated in Figure 1. The sourceof the R&D expenditures and average wages is OECD while the source of the oil prices is InternationalEnergy Agency (IEA).As the R&D expenditure is our main object of interest, we further elaborate on it. The dataare collected using the standard OECD methodology for statistics related to R&D described in theFrascati Manual (OECD, 2015). Besides the total intramural gross domestic R&D expenditures(TOTAL), we also utilize the R&D expenditures broken down by the source of funding and thesector of performance as well. There are five sources of funding: the business enterprise (FUND-BES), government (FUND-GOV), higher education (FUND-HES), private non-profit (FUND-PNP)and rest of the world (FUND-ROW). Furthermore, there are four sectors of performance: the businessenterprise (PERF-BES), government (PERF-GOV), higher education (PERF-HES) and private non-profit (PERF-PNP). The average shares of the R&D expenditures for specific sources of funding andsectors of performance are shown in Table 2. The dominant R&D segment is the self-funded businesssector with 52 percent share. Other significant segments are the higher education sector funded bythe government with 17 percent share and the self-funded government sector with 11 percent share.Figure 2 shows the composition of the source of funding in each country while Figure 3 shows thecomposition of the sector of performance.Unfortunately, not all variables are available for all OECD countries. For this reason, we analyzeonly 29 of the 36 OECD member countries over 15 years. We exclude Australia, Chile, Iceland, Israel,Latvia and Lithuania from the analysis due to missing oil prices and Turkey due to missing averagewages. Furthermore, our dataset contains some additional missing values. For the analysis of thetotal intramural R&D expenditure, we have 382 observations with all variables. This means that 53observations are missing. For the R&D expenditures with a specific source of funding or sector of5
Year P r i c e i n U S D Variable
R&D ExpenditureAverage WageOil Price
Total R&D Expenditure, Average Wage and Oil Price in Time
Figure 1: The total R&D expenditure per capita, average wage and oil price averaged over 29 OECDcountries. Source of FundingSector of Perf. BES GOV HES PNP ROW TotalBES 51.80 4.50 0.04 0.13 6.51 62.98GOV 0.84 10.57 0.05 0.14 0.78 12.38HES 1.32 17.20 2.23 0.83 1.38 22.95PNP 0.25 0.70 0.01 0.53 0.19 1.69Total 54.20 32.97 2.34 1.63 8.86 100.00Table 2: The shares of R&D expenditures per capita in percents averaged over 29 OECD countriesfrom 2003 to 2017.performance, we have between 334 and 385 observations. The exception is FUND-HES variable withonly 293 observations and the PERF-PNP variable with only 285 observations. In these two cases,some countries are entirely missing due to differences in data collection methodology.
To analyze the influence of the economic environment on the R&D expenditure, we utilize the panelregression. As the dependent variable, we consider the total intramural R&D expenditure per capitaas well as R&D expenditures per capita with a specific source of funding and R&D expenditures percapita with a specific sector of performance. Therefore, we build ten models in total. In all cases,we consider the average wage and the oil price as the independent variables. We find that strongerresults are obtained when the independent variables are lagged by one year. This suggest that thereis a one-year delay before the decision to invest and the actual investment.The considered specifications of the panel regression model are as follows. First, we find outwhether individual and time effects are needed. For this purpose, we utilize the Lagrange multipliertest of Honda (1985). The p-values of the individual effects test for all dependent variables arevirtually zero. In contrast, the p-value of the time effects test for the total intramural expenditurevariable is 0.98 and is similarly high for all other dependent variables according to Table 3. Therefore,we include only individual effects in the model. 6 A U T BE L C A N CH E C Z E D E U DN
K ESP ES
T F I N F R A G B R G RC HUN I R L I T A J P N K O R L U X M EX N L D N O R N Z L P O L P R T SVK SV N S W E U SA Country A v e r age E x pend i t u r e pe r C ap i t a Source of Funding
