留言板

尊敬的读者、作者、审稿人, 关于本刊的投稿、审稿、编辑和出版的任何问题, 您可以本页添加留言。我们将尽快给您答复。谢谢您的支持!

姓名
邮箱
手机号码
标题
留言内容
验证码

Polarization or Diffusion? Spatio-temporal Evolution of Urban Technological Innovation Capacity in China’s Five Urban Agglomerations

Jinxian CAO Shengning LI Qingyuan YANG

CAO Jinxian, LI Shengning, YANG Qingyuan, 2022. Polarization or Diffusion? Spatio-temporal Evolution of Urban Technological Innovation Capacity in China’s Five Urban Agglomerations. Chinese Geographical Science, 32(6): 946−962 doi:  10.1007/s11769-022-1309-x
Citation: CAO Jinxian, LI Shengning, YANG Qingyuan, 2022. Polarization or Diffusion? Spatio-temporal Evolution of Urban Technological Innovation Capacity in China’s Five Urban Agglomerations. Chinese Geographical Science, 32(6): 946−962 doi:  10.1007/s11769-022-1309-x

doi: 10.1007/s11769-022-1309-x

Polarization or Diffusion? Spatio-temporal Evolution of Urban Technological Innovation Capacity in China’s Five Urban Agglomerations

More Information
    • 关键词:
    •  / 
    •  / 
    •  / 
    •  
  • Figure  1.  Locations and cities of study area. Three county-level cities (Xiantao, Qianjiang, Tianmen) in MYR are not included because of data limitation. Beijing-Tianjin-Hebei Urban Agglomeration (BTH), Yangtze River Delta Urban Agglomeration (YRD), Pearl River Delta Urban Agglomeration (PRD), Chengdu-Chongqing Urban Agglomeration (CYA), and Urban Agglomeration in Middle Reaches of the Yangtze River (MYR)

    Figure  2.  Intra- and inter-UA technological innovation Gini coefficients during 2003–2016. The abbreviation means the same as in Fig. 1

    Figure  3.  Spatial distribution of urban technological innovation capacity in China’s five UAs in 2003 and 2016. Scale 1 is set for maps of BTH, YRD, CYA, MYR. Scale 2 applies to PRD. The abbreviation means the same as in Fig. 1

    Figure  4.  Spatial distribution of GDP per capita in China’s five urban agglomerations in 2003 and 2016. Scale 1 is set for maps of BTH, YRD, CYA, MYR. Scale 2 applies to PRD

    Figure  5.  Evolution of inter-city co-patents network in China’s five urban agglomerations in 2003 and 2016. Scale 1 is set for maps of BTH, YRD, CYA, MYR. Scale 2 applies to PRD

    Figure  6.  Spatial patterns of innovation spillovers in YRD and MYR. The results of YRD stopped within radius of 400 km and 500 km for MYR, because SAR model failed to convergence after the two ranges since cities outside the scope are rare and spillovers become negligible. 95% CI in these two figures mean 95% confidence intervals

    Table  1.   Measurements and codes of all variables

    VariableCodeMeasurementUnit
    City technological innovation capacity lninnoindex City technological innovation index /
    City population scale lnpopulation Total urban population at the end of the year 104 people
    City economic scale lnGDP Gross regional product 104 yuan (RMB)
    City land scale lnland Administrative region area of built districts km2
    Foreign direct investment (FDI) lnFDI The amount of foreign capital actually used 104 dollars (USD)
    Industrial structure indstru The ratio of added value of the tertiary industry to that of the secondary industry %
    Fiscal investment for S&T govfiscal The proportion of science and technology expenditure in government fiscal expenditure %
    S&T human capital lnscihc The number of scientific research and technical service industry employees at the end of the year 104 people
    Infrastructure facilities Internet infrastructure lninternet Internet users at the end of the year 104 users
    Road infrastructure lnroad Area of paved roads per capita km2
    Notes: all variables related to price are deflated by GDP index. FDI is deflated by exchange rate of RMB against USD. Abbreviations with ‘ln’ means that variables are in logarithms
    下载: 导出CSV

    Table  2.   Moran’s I indexes under geographical distance matrix of China’s five UAs from 2003 to 2016

    YearBTHYRDPRDCYAMYR
    2003 –0.112 (0.195) –0.034 (0.830) –0.176 (0.658) –0.083 (0.744) –0.063 (0.449)
    2004 –0.110 (0.219) –0.033 (0.812) –0.164 (0.730) –0.083 (0.724) –0.061 (0.424)
    2005 –0.109 (0.270) –0.035 (0.848) –0.159 (0.756) –0.082 (0.737) –0.062 (0.429)
    2006 –0.108 (0.316) –0.035 (0.845) –0.151 (0.806) –0.082 (0.728) –0.061 (0.441)
    2007 –0.108 (0.329) –0.034 (0.820) –0.138 (0.888) –0.082 (0.737) –0.063 (0.451)
    2008 –0.107 (0.349) –0.033 (0.813) –0.121 (0.954) –0.084 (0.729) –0.064 (0.484)
    2009 –0.107 (0.362) –0.032 (0.765) –0.111 (0.821) –0.086 (0.715) –0.064 (0.501)
    2010 –0.107 (0.327) –0.029 (0.728) –0.109 (0.784) –0.086 (0.714) –0.064 (0.510)
    2011 –0.108 (0.309) –0.025 (0.640) –0.109 (0.772) –0.087 (0.721) –0.064 (0.505)
    2012 –0.108 (0.297) –0.021 (0.559) –0.109 (0.780) –0.087 (0.723) –0.064 (0.501)
    2013 –0.108 (0.290) –0.017 (0.499) –0.109 (0.780) –0.087 (0.723) –0.064 (0.506)
    2014 –0.108 (0.285) –0.014 (0.456) –0.109 (0.781) –0.086 (0.727) –0.064 (0.503)
    2015 –0.108 (0.282) –0.008 (0.378) –0.108 (0.779) –0.085 (0.733) –0.063 (0.507)
    2016 –0.108 (0.278) –0.004 (0.328) –0.108 (0.782) –0.084 (0.738) –0.063 (0.506)
    Notes: P values are in parentheses. The abbreviation means the same as in Fig. 1
    下载: 导出CSV

    Table  3.   Moran’s I indexes under economic distance matrix of China’s five UAs from 2003 to 2016

    YearBTHYRDPRDCYAMYR
    2003 0.062 (0.000) 0.257 (0.000) 0.026 (0.346) 0.043 (0.085) 0.144 (0.014)
    2004 0.069 (0.000) 0.242 (0.000) 0.031 (0.322) 0.041 (0.071) 0.124 (0.013)
    2005 0.102 (0.000) 0.232 (0.000) 0.031 (0.309) 0.042 (0.069) 0.128 (0.012)
    2006 0.137 (0.000) 0.215 (0.000) 0.028 (0.297) 0.040 (0.069) 0.132 (0.012)
    2007 0.146 (0.000) 0.211 (0.000) 0.023 (0.272) 0.041 (0.079) 0.149 (0.012)
    2008 0.164 (0.000) 0.221 (0.000) 0.010 (0.214) 0.043 (0.086) 0.170 (0.012)
    2009 0.166 (0.000) 0.227 (0.000) 0.000 (0.177) 0.044 (0.097) 0.183 (0.012)
    2010 0.138 (0.000) 0.246 (0.000) –0.002 (0.165) 0.045 (0.130) 0.184 (0.012)
    2011 0.125 (0.000) 0.263 (0.000) –0.003 (0.161) 0.046 (0.114) 0.183 (0.012)
    2012 0.114 (0.000) 0.276 (0.000) –0.002 (0.162) 0.047 (0.117) 0.184 (0.012)
    2013 0.104 (0.000) 0.285 (0.000) –0.001 (0.161) 0.047 (0.117) 0.184 (0.012)
    2014 0.100 (0.000) 0.294 (0.000) 0.000 (0.161) 0.046 (0.112) 0.183 (0.012)
    2015 0.095 (0.000) 0.306 (0.000) 0.000 (0.163) 0.045 (0.109) 0.180 (0.012)
    2016 0.090 (0.000) 0.307 (0.000) 0.001 (0.170) 0.045 (0.104) 0.178 (0.012)
    下载: 导出CSV

    Table  4.   Moran’s I indexes under innovation collaboration matrix of China’s five UAs from 2003 to 2016

    YearBTHYRDPRDCYAMYR
    2003 –0.544 (0.005) –0.199 (0.012) –0.400 (0.096) –0.428 (0.004) –0.727 (0.000)
    2004 –0.538 (0.005) –0.201 (0.011) –0.424 (0.067) –0.439 (0.004) –0.719 (0.000)
    2005 –0.517 (0.008) –0.207 (0.008) –0.434 (0.055) –0.444 (0.003) –0.717 (0.000)
    2006 –0.493 (0.011) –0.211 (0.007) –0.439 (0.046) –0.445 (0.003) –0.719 (0.000)
    2007 –0.487 (0.012) –0.209 (0.008) –0.442 (0.036) –0.434 (0.004) –0.731 (0.000)
    2008 –0.475 (0.015) –0.204 (0.010) –0.438 (0.023) –0.422 (0.005) –0.713 (0.000)
    2009 –0.473 (0.015) –0.197 (0.014) –0.429 (0.020) –0.401 (0.007) –0.685 (0.000)
    2010 –0.492 (0.011) –0.185 (0.024) –0.420 (0.023) –0.391 (0.008) –0.674 (0.000)
    2011 –0.500 (0.010) –0.168 (0.047) –0.412 (0.026) –0.373 (0.011) –0.680 (0.000)
    2012 –0.508 (0.009) –0.156 (0.073) –0.403 (0.031) –0.366 (0.013) –0.679 (0.000)
    2013 –0.514 (0.008) –0.148 (0.095) –0.393 (0.038) –0.366 (0.013) –0.675 (0.000)
    2014 –0.517 (0.008) –0.142 (0.117) –0.387 (0.043) –0.378 (0.010) –0.682 (0.000)
    2015 –0.520 (0.007) –0.127 (0.185) –0.381 (0.049) –0.388 (0.008) –0.685 (0.000)
    2016 –0.523 (0.007) –0.115 (0.254) –0.375 (0.056) –0.399 (0.007) –0.691 (0.000)
    下载: 导出CSV