Business EnterpriseGovernmentHigher EducationPrivate Non−ProfitRest of the World
R&D Expenditure by Country and Source of Funding
Figure 2: The R&D expenditures per capita averaged over time from 2003 to 2017 and broken downby the source of funding. A U T BE L C A N CH E C Z E D E U DN
K ESP ES
T F I N F R A G B R G RC HUN I R L I T A J P N K O R L U X M EX N L D N O R N Z L P O L P R T SVK SV N S W E U SA Country A v e r age E x pend i t u r e pe r C ap i t a Sector of Performance
Business EnterpriseGovernmentHigher EducationPrivate Non−Profit
R&D Expenditure by Country and Sector of Performance
Figure 3: The R&D expenditures per capita averaged over time from 2003 to 2017 and broken downby the sector of performance. 7ffects Test Ser. Cor. Test Hetero. Test R-SquaredModel Ind. Time Diff. With. Diff. With. Diff. With.TOTAL 0.0000 0.9768 0.1684 0.0000 0.0000 0.0405 0.1997 0.7129FUND-BES 0.0000 0.9451 0.0001 0.0000 0.0255 0.7323 0.1641 0.3838FUND-GOV 0.0000 0.9452 0.7296 0.0118 0.3297 0.3017 0.2330 0.5779FUND-HES 0.0000 0.4788 0.3130 0.0312 0.0000 0.0253 0.0112 0.3290FUND-PNP 0.0000 0.9863 0.0523 0.0000 0.0135 0.0202 0.0270 0.3503FUND-ROW 0.0000 0.9719 0.4606 0.0004 0.0001 0.0796 0.0070 0.3435PERF-BES 0.0000 0.9791 0.0336 0.0000 0.0000 0.2687 0.1138 0.4760PERF-GOV 0.0000 0.9922 0.8307 0.0000 0.0417 0.0939 0.0112 0.2885PERF-HES 0.0000 0.9763 0.2988 0.0000 0.0865 0.3058 0.1262 0.7391PERF-PNP 0.0000 0.9931 0.1866 0.0000 0.7675 0.2294 0.0456 0.1085Table 3: The specification tests and R-squared statistics for the panel regression.Next, we choose between the first-differences estimator and the within estimator. Both estimatorsdeal with the fixed effects in the panel regression. Our variables are clearly non-stationary (seeFigure 1) and therefore we resort to the first-differences estimator as it removes any unit roots in thedependent and independent variables as well. In contrast, a spurious relation between the variablesis a serious issue for the within transformation. The auxiliary-regression-based Hausman test ofWooldridge (2010, Section 10.7.3) suggests that random effects can be considered as well. However,the model with random effects would face the same issues as in the case of the within estimator.Finally, we investigate the structure of the error terms. We adopt the heteroskedasticity test ofBreusch and Pagan (1979) and the serial correlation test of Wooldridge (2010, Section 10.6.3) basedon first differences. The p-values of the tests are presented in Table 3 for both the first-differencesand within estimator. We find that there is no universal behavior across our models in terms of het-eroskedasticity and serial correlation. In general, however, the first-differences transformation removesserial correlation much better than the within transformation. This is another major motivation forthe first-differences estimator. Furthermore, heteroskedasticity is present in many models for bothestimators. To account for heteroskedasticity and remaining serial correlation in the error terms, weutilize the White method of Arellano (1987) for robust estimation of the parameter covariance matrix.The resulting panel model is as follows. Let N denote the number of countries, T the numberof time periods and M the number of independent variables. Further, let y i,t denote the dependentvariable of country i in time t and x j,i,t the independent variable j of country i in time t . The linearpanel model with first differences and lagged independent variables is then given by y i,t − y i,t − = β + M (cid:88) j =1 β j ( x j,i,t − − x j,i,t − ) + e i,t , i = 1 , . . . , N, t = 3 , . . . , T, where β j are the unknown coefficients and e i,t are the error terms. Note that the first two timeperiods are used for inicialization of the lagged variables and first differences. The estimated coefficients with standard deviations and p-values for the first-differences estimator arereported in Table 4 while the R-squared statistics are reported in Table 3. Note that the R-squaredstatistic is relatively low for the first-differences estimator as it explains the change in the dependentvariable. On the other hand, the R-squared is much higher for the within estimator as it explainsthe absolute value of the dependent variable which has a