    Table  5.   Main results of drivers of urban technological innovation capacity

    UABTHYRDPRDCYAMYR
    $ {W}_{ECON} $$ {W}_{INNO} $$ {W}_{ECON} $$ {W}_{INNO} $$ {W}_{INNO} $$ {W}_{ECON} $$ {W}_{INNO} $$ {W}_{ECON} $$ {W}_{INNO} $
    (1)(2)(3)(4)(5)(6)(7)(8)(9)
    lnGDP –0.483***
    (0.148)
    –0.457***
    (0.149)
    –0.594***
    (0.167)
    –0.511***
    (0.165)
    0.241
    (0.240)
    –1.101***
    (0.302)
    –1.074***
    (0.296)
    0.036
    (0.169)
    –0.023
    (0.164)
    lnpopulation 0.844
    (0.721)
    0.816
    (0.716)
    0.464
    (0.510)
    0.333
    (0.508)
    –0.230
    (0.184)
    0.593
    (1.223)
    0.449
    (1.196)
    –1.905***
    (0.717)
    –2.032***
    (0.691)
    lnland 0.289
    (0.521)
    0.162
    (0.504)
    0.076
    (0.376)
    0.172
    (0.375)
    –5.944*
    (3.466)
    0.324
    (1.443)
    0.477
    (1.413)
    –2.896**
    (1.406)
    –2.564*
    (1.359)
    lnFDI –0.002
    (0.038)
    0.001
    (0.038)
    –0.024
    (0.035)
    –0.047
    (0.034)
    –0.229*
    (0.117)
    –0.017
    (0.021)
    –0.019
    (0.021)
    –0.065*
    (0.039)
    –0.067*
    (0.037)
    indstru 0.000
    (0.011)
    0.000
    (0.011)
    0.035**
    (0.017)
    0.040**
    (0.017)
    –0.068**
    (0.032)
    –0.011
    (0.018)
    –0.012
    (0.017)
    –0.032*
    (0.018)
    –0.033*
    (0.018)
    govfiscal 15.584***
    (2.814)
    16.778***
    (2.600)
    12.309***
    (1.600)
    12.106***
    (1.595)
    5.379**
    (2.400)
    9.854**
    (4.569)
    9.229**
    (4.450)
    2.128
    (1.390)
    2.056
    (1.338)
    lnscihc –0.016
    (0.073)
    –0.011
    (0.072)
    0.111*
    (0.066)
    0.076
    (0.065)
    0.288***
    (0.091)
    0.028
    (0.070)
    0.034
    (0.068)
    0.067
    (0.049)
    0.071
    (0.047)
    lninternet –0.014
    (0.020)
    –0.014
    (0.020)
    0.133**
    (0.058)
    0.155***
    (0.059)
    –0.022
    (0.092)
    –0.103***
    (0.038)
    –0.098***
    (0.037)
    –0.107**
    (0.042)
    –0.117***
    (0.041)
    lnroad –0.076
    (0.102)
    –0.089
    (0.101)
    0.367***
    (0.061)
    0.385***
    (0.061)
    0.083
    (0.133)
    0.159**
    (0.073)
    0.166**
    (0.072)
    0.157**
    (0.061)
    0.167***
    (0.059)
    rho 0.128
    (0.115)
    –0.061
    (0.179)
    0.223***
    (0.066)
    –0.163
    (0.141)
    0.058
    (0.131)
    –0.115
    (0.154)
    –0.184
    (0.166)
    –0.058
    (0.083)
    –0.281***
    (0.103)
    Time fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes
    City fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes
    n 182 182 364 364 126 224 224 392 392
    R2 0.0849 0.0754 0.8876 0.8157 0.6546 0.5255 0.5846 0.5675 0.5722
    Notes: 1) The robust standard errors are in parentheses; * P < 0.1, ** P < 0.05, *** P < 0.01; 2) The regression results based on economic distance matrix are not included in PRD since it failed to pass the spatial autocorrelation significance tests under economic distance matrix
    下载: 导出CSV
  • [1] Akamatsu K, 1962. A historical pattern of economic growth in developing countries. The Developing Economies, 1: 3–25. doi:  10.1111/j.1746-1049.1962.tb01020.x
    [2] Ang Y Y, 2018. Domestic flying geese: industrial transfer and delayed policy diffusion in China. The China Quarterly, 234: 420–443. doi:  10.1017/S0305741018000516
    [3] Audretsch D B, Feldman M P, 1996. R&D spillovers and the geography of innovation and production. American Economic Review, 86(3): 630–640.
    [4] Autant-Bernard C, LeSage J P, 2011. Quantifying knowledge spillovers using spatial econometric models. Journal of Regional Science, 51(3): 471–496. doi:  10.1111/j.1467-9787.2010.00705.x
    [5] Balland P A, Boschma R, Frenken K, 2015. Proximity and innovation: from statics to dynamics. Regional Studies, 49(6): 907–920. doi:  10.1080/00343404.2014.883598
    [6] Baycan T, Nijkamp P, Stough R, 2017. Spatial spillovers revisited: innovation, human capital and local dynamics. International Journal of Urban and Regional Research, 41(6): 962–975. doi:  10.1111/1468-2427.12557
    [7] Boschma R, 2005. Proximity and innovation: a critical assessment. Regional Studies, 39(1): 61–74. doi:  10.1080/0034340052000320887
    [8] Bottazzi L, Peri G, 2003. Innovation and spillovers in regions: evidence from European patent data. European Economic Review, 47(4): 687–710. doi:  10.1016/S0014-2921(02)00307-0
    [9] Cai H C, Feng Z J, Zhou W et al. , 2022. Understanding the spatial polarization pattern of technology transfer network in the Guangdong-Hong Kong-Macao Greater Bay area. Growth and Change. doi:  10.1111/grow.12636
    [10] Capello R, 2001. Urban innovation and collective learning: theory and evidence from five metropolitan cities in Europe. In: Fischer M M, Fröhlich J (eds). Knowledge, Complexity and Innovation Systems. Berlin, Heidelberg: Springer, 181–208. doi:  10.1007/978-3-662-04546-6_10
    [11] Carlino G, Kerr W R, 2015. Agglomeration and innovation. In: Duranton G, Henderson J V, Strange W (eds). Handbook of Regional and Urban Economics, 5: 349–404. doi:  10.1016/B978-0-444-59517-1.00006-4
    [12] Chen Y, Li W W, Yi P T, 2020. Evaluation of city innovation capability using the TOPSIS-based order relation method: the case of Liaoning Province, China. Technology in Society, 63: 101330. doi:  10.1016/j.techsoc.2020.101330
    [13] Chen Z Y, Zhang J, 2019. Types of patents and driving forces behind the patent growth in China. Economic Modelling, 80: 294–302. doi:  10.1016/j.econmod.2018.11.015
    [14] Choi H S, Sohn S Y, Yeom H J, 2018. Technological composition of US metropolitan statistical areas with high-impact patents. Technological Forecasting and Social Change, 134: 72–83. doi:  10.1016/j.techfore.2018.05.013
    [15] Crescenzi R, Rodríguez-Pose A, Storper M, 2012. The territorial dynamics of innovation in China and India. Journal of Economic Geography, 12(5): 1055–1085. doi:  10.1093/jeg/lbs020
    [16] Crescenzi R, Rodríguez-Pose A, 2017. The geography of innovation in China and India. International Journal of Urban and Regional Research, 41(6): 1010–1027. doi:  10.1111/1468-2427.12554
    [17] De Groot H L F, Poot J, Smit M J, 2009. Agglomeration externalities, innovation and regional growth: theoretical perspectives and meta-analysis. In: Capello R, Nijkamp P (eds). Handbook of Regional Growth and Development Theories. Northampton, MA: Edward Elgar Publishing, 256–281.
    [18] dos Santos Silvestre B, Dalcol P R T, 2009. Geographical proximity and innovation: evidences from the Campos Basin oil & gas industrial agglomeration: Brazil. Technovation, 29(8): 546–561. doi:  10.1016/j.technovation.2009.01.003
    [19] Drivas K, Economidou C, Karamanis D et al., 2020. Mobility of highly skilled individuals and local innovation activity. Technological Forecasting and Social Change, 158: 120144. doi:  10.1016/j.techfore.2020.120144
    [20] Duan D Z, Du D B, Liu C L et al., 2016. Spatio-temporal evolution of urban innovation structure based on zip code geodatabase: an empirical study from Shanghai and Beijing. Journal of Geographical Sciences, 26(12): 1707–1724. doi:  10.1007/s11442-016-1354-4
    [21] Ellison G, Glaeser E L, Kerr W R, 2010. What causes industry agglomeration? Evidence from coagglomeration patterns. American Economic Review, 100(3): 1195–1213. doi:  10.1257/aer.100.3.1195
    [22] Faber B, 2014. Trade integration, market size, and industrialization: evidence from China’s National Trunk Highway System. The Review of Economic Studies, 81(3): 1046–1070. doi:  10.1093/restud/rdu010
    [23] Fan F, Dai S Z, Zhang K K et al., 2021. Innovation agglomeration and urban hierarchy: evidence from Chinese cities. Applied Economics, 53(54): 6300–6318. doi:  10.1080/00036846.2021.1937507
    [24] Fritsch M, 2002. Measuring the quality of regional innovation systems: a knowledge production function approach. International Regional Science Review, 25(1): 86–101. doi:  10.1177/016001702762039394
    [25] García F, Jin B, Salomon R, 2013. Does inward foreign direct investment improve the innovative performance of local firms? Research Policy, 42(1): 231–244. doi:  10.1016/j.respol.2012.06.005
    [26] Gonçalves E, Almeida E, 2009. Innovation and spatial knowledge spillovers: evidence from Brazilian patent data. Regional Studies, 43(4): 513–528. doi:  10.1080/00343400701874131
    [27] Görg H, Greenaway D, 2004. Much ado about nothing? Do domestic firms really benefit from foreign direct investment? The World Bank Research Observer, 19(2): 171–197. doi:  10.1093/wbro/lkh019
    [28] Greunz L, 2005. Intra- and inter-regional knowledge spillovers: evidence from European regions. European Planning Studies, 13(3): 449–473. doi:  10.1080/09654310500089746
    [29] Hamidi S, Zandiatashbar A, 2019. Does urban form matter for innovation productivity? A national multi-level study of the association between neighbourhood innovation capacity and urban sprawl. Urban Studies, 56(8): 1576–1594. doi:  10.1177/0042098018767002
    [30] Howells J, Bessant J, 2012. Introduction: innovation and economic geography: a review and analysis. Journal of Economic Geography, 12(5): 929–942. doi:  10.1093/jeg/lbs029
    [31] Jaffe A B, Trajtenberg M, Henderson R, 1993. Geographic localization of knowledge spillovers as evidenced by patent citations. The Quarterly journal of Economics, 108(3): 577–598. doi:  10.2307/2118401
    [32] Kaygalak I, Reid N, 2016. Innovation and knowledge spillovers in Turkey: the role of geographic and organizational proximity. Regional Science Policy & Practice, 8(1–2): 45–60. doi:  10.1111/rsp3.12072
    [33] Koch L, Simmler M, 2020. How important are local knowledge spillovers of public R&D and what drives them? Research Policy, 49(7): 104009. doi:  10.1016/j.respol.2020.104009
    [34] Kou Zonglai, Liu Xueyue, 2017. FIND Report on City and Industrial Innovation in China (2017), Fudan Institute of Industrial Development, School of Economics, Fudan University. Available at: http://kouzonglai.blog.caixin.com/archives/176063. (in Chinese)
    [35] Lakshmanan T R, Anderson W P, Song Y, 2015. Rise of megalopolis as a mega knowledge region: interactions of innovations in transport, information, production, and organizations. In: Nijkamp P, Rose A, Kourtit K (eds). Regional Science Matters, Cham: Springer,373–398. doi:  10.1007/978-3-319-07305-7_18
    [36] Li Y, 2020. Innovative firms and high-tech industrial hubs in China. In: Oqubay A, Lin J Y (eds). The Oxford Handbook of Industrial Hubs and Economic Development. Oxford: Oxford University Press.
    [37] Lin M, Kwan Y K, 2016. FDI technology spillovers, geography, and spatial diffusion. International Review of Economics & Finance, 43: 257–274. doi:  10.1016/j.iref.2016.02.014
    [38] Liu Y L, Zhang X H, Pan X Y et al., 2020. The spatial integration and coordinated industrial development of urban agglomerations in the Yangtze River Economic Belt, China. Cities, 104: 102801. doi:  10.1016/j.cities.2020.102801
    [39] Lyu L, Sun F X, Huang R, 2019. Innovation-based urbanization: evidence from 270 cities at the prefecture level or above in China. Journal of Geographical Sciences, 29(8): 1283–1299. doi:  10.1007/s11442-019-1659-1
    [40] Mattes J, 2012. Dimensions of proximity and knowledge bases: innovation between spatial and non-spatial factors. Regional Studies, 46(8): 1085–1099. doi:  10.1080/00343404.2011.552493
    [41] Miguélez E, Moreno R, 2015. Knowledge flows and the absorptive capacity of regions. Research Policy, 44(4): 833–848. doi:  10.1016/j.respol.2015.01.016
    [42] Min S J, Kim J, Sawng Y W, 2020. The effect of innovation network size and public R&D investment on regional innovation efficiency. Technological Forecasting and Social Change, 155: 119998. doi:  10.1016/j.techfore.2020.119998
    [43] Moreno R, Paci R, Usai S, 2005. Spatial spillovers and innovation activity in European regions. Environment and Planning A:Economy and Space, 37(10): 1793–1812. doi:  10.1068/a37341
    [44] Ning L T, Wang F, Li J, 2016. Urban innovation, regional externalities of foreign direct investment and industrial agglomeration: evidence from Chinese cities. Research Policy, 45(4): 830–843. doi:  10.1016/j.respol.2016.01.014
    [45] National Bureau of Statistics of China, 2001–2017. China City Statistical Yearbook 2001–2017. China: China Statistics Press. (in Chinese)
    [46] National Bureau of Statistics of China, 2001–2017. China Statistical Yearbook 2001−2017. China: China Statistics Press. (in Chinese)
    [47] Paci R, Marrocu E, Usai S, 2014. The complementary effects of proximity dimensions on knowledge spillovers. Spatial Economic Analysis, 9(1): 9–30. doi:  10.1080/17421772.2013.856518
    [48] Qin Y, 2017. ‘No county left behind?’ The distributional impact of high-speed rail upgrades in China. Journal of Economic Geography, 17(3): 489–520. doi:  10.1093/jeg/lbw013
    [49] Schankerman M, Pakes A, 1986. Estimates of the value of patent rights in European countries during the post-1950 period. The Economic Journal, 96(384): 1052–1076. doi:  10.2307/2233173
    [50] Shang Q Y, Poon J P H, Yue Q T, 2012. The role of regional knowledge spillovers on China’s innovation. China Economic Review, 23(4): 1164–1175. doi:  10.1016/j.chieco.2012.08.004
    [51] Sheng Y W, Zhao J L, Zhang X B et al., 2019. Innovation efficiency and spatial spillover in urban agglomerations: a case of the Beijing-Tianjin-Hebei, the Yangtze River Delta, and the Pearl River Delta. Growth and Change, 50(4): 1280–1310. doi:  10.1111/grow.12329
    [52] Song H S, Zhang M, 2017. Spatial spillovers of regional innovation: evidence from Chinese Provinces. Emerging Markets Finance and Trade, 53(9): 2104–2122. doi:  10.1080/1540496X.2017.1284061
    [53] Squicciarini M P, Voigtländer N, 2015. Human capital and industrialization: evidence from the age of enlightenment. The Quarterly Journal of Economics, 130(4): 1825–1883. doi:  10.1093/qje/qjv025
    [54] Tan D D, Cheng C, Lei M J et al., 2017. Spatial distributions and determinants of regional innovation in China: evidence from Chinese metropolitan data. Emerging Markets Finance and Trade, 53(6): 1442–1454. doi:  10.1080/1540496X.2017.1283215
    [55] Tang S, Zhang J X, Niu F Q, 2020. Spatial-temporal evolution characteristics and countermeasures of urban innovation space distribution: an empirical study based on data of Nanjing high-tech enterprises. Complexity, 2020: 2905482. doi:  10.1155/2020/2905482
    [56] Tojeiro-Rivero D, Moreno R, 2019. Technological cooperation, R&D outsourcing, and innovation performance at the firm level: the role of the regional context. Research Policy, 48(7): 1798–1808. doi:  10.1016/j.respol.2019.04.006
    [57] Vernon R, 1966. International investment and international trade in the product cycle. The Quarterly Journal of Economics, 80(2): 190–207. doi:  10.2307/1880689
    [58] Wang J T, Cai S Y, 2020. The construction of high-speed railway and urban innovation capacity: based on the perspective of knowledge Spillover. China Economic Review, 63: 101539. doi:  10.1016/j.chieco.2020.101539
    [59] Wang S, Shi X, Wang T et al. , 2022. Nonlinear spatial innovation spillovers and regional open innovation: evidence from China. R&D Management. doi:  10.1111/radm.12527
    [60] Wang Yue, Wang Chengyun, Mao Xiyan et al., 2021. Spatial pattern and benefit allocation in regional collaborative innovation of the Yangtze River Delta, China. Chinese Geographical Science, 31(5): 900–914. doi:  10.1007/s11769-021-1224-6
    [61] Wang Y D, Ning L T, Li J et al., 2016a. Foreign direct investment spillovers and the geography of innovation in Chinese regions: the role of regional industrial specialization and diversity. Regional Studies, 50(5): 805–822. doi:  10.1080/00343404.2014.933800
    [62] Wang Z Y, Cheng Y Q, Ye X Y et al., 2016b. Analyzing the space–time dynamics of innovation in china: ESDA and spatial panel approaches. Growth and Change, 47(1): 111–129. doi:  10.1111/grow.12115
    [63] Wilson D J, 2009. Beggar thy neighbor? The in-state, out-of-state, and aggregate effects of R&D tax credits. The Review of Economics and Statistics, 91(2): 431–436. doi:  10.1162/rest.91.2.431
    [64] Xie Q J, Su J, 2021. The spatial-temporal complexity and dynamics of research collaboration: Evidence from 297 cities in China (1985−2016). Technological Forecasting and Social Change, 162: 120390. doi:  10.1016/j.techfore.2020.120390
    [65] Yan Y Y, Wu Z C, 2020. Regional innovation distribution and its dynamic evolution: policy impact and spillover effect—Based on the perspective of innovation motivation. PLoS One, 15(7): e0235828. doi:  10.1371/journal.pone.0235828
    [66] Yang W Y, Fan F, Wang X L et al., 2021. Knowledge innovation network externalities in the Guangdong-Hong Kong-Macao Greater Bay Area: borrowing size or agglomeration shadow. Technology Analysis & Strategic Management, 1–18. doi:  10.1080/09537325.2021.1940922
    [67] Yao L, Li J, Li J, 2020. Urban innovation and intercity patent collaboration: a network analysis of China’s national innovation system. Technological Forecasting and Social Change, 160: 120185. doi:  10.1016/j.techfore.2020.120185
    [68] Zhang G Y, Zhao S K, Xi Y J et al., 2018. Relating science and technology resources integration and polarization effect to innovation ability in emerging economies: an empirical study of Chinese enterprises. Technological Forecasting and Social Change, 135: 188–198. doi:  10.1016/j.techfore.2017.09.009
    [69] Zhang K H, 2014. How does foreign direct investment affect industrial competitiveness? Evidence from China. China Economic Review, 30: 530–539. doi:  10.1016/j.chieco.2013.08.003
    [70] Zhang W K, Tian X L, Yu A, 2020. Is high-speed rail a catalyst for the fourth industrial revolution in China? Story of enhanced technology spillovers from venture capital. Technological Forecasting and Social Change, 161: 120286. doi:  10.1016/j.techfore.2020.120286
    [71] Zhou Y, Guo Y Z, Liu Y S, 2018. High-level talent flow and its influence on regional unbalanced development in China. Applied Geography, 91: 89–98. doi:  10.1016/j.apgeog.2017.12.023
  • [1] Chunliang XIU, Ying JIN.  Issues with Spatial Scale in Urban Research . Chinese Geographical Science, 2022, 32(3): 373-388. doi: 10.1007/s11769-022-1274-4
    [2] Weiwei CHEN, Lingjian DUANMU, Yang QIN, Hongwu YANG, Jing FU, Chengwei LU, Wei FENG, Li GUO.  Lockdown-induced Urban Aerosol Change over Changchun, China During COVID-19 Outbreak with Polarization LiDAR . Chinese Geographical Science, 2022, 32(5): 824-833. doi: 10.1007/s11769-022-1303-3
    [3] Yan CHEN, Miaomiao XIE, Bin CHEN, Huihui WANG, Yali TENG.  Surface Regional Heat (Cool) Island Effect and Its Diurnal Differences in Arid and Semiarid Resource-based Urban Agglomerations . Chinese Geographical Science, 2022, (): 1-13. doi: 10.1007/s11769-022-1324-y
    [4] Kan ZHOU, Jianxiong WU, Jie FAN, Hanchu LIU.  Drivers of Regional Environmental Pollution Load and Zoning Control: A Case Study of the Yangtze River Economic Belt, China . Chinese Geographical Science, 2022, 32(1): 31-48. doi: 10.1007/s11769-022-1257-5
    [5] Wenwen QIU, Zhangbao ZHONG, Zhaoliang LI.  Agricultural Non-point Source Pollution in China: Evaluation, Convergence Characteristics and Spatial Effects . Chinese Geographical Science, 2021, 31(3): 571-584. doi: 10.1007/s11769-021-1200-1
    [6] LI Xiaolan, HU Xiaoming, SHI Shuaiyi, SHEN Lidu, LUAN Lan, MA Yanjun.  Spatiotemporal Variations and Regional Transport of Air Pollutants in Two Urban Agglomerations in Northeast China Plain . Chinese Geographical Science, 2019, 29(6): 917-933. doi: 10.1007/s11769-019-1081-8
    [7] HE Xingyuan, XU Sheng, XU Wenduo, CHEN Wei, HUANG Yanqing, WEN Hua.  Effects of Climate Warming on Phenological Characteristics of Urban Forest in Shenyang City, China . Chinese Geographical Science, 2016, 26(1): 1-9. doi: 10.1007/s11769-015-0782-x
    [8] ZHANG Wei, JIANG Jingang, ZHU Yubi.  Change in Urban Wetlands and Their Cold Island Effects in Response to Rapid Urbanization . Chinese Geographical Science, 2015, 25(4): 462-471. doi: 10.1007/s11769-015-0764-z
    [9] WANG Kaiyong, DENG Yu, SUN Daowei, SONG Tao.  Evolution and Spatial Patterns of Spheres of Urban Influence in China . Chinese Geographical Science, 2014, 0(1): 126-136. doi: 10.1007/s11769-013-0635-4
    [10] LU Lachang, HUANG Ru.  Urban Hierarchy of Innovation Capability and Inter-city Linkages of Knowledge in Post-reform China . Chinese Geographical Science, 2012, 22(5): 602-616.
    [11] QI Weifeng, FANG Chuanglin, SONG Jitao.  Measurement and Spatial Distribution of Urban Agglomeration Industrial Compactness in China . Chinese Geographical Science, 2008, 18(4): 291-299. doi: 10.1007/s11769-008-0291-2
    [12] FANG Chuanglin, SONG Jitao, SONG Dunjiang.  Stability of Spatial Structure of Urban Agglomeration in China Based on Central Place Theory . Chinese Geographical Science, 2007, 17(3): 193-202. doi: 10.1007/s11769-007-0193-8
    [13] XU Jiangang, LIAO Banggu, SHEN Qing, ZHANG Feng, MEI Anxin.  Urban Spatial Restructuring in Transitional Economy——Changing Land Use Pattern in Shanghai . Chinese Geographical Science, 2007, 17(1): 19-27. doi: 10.1007/s11769-007-0019-8
    [14] GUO Rongchao, MIAO Changhong, LI Xuexin, CHEN Deguang.  Eco-spatial Structure of Urban Agglomeration . Chinese Geographical Science, 2007, 17(1): 28-33. doi: 10.1007/s11769-007-0028-7
    [15] LIU Jisheng, CHEN Yanguang.  Spatial Autocorrelation and Localization of Urban Development . Chinese Geographical Science, 2007, 17(1): 34-39. doi: 10.1007/s11769-007-0034-9
    [16] ZHANG Luo-cheng, ZHU Tian-ming, YAO Shi-mou.  EVOLVEMENT OF URBAN SPATIAL STRUCTURE AND MAIN DRIVES IN WUXI, CHINA . Chinese Geographical Science, 2005, 15(4): 309-314.
    [17] ZHOU De-min, XU Jian-chun, John RADKE, MU Lan.  A SPATIAL CLUSTER METHOD SUPPORTED BY GIS FOR URBAN-SUBURBAN-RURAL CLASSIFICATION . Chinese Geographical Science, 2004, 14(4): 337-342.
    [18] SONG Yu-xiang, ZHENG Hong-bo.  STUDY ON THE OPTIMUM PATTERN OF DAQING URBAN SPATIAL STRUCTURE . Chinese Geographical Science, 2002, 12(3): 206-211.
    [19] 宁越敏, 严重敏.  THE UNEVEN DEVELOPMENT AND SPATIAL DIFFUSION OF CHINESE CENTRAL CITIES . Chinese Geographical Science, 1994, 4(2): 119-128.
    [20] 李若建.  ON SPATIAL AND TEMPORAL CHANGES OF PRESENT URBAN POPULATION IN CHINA . Chinese Geographical Science, 1993, 3(4): 326-333.
  • caojinxian-supplementary.pdf
  • 加载中
图(6) / 表ll (5)
计量
  • 文章访问数:  542
  • HTML全文浏览量:  267
  • PDF下载量:  18
  • 被引次数: 0
出版历程
  • 收稿日期:  2021-11-17
  • 录用日期:  2022-03-15
  • 网络出版日期:  2022-11-03
  • 刊出日期:  2022-11-05