clear trend in time. Nevertheless, modelingchanges is more meaningful in our situation while it also proves to be more challenging.First, let us focus on the total intramural R&D expenditures per capita (the TOTAL model). Theaverage wage has significantly positive effect on the R&D expenditure in this model. Specifically, the8nnual R&D expenditure per capita increases by 0.27 USD when the average monthly wage increasesby 1 USD. The effect of the oil price is insignificant. The model explains 20 percent of the variancein the total R&D expenditure per capita changes.Next, let us consider breakdown by the source of funding. The dominant source of funding isthe business enterprise sector with 54 percent share of the total R&D expenditures on average. Inthe FUND-BES model, the average wage has significantly positive effect on the R&D expenditurewhile the oil price has significantly negative effect. The annual R&D expenditure per capita by thebusiness enterprise sector increases by 0.22 USD when the average monthly wage increases by 1 USD.The annual R&D expenditure per capita decreases by 0.03 USD when the oil price per litresincreases by 1 USD. The FUND-BES model explains 16 percent of the variance. Another majorsource of funding is the government sector with 33 percent share of the total R&D expenditures onaverage. In the FUND-GOV model, the average wage has significantly positive effect on the R&Dexpenditure while the oil price has no significant effect. The annual R&D expenditure per capita bythe government sector increases by 0.10 USD when the average monthly wage increases by 1 USD.The FUND-GOV model explains 23 percent of the variance. Other sources of funding have lowershare of the R&D expenditure and are not explained well by our model due to very low R-squaredstatistic and insignificance of regressors.Finally, let us consider breakdown by the sector of performance. The dominant sector of per-formance is the business enterprise sector with 63 percent share of the total R&D expenditures onaverage. The behavior of the PERF-BES model is similar to the FUND-BES model although lesspronounced. The annual R&D expenditure per capita in the business enterprise sector increases by0.17 USD when the average monthly wage increases by 1 USD. The annual R&D expenditure percapita decreases by 0.01 USD when the oil price per litres increases by 1 USD. The PERF-BESmodel explains 11 percent of the variance. The second most important sector is the higher educationsector with 23 percent share of the total R&D expenditures on average. In the PERF-HES model,the average wage has significantly positive effect on the R&D expenditure while the oil price has nosignificant effect. The annual R&D expenditure per capita in the higher education sector increasesby 0.06 USD when the average monthly wage increases by 1 USD. The PERF-HES model explains13 percent of the variance. Other sectors of performance have lower share of the R&D expenditureand are not explained well by our model due to very low R-squared statistic and insignificance ofregressors.As a robustness check, we also report results obtained by the within estimator in Table 5. Justas with the first-differences transformation, we adopt the robust covariance matrix estimation ofArellano (1987) to account for heteroskedasticity and serial correlation in the error terms. We findthat there are no dramatical differences in the coefficients estimated by both methods. The withinestimator, however, puts more significance to the average wage variable. Most likely, this is a spuriousrelation caused by non-stationarity. Overall, the within estimator does not notably deviate from thefirst-differences estimator but is not as reliable due to potential spurious relationship.To conclude, let us relate the results of the panel regression analysis to hypotheses 1 and 2. Theoriginal Hicks hypothesis is supported in the cases of R&D funded by the government sector andR&D performed by the higher education sector. In contrast, our alternative hypothesis is supportedin the cases of R&D funded by the business enterprise sector and R&D performed by the businessenterprise sector. The other R&D expenditures are funded and performed on much lower scales andare not well captured by average wages and oil prices.
The focus of this paper is dependence of the investment to R&D on economic environment in asituation of major governmental involvement in the new millennia. We tried to answer the questionwhether the market conditions of relative prices of factors of production still hold sway as motivatingelement for the spur to innovation as suggested by our alternative Hicks hypothesis. Or, if originalHicks hypothesis apply, or any other pattern can be found.We work with the alternative Hicks hypothesis formulated as follows: innovation is spurred by9ntercept Average Wage Oil PriceModel Coef. St. D. p-Val. Coef. St. D. p-Val. Coef. St. D. p-Val.TOTAL 7.4459 11.1675 0.5054 0.2671 0.1214 0.0285 -0.0077 0.0090 0.3960FUND-BES -2.0300 6.8087 0.7658 0.2156 0.0542 0.0001 -0.0328 0.0088 0.0002FUND-GOV -0.4973 2.0693 0.8102 0.0998 0.0174 0.0000 0.0020 0.0050 0.6886FUND-HES 0.7228 0.3186 0.0242 0.0032 0.0018 0.0868 -0.0010 0.0015 0.5026FUND-PNP 0.5170 0.2448 0.0356 0.0033 0.0030 0.2713 -0.0014 0.0008 0.0860FUND-ROW 5.3768 4.8662 0.2702 -0.0068 0.0450 0.8797 0.0118 0.0068 0.0835PERF-BES 5.6545 8.1957 0.4907 0.1690 0.0894 0.0595 -0.0129 0.0048 0.0076PERF-GOV 1.9911 0.8659 0.0221 0.0088 0.0080 0.2713 0.0034 0.0028 0.2200PERF-HES 2.4450 2.1569 0.2578 0.0640 0.0238 0.0076 0.0026 0.0045 0.5656PERF-PNP -0.0445 0.1983 0.8225 0.0044 0.0020 0.0272 -0.0019 0.0012 0.1081Table 4: The results of the panel regression based on the first-difference estimator.Average Wage Oil PriceModel Coef. St. D. p-Val. Coef. St. D. p-Val.TOTAL 0.3098 0.0446 0.0000 -0.0181 0.0268 0.4990FUND-BES 0.1743 0.0397 0.0000 -0.0390 0.0189 0.0405FUND-GOV 0.0960 0.0176 0.0000 -0.0059 0.0096 0.5442FUND-HES 0.0082 0.0023 0.0006 -0.0009 0.0017 0.5979FUND-PNP 0.0081 0.0026 0.0018 -0.0030 0.0016 0.0574FUND-ROW 0.0472 0.0067 0.0000 0.0132 0.0105 0.2096PERF-BES 0.1812 0.0491 0.0003 -0.0096 0.0253 0.7036PERF-GOV 0.0284 0.0132 0.0318 0.0026 0.0068 0.7027PERF-HES 0.1007 0.0143 0.0000 -0.0113 0.0112 0.3159PERF-PNP 0.0041 0.0016 0.0094 -0.0031 0.0020 0.1237Table 5: The results of the panel regression based on the within estimator.10n increase in the relative price of one factor of production compensated by a decrease in relativeprice of another factor of production. This hypothesis was proven for part of the investment to R&Dwhich was either funded or performed by business enterprises. In this case, we found that marketsources play a major role. The finding is statistically significant and robust. Under market conditionsinnovation is only possible when business not only has the need for innovation (increased relativeprice of factor of production) but also has the resources to conduct it (decrease in relative price ofanother factor of production) or innovation can bring about such a decrease in form of for examplelabour and resources-saving techniques with minimal lag.Investment to R&D funded by government and investment performed by universities fell underoriginal Hicks hypothesis as there seems to be no need for conditions other than increase of relativeprice of factor of production. Presumably government funding is the force which allows for research tocontinue without the need for it to provide other continuous source of income. Continuous crowdingout of private investment by government funds with its less desirable properties, namely weakerperformance, might cause original Hicks hypothesis to be more relevant then before.
Acknowledgements
We would like to thank Jan Zouhar for his useful comments. We would also like to thank organizersand participants of the 29th Eurasia Business and Economics Society Conference (Lisbon, October10–12, 2019) for fruitful discussions.
Funding
The work of Vladimír Holý was supported by the Internal Grant Agency of the University of Eco-nomics, Prague under project F4/27/2020.
References
Ahmad , S. 1966. On the Theory of Induced Invention.
The Economic Journal . Volume 76. Issue302. Pages 344–357. ISSN 0013-0133. https://doi.org/10.2307/2229720 . Allen , R. C. 2011. Why the Industrial Revolution Was British: Commerce, Induced Invention, andthe Scientific Revolution.