Polarization or Diffusion? Spatio-temporal Evolution of Urban Technological Innovation Capacity in China’s Five Urban Agglomerations

doi: 10.1007/s11769-022-1309-x

English Abstract

CAO Jinxian, LI Shengning, YANG Qingyuan, 2022. Polarization or Diffusion? Spatio-temporal Evolution of Urban Technological Innovation Capacity in China’s Five Urban Agglomerations. Chinese Geographical Science, 32(6): 946−962 doi:  10.1007/s11769-022-1309-x
Citation: CAO Jinxian, LI Shengning, YANG Qingyuan, 2022. Polarization or Diffusion? Spatio-temporal Evolution of Urban Technological Innovation Capacity in China’s Five Urban Agglomerations. Chinese Geographical Science, 32(6): 946−962 doi:  10.1007/s11769-022-1309-x
    • Comprised of cities’ synergies of interactive growth, urban agglomerations (‘UAs’) are important innovation hubs in the 21st Century (Capello, 2001). Among the ‘centers of creativity’, China as the largest and most innovative emerging country, provides a unique context for the study of regional innovation (Xie and Su, 2021). Over the past few decades, booming innovation activities in Chinese urban agglomerations can be attributed to concentration of innovation factor inflows, such as knowledge, talents, foreign direct investment (Lakshmanan et al., 2015; Miguélez and Moreno, 2015; Ning et al., 2016; Drivas et al., 2020), and especially the specialization of high-tech industrial hubs (Li, 2020). The agglomeration of innovation factors and industrial specialization make some UAs in China transit to inter-city collaborative innovation and thus obtaining excessive benefits (Wang et al., 2021), however, some UAs still undergo increasing innovation intra-gaps and demonstrate extremely uneven innovation distribution (Yang et al., 2021).

      This phenomenon can trace back to tensions and interactions between two effects: innovation diffusion effect and innovation polarization effect. On the one hand, large cities may improve their neighboring cities’ innovation capacity via technological cooperation, industrial linkage and outsourcing activities (Tojeiro-Rivero and Moreno, 2019; Sheng et al., 2019). Due to externality of R&D investment and knowledge spillover, the innovation development in one city and that of its surrounding cities in the UA may be spatially intercorrelated (Gonçalves and Almeida, 2009; Shang et al., 2012; Koch and Simmler, 2020). Here, we consider this type of positive innovation spillover as innovation diffusion effect. On the other hand, cities with high innovation capacity can not always boost innovation growth in other cities in the UA, because high-innovative cities have absolute advantages in attracting resources (Zhang et al., 2018). In this case, innovation gaps within UA may not decline, by contrast, the innovation resources crowding-out effects and market rivalry may lead to negative innovation spillovers (Wang et al., 2022). For example, Wilson (2009) found that the success of research and development (R&D) inducement policy in US states is at the expense of attracting R&D spending away from neighboring states. In this paper, we call the negative innovation spillover as innovation polarization effect.

      The innovation diffusion effect and innovation polarization effect generally occur hand in hand in shaping urban innovation distribution, but why one effect predominates the other in a given time and space is still unclear (Crescenzi et al., 2012; Yan and Wu, 2020). A few literatures examined the spatial polarization and diffusion patterns in certain Chinese UAs or several large cities. For instance, Duan et al. (2016) taking Shanghai and Beijing in China as study area, and found that the spatial distribution of innovation in Shanghai shifted from single-core to a multi-core structure, while Beijing maintained a single-core oriented structure. They attributed this disparity to regional innovation patterns and more balanced development of innovation activities and resources in Shanghai. Similarly, Tang et al. (2020) focused on urban innovation space distribution in Nanjing and found that it diffused at the initial and polarized later. They also noted that excessive agglomeration of urban innovation space may lead to unbalanced development. Cai et al. (2022) found prominent polarization pattern and Matthew effect in Guangdong-Hong Kong-Macao Greater Bay area, with technology transfer mainly generated by interactions of core city nodes like Shenzhen and Guangzhou.

      From the above brief review of previous literatures, we know that spatial patterns of innovation activities have been widely discussed in Chinese cities and provinces (Ning et al., 2016; Song and Zhang, 2017; Tan et al., 2017; Chen et al., 2020), while researches at urban agglomeration level, particularly comparative studies among UAs, are still scarce. Therefore, it is of great significance to investigate urban innovation capacity development of Chinese UAs, where possess salient variations both in innovation activities and socioeconomic factors. In this paper, we aim to answer the following two questions: 1) before estimating innovation diffusion effect and innovation polarization effect, we need to know: in the past decades, what are the spatio-temporal changes of technological innovation capacity within and between Chinese UAs? What factors may drive their development? 2) Are the spatial patterns of technological innovation capacity in Chinese UAs polarized or diffused? What are possible reasons for the differences?

      In practice, we first build innovation Gini coefficient to measure innovation gaps intra- and inter-Chinese UAs, and simultaneously depict spatio-temporal evolution process of technological innovation capacity of Chinese UAs by applying spatial autocorrelation tests. Second, spatial autoregression model (SAR) is employed to explore the driving factors of that evolution processes. Third, we further displayed the polarization-diffusion patterns of technological innovation capacity of Chinese UAs by detecting potential innovation spillover effects, and meanwhile, explain possible mechanisms combining the results in SAR.

    • We review literatures related to urban innovation from two perspectives. The first involves factors influencing urban innovation progress and their polarization and diffusion patterns. The second is about measurements of technological innovation concentration and linkages between UAs.

      For the reasons of concentration of innovative activities, literatures on urban agglomeration economies originally rested on externalities in the production of goods and services, since innovative activities take place inside industrial clusters (Carlino and Kerr, 2015). Following the research of Ellison et al. (2010), scholars extended innovation agglomeration in industrial clusters to the extent to which industry pairs share goods, workers, and knowledge. They found evidence of innovation increase derived from knowledge spillovers of R&D expenditures of local area and its neighboring cities (Autant-Bernard and LeSage, 2011; Wang et al., 2016b; Chen and Zhang, 2019), especially public investments funded by government (Koch and Simmler, 2020; Min et al., 2020). Meanwhile, in their analysis about the knowledge production function, human capital comes a crucial factor deciding whether knowledge can be transformed into innovation capacity (Squicciarini and Voigtländer, 2015). While some literatures stressed that the spread of human capital and R&D investments can be highly selective and exhibit different patterns that may induce innovation polarization, when they reported positive impacts of talents flow on urban innovation output (Crescenzi et al., 2012; Crescenzi and Rodríguez-Pose, 2017; Lyu et al., 2019). In comparison, studies on infrastructure facilities and amenities, e.g., high-speed rail, supported that road infrastructure construction may indirectly accelerate innovation diffusion and knowledge spillover (Zhang et al., 2020; Wang and Cai, 2020), and the popularization of internet may decentralize the spatial network in terms of soft innovation (Howells and Bessant, 2012; Baycan et al., 2017; Yang et al., 2021).