The Economic History Review . Volume 64. Issue 2. Pages 357–384. ISSN0013-0117. https://doi.org/10.1111/j.1468-0289.2010.00532.x . Allen , R. C. 2009. Why Was the Industrial Rrevolution British?
Oxonomics . Volume 4. Issue 1.Pages 50–54. ISSN 1752-5195. https://doi.org/10.1111/j.1752-5209.2009.00032.x . Allen , R. C. 2015. The High Wage Economy and the Industrial Revolution: A Restatement.
The Economic History Review . Volume 68. Issue 1. Pages 1–22. ISSN 0013-0117. https://doi.org/10.1111/ehr.12079 . Angelini , E. C.,
Farina , F.,
Pianta , M. 2009. Innovation and Wage Polarisation in Europe.
International Review of Applied Economics . Volume 23. Issue 3. Pages 309–325. ISSN 0269-2171. https://doi.org/10.1080/02692170902811736 . Arellano , M. 1987. Computing Robust Standard Errors for Within-Groups Estimators.
OxfordBulletin of Economics and Statistics . Volume 49. Issue 4. Pages 431–434. ISSN 0305-9049. https://doi.org/10.1111/j.1468-0084.1987.mp49004006.x . Baldwin , R. E.,
Martin , P. 1999.
Two Waves of Globalisation: Superficial Similarities, Funda-mental Differences . Working Paper. .11 ogliacino , F.,
Guarascio , D.,
Cirillo , V. 2018. The Dynamics of Profits and Wages: Technol-ogy, Offshoring and Demand.
Industry and Innovation . Volume 25. Issue 8. Pages 778–808. ISSN1366-2716. https://doi.org/10.1080/13662716.2017.1349651 . Bolotov , I.,
Evan , T. 2017. A Non-Parametric Statistical Test for J. R. Hicks’ Induced InnovationHypothesis. In
Proceedings of the 11th International Days of Statistics and Economics . Prague.Melandrium. Pages 183–194. ISBN 978-80-87990-12-4. https://msed.vse.cz/msed{_}2017/article/219-Bolotov-Ilya-paper.pdf . Botham , F. W.,
Hunt , E. H. 1987. Wages in Britain During the Industrial Revolution.
TheEconomic History Review . Volume 40. Issue 3. Pages 380–399. ISSN 0013-0117. https://doi.org/0.2307/2596251 . Breusch , T. S.,
Pagan , A. R. 1979. A Simple Test for Heteroscedasticity and Random CoefficientVariation.
Econometrica . Volume 47. Issue 5. Pages 1287–1294. ISSN 0012-9682. https://doi.org/10.2307/1911963 . Broadberry , S.,
Campbell , B. M. S.,
Leeuwen , B. 2013. When Did Britain Industrialise?The Sectoral Distribution of the Labour Force and Labour Productivity in Britain, 1381-1851.
Explorations in Economic History . Volume 50. Issue 1. Pages 16–27. ISSN 0014-4983. https://doi.org/10.1016/j.eeh.2012.08.004 . Cambini , C.,
Caviggioli , F.,
Scellato , G. 2016. Innovation and Market Regulation: Evidencefrom the European Electricity Industry.
Industry and Innovation . Volume 23. Issue 8. Pages734–752. ISSN 1366-2716. https://doi.org/10.1080/13662716.2016.1206464 . Crabb , J. M.,
Johnson , D. K. 2010. Fueling Innovation: The Impact of Oil Prices and CAFEStandards on Energy-Efficient Automotive Technology.
The Energy Journal . Volume 31. Issue 1.Pages 199–216. ISSN 0195-6574. https://doi.org/10.5547/issn0195-6574-ej-vol31-no1-9 . Evan , T. 2014.
Chapters of European Economic History . Prague. Karolinum. ISBN 978-80-246-2829-5. . Fabre , G. 2014. The Real Leap Forward: China’s R&D and Innovation Strategy. In
The Globalizationof Chinese Business . Sawston. Chandos Publishing. Pages 3–25. ISBN 978-1-84334-768-2. https://doi.org/10.1533/9781780634494.1.3 . Fellner , W. 1961. Two Propositions in the Theory of Induced Innovations.