      Except for factors aforementioned, foreign direct investment (FDI), industrial structure, as well as city scale are also frequently considered as determinants of urban innovation growth. In an open economy, FDI can promote technology spillover and knowledge spillover as it incentivizes local firms’ learning through imitation (Görg and Greenaway, 2004; Zhang, 2014), but it may also crowd out local funds and talents of local enterprises, offsetting positive innovation spillovers (García et al., 2013). A few literatures linked FDI and regional industrial structure somehow. They argued that innovation spillovers of FDI are contingent on the intensity of industrial agglomeration within and across cities (Ning et al., 2016; Wang et al., 2016a). In addition, urban scale which measured by a series of indicators, such as population scale, economy scale, as well as land scale, are proved to be associated with regional innovation. For instance, Fan et al. (2021) found evidence about non-linear threshold influences of city scale and positive impacts of city income level on innovation agglomeration. The expansion of urban area provide space for sprawl and help to reduce congestion cost with respect to city scale (Hamidi and Zandiatashbar, 2019). Besides, dense population in agglomeration economies can also facilitate knowledge spillovers (De Groot et al., 2009). On the basis of these studies, we introduce fiscal investment for science and technology (S&T), S&T human capital, infrastructure facilities, FDI, and industrial structure as drivers for city innovation capacity in spatial analysis, and take city scale as control variable.

      How to measure innovation and innovation linkages is another concern for estimating regional innovation spillover. Some believe that innovation activities spread along with geographical proximity because economic activities and industrial specialization generally cluster in certain geographical zones (Audretsch and Feldman, 1996; dos Santos Silvestre and Dalcol, 2009; Kaygalak and Reid, 2016). Some highlight disparities of technological distance, social proximity, cognitive gaps (i.e., values, rules, and cultures) and institutional differences between areas will also hinder innovation spillover (Boschma, 2005; Mattes, 2012; Paci et al., 2014; Balland et al., 2015; Yang et al., 2021). Accordingly, in order to capture innovation linkages, literatures use geographical distance, socioeconomic distance, cultural distance and innovation input-output-related measurements (e.g., R&D flow, R&D personnel, patents) when handling spatial analysis (Jaffe et al., 1993; Wang et al., 2016b; Sheng et al., 2019). Recently, featuring in estimation of innovation output and innovation interactions, co-patents has gained increasing popularity in inter-city innovation narratives (Yao et al., 2020). Based on existing literature, this paper builds three spatial weighted matrixes, including geographical distance matrix, economic distance matrix and innovation collaboration matrix to define regional innovation linkages. Furthermore, we apply city innovation index calculated from patent value (Kou and Liu, 2017), which is operationalized by proportion of renewed ages and renewal fees of patents (Schankerman and Pakes, 1986), to measure city technological innovation capacity.

      All in all, previous literatures provide insightful theoretical foundation for regional innovation research. Nevertheless, they care less on divergences of driving factors for innovation patterns among UAs. In this paper, we try to make twofold contributions: 1) we retrospect the evolution process and main drivers of technological innovation capacity of five UAs in China during past 14 years; 2) we compare the innovation spatial patterns among five UAs from the theoretical lens of innovation diffusion effect and innovation polarization effect, for better understanding the phenomenon of regional innovation inequality and balance.

    • The five Chinese urban agglomerations in this paper include Beijing-Tianjin-Hebei Urban Agglomeration (BTH), Yangtze River Delta Urban Agglomeration (YRD), Pearl River Delta Urban Agglomeration (PRD), Chengdu-Chongqing Urban Agglomeration (CYA), and Urban Agglomeration in Middle Reaches of the Yangtze River (MYR), which are comprised of 92 prefecture-level divisions in total and 13 cities, 26 cities, 9 cities, 16 cities, and 28 cities respectively (Table A1). Spanning three geographical locations in eastern, central and western China (Fig. 1), the five UAs have accounted for approximately 55% of China’s GDP and 40% permanent population since 2003.

      Figure 1.  Locations and cities of study area. Three county-level cities (Xiantao, Qianjiang, Tianmen) in MYR are not included because of data limitation. Beijing-Tianjin-Hebei Urban Agglomeration (BTH), Yangtze River Delta Urban Agglomeration (YRD), Pearl River Delta Urban Agglomeration (PRD), Chengdu-Chongqing Urban Agglomeration (CYA), and Urban Agglomeration in Middle Reaches of the Yangtze River (MYR)

    • The observation period of this paper is set from 2003 to 2016 due to administrative division adjustment and data limitation (Data of dependent variable, urban technological innovation index, is currently published till 2016). Our database contains three datasets: 1) Data of spatial weighted matrixes. The geographical distance between cities are calculated by ArcGIS 10.6 through longitude and latitude of every city, fetched from Baidu Map API (http://api.map.baidu.com/). We construct the economic distance matrix by differences of annual Gross Regional Product (GRP) per capita between cities, which are from China City Statistical Yearbook (20022017) (National Bureau of Statistics, 2001–2017). Data of inter-city co-patents are requested from website of China National Intellectual Property Administration (CNIPA) (http://pss-system.cnipa.gov.cn/). 2) City technological innovation index. We use innovation index in the Report on City and Industrial Innovation in China released by China Center for Economic Studies of Fudan University (CCES). 3) Data of driving factors and city characteristics. The data of independent variables in this paper are mainly from China City Statistical Yearbook (20012017) (National Bureau of Statistics, 2001–2017) and China Statistical Yearbook (20012017) (National Bureau of Statistics, 2001–2017). Descriptive statistics for all variables are shown in Table A2–Table A6.

    • In this paper, we use Gini coefficients to compare the intra- and inter- urban agglomeration innovation gaps, referring to Choi et al. (2018). This coefficient is originally applied to measure income gaps, but to describe spatial distributions of industries and innovation activities later (Audretsch and Feldman, 1996). Greater value of this coefficient means larger innovation gaps between urban agglomerations. The formula is presented below:

      $$ {G}_{k}=\frac{1}{2{n}^{2}\mu }\sum _{i=1}^{n}\sum _{\mathrm{j}=1}^{n}|{x}_{ik}-{x}_{jk}| $$ (1)

      where Gk means the Gini coefficient of year k. xik and xjk indicates technological innovation index of city or UA i and city or UA j, respectively. n is the number of cities or UAs. μ denotes the average value of technological innovation index of cities or UAs. The coefficient ranges from 0 to 1.

    • We define three types of matrixes, including geographical distance matrix, economic distance matrix, and innovation collaboration matrix, to conduct spatial analysis referring to previous literature (Sheng et al., 2019; Yao et al., 2020). The spatial weighted matrixes are as follow:

      Geographical distance matrix (${W}_{{G}{E}{O}}$):

      $$ {W}_{ij}=(1/{d}_{ij})/\left[\sum _{j=1}^{n}(1/{d}_{ij})\right] $$ (2)

      where dij the geographical distance between city i and city j, calculated by latitude and longitude of their city government. Higher spatial weights denote shorter geographical distance.

      Economic distance matrix (${W}_{{E}{C}{O}{N}}$):

      $$ {W}_{i j}=(1/|{\overline{a}}_{i}-{\overline{a}}_{j}\left|\right)/\left[{\sum _{j=1}^{n}}(1/|{\overline{a}}_{i}-{\overline{a}}_{j}\left|\right)\right] $$ (3)

      where $\overline{a}_i $ and $\overline{a}_j$ are the means of deflated GDP per capita of city i and city j respectively. Greater weights in this matrix indicate smaller income gaps between cities.

      Innovation collaboration matrix (${W}_{{I}{N}{N}{O}}$):

      $$ {W}_{i j}=\left({patent}_{i j}\right)/ \left[\sum _{j=1}^{n}\left({patent}_{i j}\right)\right] $$ (4)

      where patentij represents the number of co-patents between city i and city j. Greater weights mean closer innovation connection.

      After setting up the spatial weighted matrixes, we adopt the Moran’s I index to test the spatial dependence of technological innovation capacity between cities. On the one hand, significant Moran’s I indexes are one of the premises of conducting spatial econometrics in the next part. On the other hand, Moran’s I index can show us that inter-city innovation linkage can be explained by which type of spatial distance. The equation of Moran’s I is shown as follow:

      $$ I=\frac{{\displaystyle\sum _{i=1}^{n}}{\displaystyle\sum _{j=1}^{n}}{W}_{i j}({x}_{i}-\overline{x})({x}_{j}-\overline{x})}{{S}^{2}{\displaystyle\sum _{i=1}^{n}}{\displaystyle\sum _{j=1}^{n}}{W}_{ij}} $$ (5)

      where xi and xj are the technological innovation indexes of city i and city j respectively. $\overline{x} $ is the average value of city technological innovation capacity in certain UA. S2=∑ni=1(xix)2/n is the sample variance of this UA. Wij is the element of spatial weighted matrixes and n indicates the number of cities in this UA. For this index, if high values clustered with high values, low values clustered with low values, it represents positive spatial correlation, vice versa. If city technological innovation capacity distributes randomly, the Moran’s I index would not be salient.

    • In this section, we apply a two-way fixed Spatial Autoregressive (SAR) specification to model the regional interconnectivity patterns of city technological innovation capacity and its driving factors. In consistent with our research aim, this model enables innovation spillover and transmission of innovation capacity across proximate territories (Paci et al., 2014), thus can detecting the innovation polarization effect and innovation diffusion effect. The model is as follow:

      $$ \begin{split} \mathrm{l}\mathrm{n}{innoindex}_{it}=&{\beta }_{0}+\rho \sum _{j\ne i}^{n}{W}_{ij}\mathrm{l}\mathrm{n}{innoindex}_{it}+{\beta }_{1}{\mathrm{l}\mathrm{n}FDI}_{it}+\\ &{\beta }_{2}{indstru}_{it}+{\beta }_{3}{govfiscal}_{it}+{\beta }_{4}{\mathrm{l}\mathrm{n}scihc}_{it}+\\ &{\beta }_{5}\mathrm{l}\mathrm{n}{internet}_{it}+{{\beta }_{6}{road}_{it}}+{\beta }_{7}{X}_{it}+\\ &{\alpha }_{i}+{\gamma }_{t}+{\varepsilon }_{it}\\[-10pt] \end{split}$$ (6)

      where lninnoindexit is the technological innovation index of city i in year t. Coefficient ρ represents the innovation spillover from its surrounding cities. If ρ is positive, it indicates innovation diffusion effect, otherwise, it represents innovation polarization effect. β1β6 are the coefficients of driving factors, including foreign direct investment (FDI), industrial structure, fiscal investment for S&T, S&T human capital, and infrastructure facilities. β7 is the coefficient of all control variables, including city population scale, economic scale, and land scale. αi is the city fixed effect. γt is the year fixed effect. εit is the error term. β0 is the constant term. The measurements and codes for all variables are shown in Table 1.

      Table 1.  Measurements and codes of all variables

      VariableCodeMeasurementUnit
      City technological innovation capacity lninnoindex City technological innovation index /
      City population scale lnpopulation Total urban population at the end of the year 104 people
      City economic scale lnGDP Gross regional product 104 yuan (RMB)
      City land scale lnland Administrative region area of built districts km2
      Foreign direct investment (FDI) lnFDI The amount of foreign capital actually used 104 dollars (USD)
      Industrial structure indstru The ratio of added value of the tertiary industry to that of the secondary industry %
      Fiscal investment for S&T govfiscal The proportion of science and technology expenditure in government fiscal expenditure %
      S&T human capital lnscihc The number of scientific research and technical service industry employees at the end of the year 104 people
      Infrastructure facilities Internet infrastructure lninternet Internet users at the end of the year 104 users
      Road infrastructure lnroad Area of paved roads per capita km2
      Notes: all variables related to price are deflated by GDP index. FDI is deflated by exchange rate of RMB against USD. Abbreviations with ‘ln’ means that variables are in logarithms

      In order to further explore the spatial patterns of regional innovation spillover, we divide the spatial weighted matrix according to geographical scope of UAs. The equation is shown below:

      $$ \begin{split} \mathrm{l}\mathrm{n}{innoindex}_{it}=&{\beta }_{0}+\rho {\sum }_{j\ne i}^{\delta }\sum _{j\ne i}^{n}{W}_{ij}^{\delta \sim \delta +50}\mathrm{l}\mathrm{n}{innoindex}_{it}+\\ &{\beta }_{1}{\mathrm{l}\mathrm{n}FDI}_{it}+{\beta }_{2}{indstru}_{it}+{\beta }_{3}{govfiscal}_{it}+\\ &{\beta }_{4}{\mathrm{l}\mathrm{n}scihc}_{it}+{\beta }_{5}{\mathrm{l}\mathrm{n}internet}_{it}+{\beta }_{6}{\mathrm{l}\mathrm{n}road}_{it}+\\ &{\beta }_{7}{X}_{it}+{\alpha }_{i}+{\gamma }_{t}+{\varepsilon }_{it} \\[-11pt] \end{split}$$ (7)

      where Wijδ~δ+50 denotes the element of spatial weighted matrixes within a geographical scope of [δ, δ+50], taking 50 km as the unit distance and δ = 0, 50, 100, …, 600 km. When the element is outside the distance range, the weights will be 0. The explanation of remaining variables is the same as Eq. (6).

    • We graph the intra-and inter-UA Gini coefficients of technological innovation capacity from 2003 to 2016 in Fig. 2. As shown in the line chart of Fig. 2, the intra-UA innovation Gini coefficients have obvious downward slopping trend in YRD, from around 0.75 in 2003 to 0.64 in 2016, which indicates that the innovation gaps in YRD had decreased in the past decades and may have higher possibility of stronger innovation diffusion effect. While in PRD, the Gini coefficients of technological innovation capacity had increased from 2003 to 2010, peaked at about 0.75, then began to decline in 2011, and finally decreased to approximately 0.70 in 2016. By contrast, the Gini coefficients of technological innovation capacity in BTH, CYA, and MYR seem maintain rising. In BTH, the Gini coefficients had escalated from approximately 0.82 in 2003 to 0.85 in 2016, whose innovation gap was the largest among the five UAs in China. The Gini coefficients demonstrated a quite similar trajectory in CYA and MYR, which had climbed from 0.70 to around 0.79 in CYA and increased from approximately 0.67 to 0.75 in MYR.

      Figure 2.  Intra- and inter-UA technological innovation Gini coefficients during 2003–2016. The abbreviation means the same as in Fig. 1

      The bar chart of Fig. 2 shows the changes of inter-UA innovation gaps during 2003–2016. From this bar chart, the inter-UA Gini coefficients of technological innovation capacity have dropped from approximately 0.40 in 2003 to 0.38 in 2009 (right-hand scale), but rebounded on around 0.39 in 2011, and descended to 0.36 in 2016. To sum up, among the five UAs in China, innovation gaps were prone to shrink in YRD and PRD during our observation period. Meanwhile, the differences of technological innovation capacity between five UAs also have the tendency to decline.

    • First, in order to display the spatial evolution of five UA’s technological innovation capacity, we mapped the spatial distribution of cities’ technological innovation indexes in 2003 and 2016 in Fig. 3. From Fig. 3, we found that most UAs have at least one or more center cities with high technological innovation capacity, for instance, in BTH (Beijing, Tianjin), YRD (Shanghai, Nanjing, Hangzhou, Suzhou), PRD (Guangzhou, Shenzhen), CYA (Chengdu), as well as MYR (Wuhan, Changsha). These center cities were surrounded by cities with comparatively low technological innovation indexes, which is consider as ‘center-periphery’ structure in economic geography and regional innovation studies (Fritsch, 2002; Moreno et al., 2005). Meanwhile, from the spatial distribution in 2016, the technological innovation capacity development in YRD seems more even than other UAs, in line with the aforementioned declining tendency of technological innovation Gini coefficients in YRD. Superficially, the evaluation of spatial distribution of five UA’s technological innovation capacity exhibited eye-catching agglomeration characteristics of high values clustered by low values during the period of 2003–2016.

      Figure 3.  Spatial distribution of urban technological innovation capacity in China’s five UAs in 2003 and 2016. Scale 1 is set for maps of BTH, YRD, CYA, MYR. Scale 2 applies to PRD. The abbreviation means the same as in Fig. 1

      Second, we applied Moran’s I index to test the spatial autocorrelation of cities’ technological innovation capacity, based on three spatial weighted matrixes. The results are reported in Tables 24, we found that the indexes under geographical distance matrixes did not pass the significant tests, indicating that there is no conspicuous agglomeration of innovation capacity in five UAs upon geographical proximity, at least in the observation period of 2003–2016. On the contrary, almost all Moran’s I indexes are positively salient, except in PRD, illustrating that following economic distance, high innovation capacity cities clustered with the high ones, low values clustered with low values. In comparison, most Moran’s I indexes based on innovation collaboration matrixes presented negatively significant and the values of indexes are rather higher. The results implied us that among cities connected by co-patents, high innovation capacity area was circled around by low innovation capacity cities. For this, we further displayed the spatial distribution of GDP per capita and the number of co-patents in 2003 and 2016 in Fig. 4 and Fig. 5.

      Table 2.  Moran’s I indexes under geographical distance matrix of China’s five UAs from 2003 to 2016

      YearBTHYRDPRDCYAMYR
      2003 –0.112 (0.195) –0.034 (0.830) –0.176 (0.658) –0.083 (0.744) –0.063 (0.449)
      2004 –0.110 (0.219) –0.033 (0.812) –0.164 (0.730) –0.083 (0.724) –0.061 (0.424)
      2005 –0.109 (0.270) –0.035 (0.848) –0.159 (0.756) –0.082 (0.737) –0.062 (0.429)
      2006 –0.108 (0.316) –0.035 (0.845) –0.151 (0.806) –0.082 (0.728) –0.061 (0.441)
      2007 –0.108 (0.329) –0.034 (0.820) –0.138 (0.888) –0.082 (0.737) –0.063 (0.451)
      2008 –0.107 (0.349) –0.033 (0.813) –0.121 (0.954) –0.084 (0.729) –0.064 (0.484)
      2009 –0.107 (0.362) –0.032 (0.765) –0.111 (0.821) –0.086 (0.715) –0.064 (0.501)
      2010 –0.107 (0.327) –0.029 (0.728) –0.109 (0.784) –0.086 (0.714) –0.064 (0.510)
      2011 –0.108 (0.309) –0.025 (0.640) –0.109 (0.772) –0.087 (0.721) –0.064 (0.505)
      2012 –0.108 (0.297) –0.021 (0.559) –0.109 (0.780) –0.087 (0.723) –0.064 (0.501)
      2013 –0.108 (0.290) –0.017 (0.499) –0.109 (0.780) –0.087 (0.723) –0.064 (0.506)
      2014 –0.108 (0.285) –0.014 (0.456) –0.109 (0.781) –0.086 (0.727) –0.064 (0.503)
      2015 –0.108 (0.282) –0.008 (0.378) –0.108 (0.779) –0.085 (0.733) –0.063 (0.507)
      2016 –0.108 (0.278) –0.004 (0.328) –0.108 (0.782) –0.084 (0.738) –0.063 (0.506)
      Notes: P values are in parentheses. The abbreviation means the same as in Fig. 1

      Table 3.  Moran’s I indexes under economic distance matrix of China’s five UAs from 2003 to 2016

      YearBTHYRDPRDCYAMYR
      2003 0.062 (0.000) 0.257 (0.000) 0.026 (0.346) 0.043 (0.085) 0.144 (0.014)
      2004 0.069 (0.000) 0.242 (0.000) 0.031 (0.322) 0.041 (0.071) 0.124 (0.013)
      2005 0.102 (0.000) 0.232 (0.000) 0.031 (0.309) 0.042 (0.069) 0.128 (0.012)
      2006 0.137 (0.000) 0.215 (0.000) 0.028 (0.297) 0.040 (0.069) 0.132 (0.012)
      2007 0.146 (0.000) 0.211 (0.000) 0.023 (0.272) 0.041 (0.079) 0.149 (0.012)
      2008 0.164 (0.000) 0.221 (0.000) 0.010 (0.214) 0.043 (0.086) 0.170 (0.012)
      2009 0.166 (0.000) 0.227 (0.000) 0.000 (0.177) 0.044 (0.097) 0.183 (0.012)
      2010 0.138 (0.000) 0.246 (0.000) –0.002 (0.165) 0.045 (0.130) 0.184 (0.012)
      2011 0.125 (0.000) 0.263 (0.000) –0.003 (0.161) 0.046 (0.114) 0.183 (0.012)
      2012 0.114 (0.000) 0.276 (0.000) –0.002 (0.162) 0.047 (0.117) 0.184 (0.012)
      2013 0.104 (0.000) 0.285 (0.000) –0.001 (0.161) 0.047 (0.117) 0.184 (0.012)
      2014 0.100 (0.000) 0.294 (0.000) 0.000 (0.161) 0.046 (0.112) 0.183 (0.012)
      2015 0.095 (0.000) 0.306 (0.000) 0.000 (0.163) 0.045 (0.109) 0.180 (0.012)
      2016 0.090 (0.000) 0.307 (0.000) 0.001 (0.170) 0.045 (0.104) 0.178 (0.012)

      Table 4.  Moran’s I indexes under innovation collaboration matrix of China’s five UAs from 2003 to 2016