The Economic Journal .Volume 71. Issue 282. Pages 305–308. ISSN 0013-0133. https://doi.org/10.2307/2228769 . Fellner , W. 1971. Empirical Support for the Theory of Induced Innovations.
Quarterly Journalof Economics . Volume 85. Issue 4. Pages 580–604. ISSN 0033-5533. https://doi.org/10.2307/1882269 . Funk , P. 2002. Induced Innovation Revisited.
Economica . Volume 69. Issue 273. Pages 155–171.ISSN 0013-0427. https://doi.org/10.1111/1468-0335.00275 . Gallouj , F.,
Savona , M. 2009. Innovation in Services: A Review of the Debate and a ResearchAgenda.
Journal of Evolutionary Economics . Volume 19. Issue 2. Pages 149–172. ISSN 0936-9937. https://doi.org/10.1007/s00191-008-0126-4 . Griliches , Z.,
Nordhaus , W. D.,
Scherer , F. M. 1989. Patents: Recent Trends and Puzzles.
Brookings Papers on Economic Activity. Microeconomics . Volume 1989. Issue 1989. Pages 291–330. https://doi.org/10.2307/2534723 . Hayami , Y.,
Ruttan , V. W. 1971.
Agricultural Development: An International Perspective . Bal-timore. John Hopkins University Press. ISBN 978-0-8018-1259-0. . 12 icks , J. R. 1963.
The Theory of Wages . Second Edition. London. MacMillan. ISBN 978-1-349-00189-7. https://doi.org/10.1007/978-1-349-00189-7 . Holý , V.,
Šafr , K. 2018. Are Economically Advanced Countries More Efficient in Basic and AppliedResearch?
Central European Journal of Operations Research . Volume 26. Issue 4. Pages 933–950.ISSN 1613-9178. https://doi.org/10.1007/s10100-018-0559-2 . Honda , Y. 1985. Testing the Error Components Model with Non-Normal Disturbances.
Review ofEconomic Studies . Volume 52. Issue 4. Pages 681–690. ISSN 0034-6527. https://doi.org/10.2307/2297739 . Jang , H., Du , X. 2013. Price- and Policy-Induced Innovations: The Case of U.S. Biofuel. Journalof Agricultural and Resource Economics . Volume 38. Issue 3. Pages 299–311. ISSN 1068-5502. . Kelly , M.,
Mokyr , J.,
Gráda , C. Ó. 2014. Precocious Albion: A New Interpretation of the BritishIndustrial Revolution.
Annual Review of Economics . Volume 6. Issue 1. Pages 363–389. ISSN1941-1383. https://doi.org/10.1146/annurev-economics-080213-041042 . Kennedy , C. 1967. On the Theory of Induced Invention - A Reply.
The Economic Journal . Volume77. Issue 308. Pages 958–960. ISSN 0013-0133. https://doi.org/10.2307/2229618 . Landes , D. S. 1998.
The Wealth and Poverty of Nations: Why Some Are So Rich and Some SoPoor . First Edition. New York. W. W. Norton & Company. ISBN 978-0-393-04017-3. https://wwnorton.com/books/The-Wealth-and-Poverty-of-Nations/ . Lee , K.,
Kang , S.-M. 2007. Innovation Types and Productivity Growth: Evidence from KoreanManufacturing Firms.
Global Economic Review . Volume 36. Issue 4. Pages 343–359. ISSN1226-508X. https://doi.org/10.1080/12265080701694512 . Lin , H.-F.,
Lin , S.-M. 2008. Determinants of E-Business Diffusion: A Test of the TechnologyDiffusion Perspective.
Technovation . Volume 28. Issue 3. Pages 135–145. ISSN 0166-4972. https://doi.org/10.1016/j.technovation.2007.10.003 . Maddison , A. 2007.