      YearBTHYRDPRDCYAMYR
      2003 –0.544 (0.005) –0.199 (0.012) –0.400 (0.096) –0.428 (0.004) –0.727 (0.000)
      2004 –0.538 (0.005) –0.201 (0.011) –0.424 (0.067) –0.439 (0.004) –0.719 (0.000)
      2005 –0.517 (0.008) –0.207 (0.008) –0.434 (0.055) –0.444 (0.003) –0.717 (0.000)
      2006 –0.493 (0.011) –0.211 (0.007) –0.439 (0.046) –0.445 (0.003) –0.719 (0.000)
      2007 –0.487 (0.012) –0.209 (0.008) –0.442 (0.036) –0.434 (0.004) –0.731 (0.000)
      2008 –0.475 (0.015) –0.204 (0.010) –0.438 (0.023) –0.422 (0.005) –0.713 (0.000)
      2009 –0.473 (0.015) –0.197 (0.014) –0.429 (0.020) –0.401 (0.007) –0.685 (0.000)
      2010 –0.492 (0.011) –0.185 (0.024) –0.420 (0.023) –0.391 (0.008) –0.674 (0.000)
      2011 –0.500 (0.010) –0.168 (0.047) –0.412 (0.026) –0.373 (0.011) –0.680 (0.000)
      2012 –0.508 (0.009) –0.156 (0.073) –0.403 (0.031) –0.366 (0.013) –0.679 (0.000)
      2013 –0.514 (0.008) –0.148 (0.095) –0.393 (0.038) –0.366 (0.013) –0.675 (0.000)
      2014 –0.517 (0.008) –0.142 (0.117) –0.387 (0.043) –0.378 (0.010) –0.682 (0.000)
      2015 –0.520 (0.007) –0.127 (0.185) –0.381 (0.049) –0.388 (0.008) –0.685 (0.000)
      2016 –0.523 (0.007) –0.115 (0.254) –0.375 (0.056) –0.399 (0.007) –0.691 (0.000)

      Figure 4.  Spatial distribution of GDP per capita in China’s five urban agglomerations in 2003 and 2016. Scale 1 is set for maps of BTH, YRD, CYA, MYR. Scale 2 applies to PRD

      Figure 5.  Evolution of inter-city co-patents network in China’s five urban agglomerations in 2003 and 2016. Scale 1 is set for maps of BTH, YRD, CYA, MYR. Scale 2 applies to PRD

      As shown in Fig. 4, the distributions of high values of GDP per capita are highly consistent with that of technological innovation indexes in five UAs and the feature of spatial concentration is also obvious. High GDP per capita cities in five UAs are BTH (Beijing, Tianjin), YRD (Shanghai, Hangzhou, Suzhou, Nanjing, Wuxi, Ningbo), PRD (Guangzhou, Shenzhen, Foshan, Zhuhai, Zhongshan), CYA (Chengdu), MYR (Wuhan, Changsha). Fig. 5 presents the inter-city patent collaboration in 2003 and 2006. It shows that the innovation collaboration networks have expanded a lot with time goes by. In 2003, the co-patents networks were suffused with structural holes since many cities were outside networks, especially in CYA and MYR. While in 2016, the innovation collaboration networks were denser, with YRD as the most intensive one, whose high co-patenting cities (the red line) count more. Moreover, the paradigm of agglomeration is also remarkable as high co-patenting lines are generated by center cities, such as Beijing, where nearly bridged the whole network in BTH, Chengdu in CYA, as well as Wuhan in MYR.

      In short, there are two takeaways from spatial autocorrelation tests: 1) the five UAs’ technological innovation capacity in 2003 to 2016 showed little geographical proximity but high dependence on economic distance and innovation collaboration, which is similar to the results of previous studies (Paci et al., 2014; Yang et al., 2021). 2) Moran’s I indexes illustrate that economic distance matrix and innovation collaboration matrix may better catch the innovation linkage between cities. Hence, we employed the economic distance matrix and innovation collaboration matrix to carry out following spatial autoregression model (SAR).

    • The results of two-way fixed Spatial Autoregressive (SAR) model for panel data of five UAs are reported in Table 5. In this section, we explained the results of our interested driving factors according to UAs one by one, and meanwhile, compared differences between main determinants for their technological innovation capacity.

      Table 5.  Main results of drivers of urban technological innovation capacity

      UABTHYRDPRDCYAMYR
      $ {W}_{ECON} $$ {W}_{INNO} $$ {W}_{ECON} $$ {W}_{INNO} $$ {W}_{INNO} $$ {W}_{ECON} $$ {W}_{INNO} $$ {W}_{ECON} $$ {W}_{INNO} $
      (1)(2)(3)(4)(5)(6)(7)(8)(9)
      lnGDP –0.483***
      (0.148)
      –0.457***
      (0.149)
      –0.594***
      (0.167)
      –0.511***
      (0.165)
      0.241
      (0.240)
      –1.101***
      (0.302)
      –1.074***
      (0.296)
      0.036
      (0.169)
      –0.023
      (0.164)
      lnpopulation 0.844
      (0.721)
      0.816
      (0.716)
      0.464
      (0.510)
      0.333
      (0.508)
      –0.230
      (0.184)
      0.593
      (1.223)
      0.449
      (1.196)
      –1.905***
      (0.717)
      –2.032***
      (0.691)
      lnland 0.289
      (0.521)
      0.162
      (0.504)
      0.076
      (0.376)
      0.172
      (0.375)
      –5.944*
      (3.466)
      0.324
      (1.443)
      0.477
      (1.413)
      –2.896**
      (1.406)
      –2.564*
      (1.359)
      lnFDI –0.002
      (0.038)
      0.001
      (0.038)
      –0.024
      (0.035)
      –0.047
      (0.034)
      –0.229*
      (0.117)
      –0.017
      (0.021)
      –0.019
      (0.021)
      –0.065*
      (0.039)
      –0.067*
      (0.037)
      indstru 0.000
      (0.011)
      0.000
      (0.011)
      0.035**
      (0.017)
      0.040**
      (0.017)
      –0.068**
      (0.032)
      –0.011
      (0.018)
      –0.012
      (0.017)
      –0.032*
      (0.018)
      –0.033*
      (0.018)
      govfiscal 15.584***
      (2.814)
      16.778***
      (2.600)
      12.309***
      (1.600)
      12.106***
      (1.595)
      5.379**
      (2.400)
      9.854**
      (4.569)
      9.229**
      (4.450)
      2.128
      (1.390)
      2.056
      (1.338)
      lnscihc –0.016
      (0.073)
      –0.011
      (0.072)
      0.111*
      (0.066)
      0.076
      (0.065)
      0.288***
      (0.091)
      0.028
      (0.070)
      0.034
      (0.068)
      0.067
      (0.049)
      0.071
      (0.047)
      lninternet –0.014
      (0.020)
      –0.014
      (0.020)
      0.133**
      (0.058)
      0.155***
      (0.059)
      –0.022
      (0.092)
      –0.103***
      (0.038)
      –0.098***
      (0.037)
      –0.107**
      (0.042)
      –0.117***
      (0.041)
      lnroad –0.076
      (0.102)
      –0.089
      (0.101)
      0.367***
      (0.061)
      0.385***
      (0.061)
      0.083
      (0.133)
      0.159**
      (0.073)
      0.166**
      (0.072)
      0.157**
      (0.061)
      0.167***
      (0.059)
      rho 0.128
      (0.115)
      –0.061
      (0.179)
      0.223***
      (0.066)
      –0.163
      (0.141)
      0.058
      (0.131)
      –0.115
      (0.154)
      –0.184
      (0.166)
      –0.058
      (0.083)
      –0.281***
      (0.103)
      Time fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes
      City fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes
      n 182 182 364 364 126 224 224 392 392
      R2 0.0849 0.0754 0.8876 0.8157 0.6546 0.5255 0.5846 0.5675 0.5722
      Notes: 1) The robust standard errors are in parentheses; * P < 0.1, ** P < 0.05, *** P < 0.01; 2) The regression results based on economic distance matrix are not included in PRD since it failed to pass the spatial autocorrelation significance tests under economic distance matrix

      First, from column (1) and (2) of Table 5, effects of fiscal investment for science and technology (S&T) are positively significant in BTH, whose magnitude under innovation collaboration matrix is 1.2% higher than the equivalent under economic distance matrix. While other influencing factors are not salient in BTH, whether under the economic distance matrix nor the innovation collaboration matrix.

      Second, as shown in column (3) and (4) for the results of YRD, coefficients of industrial structure, fiscal investment for S&T, as well as two types of infrastructure facilities are consistently positively significant, no matter which model specification, with the largest elasticity of fiscal investment for S&T (around 12%), followed by internet infrastructure (0.13%–0.15%), road infrastructure (0.37%–0.38%) and industrial structure (approximately 0.04%). The effect of human capital for S&T is positively significant in YRD when use economic distance, and the effect size is about 0.11%. Notably, positive impacts of infrastructure facility are most prominent in YRD. The effect of road infrastructure in YRD is almost 2.3 times of the equivalent of other UAs. Coefficients of internet infrastructure are just positive significant in YRD.

      Third, we used innovation collaboration matrix to handle spatial autoregression in PRD since its insignificant Moran’s I indexes of other matrixes (see section 4.1.2). From column (5) of Table 5, the significant effect of fiscal investment for S&T is nearly 2 times of influence of human capital for S&T. While the industrial structure and FDI are negatively correlated with technological innovation capacity. 1% increase of FDI will cause around 0.23% decrease of local technological innovation capacity. The coefficient of industrial structure is around –0.07. Forth, in CYA, coefficients of fiscal investment for S&T and road infrastructure are positively salient, while the sign of internet infrastructure is negative. Magnitude of effect of fiscal investment for S&T still outweigh the equivalent of other variables. Fifth, in sharp contrast to YRD, the effect of fiscal investment for S&T in MYR is inconspicuous, even though the sign is positive. Moreover, industrial structure, FDI as well as internet infrastructure are not conducive to technological innovation capacity in MYR. Among all significant variables, just the influence of road infrastructure is positive and its magnitude is around 0.16, similar to CYA.

      Last, from the results of coefficients of spatial autocorrelation (rho) reported in Table 5, we found that salient innovation spillover just emerged in two specifications, one is the spatial autoregression for YRD based on economic distance matrix, the other is the estimation in MYR bridged by inter-city innovation collaboration. The spatial autocorrelation coefficient of technological innovation capacity is positively significant with the effect size of approximately 0.22, indicating that 1% increase of technological innovation capacity of center cities can generate 0.22% extra technological innovation capacity increase of adjacent areas. While the magnitude of negative innovation spillover in MYR is around –0.28, which means that innovation growth in MYR is at the expenses of resources crowding out and may imply us that innovation polarization effect dominated innovation diffusion effect in MYR.

      We summarized these results into two findings: 1) the regression results under economic distance matrix and innovation collaboration matrix are basically the same. It proves that our results are quite robust. Furthermore, we reported that decomposition results of SAR model in Appendix Table A7including direct effects, indirect effects and total effects of all driving factors. The decomposition results are also in accordance with main results calculated by SAR. 2) Common patterns show in substantial positive impacts of fiscal investment for S&T and negative influence of FDI in almost all five UAs. While disparities appear concerning effects of industrial structure, human capital for S&T as well as infrastructure facility. The differences of driving factors between UAs may provide empirical evidence for better understanding why innovation diffusion effect or innovation polarization effect predominated each other in certain UAs.