Contours of the World Economy 1-2030 AD: Essays in Macro-EconomicHistory . Oxford. Oxford University Press. ISBN 978-0-19-922721-1. https://global.oup.com/ukhe/product/contours-of-the-world-economy-1-2030-ad-9780199227204 . Milyaeva , L.,
Fedorkevich , D. 2015. Analysis of Innovative Industrial Competitiveness: Method-ological and Applied Aspects.
Economic Annals-XXI . Volume 7-8. Issue 1. Pages 58–61. ISSN1728-6220. http://soskin.info/en/ea/2015/7-8-1/contents{_}14.html . Newell , R. G.,
Jaffe , A. B.,
Stavins , R. N. 1999. The Induced Innovation Hypothesis andEnergy-Saving Technological Change.
Quarterly Journal of Economics . Volume 114. Issue 3.Pages 941–975. ISSN 0033-5533. https://doi.org/10.1162/003355399556188 . Nordhaus , W. D. 1973. Some Skeptical Thoughts on the Theory of Induced Innovation.
QuarterlyJournal of Economics . Volume 87. Issue 2. Pages 208–219. ISSN 0033-5533. https://doi.org/10.2307/1882184 . OECD . 2015.
Frascati Manual 2015: Guidelines for Collecting and Reporting Data on Research andExperimental Development, The Measurement of Scientific, Technological and Innovation Activities .Technical Report. https://doi.org/10.1787/9789264239012-en . Pomeranz , K. 2001.
The Great Divergence: China, Europe, and the Making of the Modern WorldEconomy . Princeton. Princeton University Press. ISBN 978-0-691-09010-8. https://doi.org/10.2307/j.ctt7sv80 . 13 opp , D. 2002. Induced Innovation and Energy Prices.
American Economic Review . Volume 92.Issue 1. Pages 160–180. ISSN 0002-8282. https://doi.org/10.2307/3083326 . Pritchett , L. 1997. Divergence, Big Time.
Journal of Economic Perspectives . Volume 11. Issue 3.Pages 3–17. ISSN 0895-3309. https://doi.org/10.1257/jep.11.3.3 . Rostow , W. W. 1991.
The Stages of Economic Growth: A Non-Communist Manifesto .Third Edition. Cambridge University Press. ISBN 978-0-521-40928-5. . Ruttan , V. W. 2001.
Technology, Growth, and Development: An Induced Innovation Perspective .New York. Oxford University Press. ISBN 978-0-19-511871-1. https://books.google.com/books/about/Technology{_}Growth{_}and{_}Development.html?id=Z8u7QHsa72IC . Ruttan , V. W.,
Hayami , Y. 1984. Toward a Theory of Induced Institutional Innovation.
TheJournal of Development Studies . Volume 20. Issue 4. Pages 203–223. ISSN 0022-0388. https://doi.org/10.1080/00220388408421914 . Salter , W. E. G.,
Reddaway , W. B. 1966.
Productivity and Technical Change . Cambridge.Cambridge University Press. http://library.wur.nl/WebQuery/clc/293908 . Samuelson , P. A. 1965. A Theory of Induced Innovation along Kennedy-Weisäcker Lines.
TheReview of Economics and Statistics . Volume 47. Issue 4. Pages 343–356. ISSN 0034-6535. https://doi.org/10.2307/1927763 . Savona , M.,
Steinmueller , W. E. 2013. Service Output, Innovation and Productivity: A Time-Based Conceptual Framework.
Structural Change and Economic Dynamics . Volume 27. Pages118–132. ISSN 0954-349X. https://doi.org/10.1016/j.strueco.2013.06.006 . Tomory , L. 2016. Technology in the British Industrial Revolution.
History Compass . Volume 14.Issue 4. Pages 152–167. ISSN 1478-0542. https://doi.org/10.1111/hic3.12306 . Wooldridge , J. M. 2010.
Econometric Analysis of Cross Section and Panel Data . Second Edi-tion. Cambridge. MIT Press. ISBN 978-0-262-23258-6. https://mitpress.mit.edu/books/econometric-analysis-cross-section-and-panel-data-second-editionhttps://mitpress.mit.edu/books/econometric-analysis-cross-section-and-panel-data-second-edition