    • As discussed above, innovation diffusion effects dominated in YRD since its positive innovation spillover under economic distance matrix verified in SAR (Table 5), and innovation polarization connected by co-patents emerged in MYR. In order to investigate detailed spatial patterns of innovation spillover in these two UAs, we reset economic distance matrix and innovation collaboration matrix benchmarking inter-city geographical distance with 50 km as unit distance, and then we estimated SAR again (Eq.7). The left-hand graph in Fig. 6 shows spatial patterns of innovation spillover measured by economic distance in YRD and the right-hand one displays spatial innovation spillovers in different distance ranges linked by co-patents in MYR.

      Figure 6.  Spatial patterns of innovation spillovers in YRD and MYR. The results of YRD stopped within radius of 400 km and 500 km for MYR, because SAR model failed to convergence after the two ranges since cities outside the scope are rare and spillovers become negligible. 95% CI in these two figures mean 95% confidence intervals

      From the left graph in Fig. 6, spatial pattern of innovation spillover in YRD is inverted ‘N’ shape as a whole. In the scope of 0–100 km, the magnitude of positive innovation spillover in YRD is around 0.30, which means 1% increase of technological innovation capacity of center cities can leverage 0.30% innovation growth of their peripheral areas within the UA. After 100 km, the innovation spillover begins to decay rapidly with distance and in the range of 100–150 km, the innovation spillover changed from positive to negative with an elasticity of –0.20 in YRD. Nonetheless, the innovation diffusion effect rebound ranging from 150–250 km, and then attenuated with distance to negative innovation spillover. The effects after 300 km are not significant any more.

      As presented in right chart of Fig. 6, innovation spillover in MYR bears a very similar spatial paradigm to that in YRD, even though the positive innovation spillover demonstrated a recovering trend twice in MYR, which makes spatial patterns of innovation spillover in MYR looks like a ‘W’ shape. Similarly, innovation diffusion mainly shows in the scope of 0–100 km, and plunged to negative spillover after 100 km. Then, innovation diffusion effect rises from 150–250 km. However, distinguished from YRD, the positive innovation spillover has an upward trend between 300 km and 350 km, and spatial effects are not salient after 350 km. Intuitively, disparities between innovation spatial patterns of YRD and MYR may be attributed to different city sizes and cities’ agglomeration in different geographical distance of two UAs. This speculation is consolidated by our real data. We found that most cities are connected at the scope of 200–250 km in YRD (about 53 links in our matrix), while that distance in MYR is 300–350 km with approximately 63 links. The range for cities agglomerated happened to be places where positive innovation spillover rebound.

      In general, we found that 1) innovation diffusion effect represented by positive innovation spillover and innoavation poliarization effect proxied by negative innovation spillover emerge in cities located at different gepgraphical distance when shaping spatial patterns of innovation spillovers, whether in YRD or in MYR. 2) The non-linear spatial pattern of innovation spillovers in YRD is inverted ‘N’ shape, and demonstrated a ‘W’ shape in MYR, following the rule of attenuation with geographical distance. These findings are similar to conclusions of previous studies (Sheng et al., 2019; Wang et al., 2022). 3) The innovation spillovers are restrained by radius of 300 km in YRD and 350 km in MYR. The scope here are close to previous estimates about a circle of 250–400 km of knowledge spillovers (Bottazzi and Peri, 2003; Greunz, 2005; Wang et al., 2022).

    • Just as mentioned in section 4.2, we concluded similarities and disparities upon the drivers of technological innovation capacity of five UAs. In this part, we try to obtain some insights for policy-making by these common patterns, and also to explain reasons and inner mechanisms behind the divergences.

      For the similar results of drivers, we found that elasticity of fiscal investment for S&T to innovation growth is considerable in almost all five UAs, albeit with differences in returns between UAs. This finding echoed previous studies’ statements on government’s crucial role in funding R&D activities (Shang et al, 2012). They argued that regional innovation development can be achieved via direct knowledge and technological spillovers transmitted by public institutions that funded by governments (e.g., research institutes, universities, etc.), and indirect interplay of firm’s collaboration with these public institutions (Koch and Simmler, 2020; Min et al., 2020). While concerning the negative influence of FDI inflows in almost all UAs, it may reflect a phenomenon documented by literature: increased competition that comes with foreign entry may relegate local firms to less innovative market niches or crowd local firms out of the market (García et al., 2013; Ning et al., 2016), which reminds us that rather than blindly introduce FDI, policies for regional innovation should lay more emphasis on FDI-affiliated technology transfer and knowledge spillover (Lin and Kwan, 2016).

      As regard to disparities of drivers among UAs, firstly, positive consequences of industrial structure and internet infrastructure just shown in YRD, which may originate from well-developed industrial integration and industrial collaboration, and stampede of internet companies (e.g., Alibaba Group, NetEase Group, etc.) in YRD. One the one hand, featuring in abundant innovation resources and actors of innovation collaboration network, such as universities, research institutes, as well as firms, center cities in YRD like Shanghai, Nanjing, Hangzhou, Suzhou, pose tremendous demonstration effect on adjacent cities when improving their own innovation capabilities. On the other hand, similar industrial structure with local distinctiveness and resources complementarity increase the possibility of industrial transfer (Liu et al., 2020). For instance, the research of Wang et al. (2021) found that the soaring of innovation scores in YRD cities (e.g., Chuzhou, Hefei, Wuhu, Anqing) actually came from industrial transfer to a large extent.

      Second, human capital for S&T promoted regional innovation capacity just in YRD and PRD. Most signs of this variable are not salient, even though most signs are positive in SAR. Here we think it may because talent races and talent competition in other UAs. Center cities in these UAs may attract high-skilled labor from neighboring cities, thus resulting in human capital shortfalls of peripheral cities (Zhou et al., 2018). In comparison, comparatively low wage gaps within YRD and PRD make them suffered less by unbalanced talents flow. Our inference upon this issue can be verified by the decomposition results of SAR, as the indirect effects of human capital are positive in YRD and PRD, while negative in other UAs. This reminder tells us that more efforts should be made by governments to build reasonable salary distribution system and institutions for inter-city talents transfer, thus compensating brain drain caused by regional salary differences.

      Third, the function of road infrastructure is extremely distinctive in YRD, CYA and MYR, confirmed the pivotal role of public transport, particularly the highway, in regional development (Faber, 2014), especially for remote areas with urgent needs for transportation system. Nevertheless, we also note that we should be cautious about regional inequality of bypassed region of new infrastructure construction projects we launch (Qin, 2017).

    • The spatial autoregression coefficients in SAR verified that innovation diffusion effect dominated in YRD (coefficient rho in Table 2), but innovation polarization effect prodomintes the other in MYR in our sample period, although we found positive innvation spillover in some geographical divisions of MYR is even higher than that of YRD (see section 4.3). It may imply us that innovation diffusion effect of center cities in MYR failed to fully functinate. One explanation for these contradictory results is ‘beggar thy neighbor effect’ of resources competition between center cities and their peripheral cities in MYR. The innovation production factors, such as industries, talents as well as investment, may not be well coupled in this area. This conjecture can be partially cross-validate by the results of SAR model as coefficients of most influencing factors in MYR are negative.

      Meanwhile, if we combine the results of autocorrelation coefficients with estimates of driving factors, deeper reasons for oppsite innovation spillovers in these two UAs seem clearer. Differences in industrial structure and internet infrastructure may explain why innovation diffusion effect predominates innovation polarization effect in YRD and vice versa in MYR, because coefficients of these two variables are reversed in these two UAs. This conjecture keeps consistent with our analysis in section 5.1 about YRD’s advantage in industry transfer and recent rise of internet companies, and also in line with studies about dependence of balanced regional innovation development on industrial specialization and industrial integration ( Ning et al., 2016; Wang et al., 2016a).

      As for why the curves of innovation spillover are inverted-N shape or W shape, in other words, the recovering trend of positive innovation spillover at radius of 150–250 km, we discussed it according to gradient theory of domestic industry transfer in economic geography. The domestic industrial transfer is adapted from product cycle theory (Vernon, 1966) and ‘flying geese’ model (Akamatsu, 1962), which explained industry transfer from the perspective of comparative advantage. They think developed regions would transfer industries that they have no comparative advantages to developinng areas, where have low-cost labor and resources (Ang, 2018). It means industry prones to transfer between cities with cetain economic distance. If industry transfers to regions where have same economic level, costs are too high, while if transfer to cities with large economic gaps, the contractors may lack the ability to undertake. Moreover, gradient industry transfer occurs more frequently in labor-intensive and export-oriented manufacturing. Combining the real situation of industrial specification and industrial linkage in YRD, it is no wonder to find an upward trend of innovation diffusion at the middle distance, for example, leather in Haining, hardware in Yongkang, and light textile in Shaoxing. These labor-intensive industries in YRD increases likelihood for these cities to be foundries and undertake industrial transfers. The same rationale can also apply to MYR. While the rising tendency at the scope of 300–350 km in MYR is more likely related to spatial agglomeration patterns of cities in MYR. As we discussed in section 4.3, a majority of cities are connected by innovation collaboration in the radius of 300–350 km.

    • Based on balanced panel data from 2003 to 2016 of five urban agglomerations’ 92 prefecture-level cities in China, this paper shows spatio-temporal evolution process of their technological innovation capacity with discussing polarization-diffusion patterns, and simultaneously examines driving factors of that evolution processes. We mainly report following three findings: 1) first, there is a high degree of concentration in technological innovation capacity distribution within all China’s five UAs, linked by economic and innovation collaborations. 2) Second, innovation capacity increase in China’s five UAs is driven by government’s investment in S&T to a large extent, followed by influences of infrastructure facilities construction, human capital for S&T, as well as transformation of industrial structure, with great disparities emerged between UAs in our sample period. 3) Third, while the intra-region innovation gaps are reducing in Yangtze River Delta Urban Agglomeration (YRD) with obvious innovation diffusion, Urban Agglomeration in Middle Reaches of the Yangtze River (MYR) is still dominated by innovation polarized growth. Differences in polarization-diffusion patterns between these two UAs may be explained by opposite returns of industrial structure and internet infrastructure in two UAs.

      This paper contributes to better understanding why some UAs realized balanced regional innovation development, while innovative activities tend to polarize in other UAs. These findings can be insightful for more effective policy-making in promoting innovation development while reducing regional inequality through innovation diffusion. We emphasize policies for regional innovation development should focus more on public investment in innovative activities and public infrastructure, especially in regions where have less well-developed road infrastructure. Meanwhile, institutions for encouraging industrial cooperation and industrial integration within UAs, as well as inter-city talents transfer are also extremely important in generating positive innovation spillovers. Besides, we should be cautious about pros and cons of FDI inflows, and pay more attention on FDI-affiliated technology transfer and knowledge spillover. Anyway, there is no one-size-fits-all policy and innovation policies can be more regional-specific and oriented to local milieu, given great disparities in driving factors for innovation development between UAs.

      There are several limitations in our study. One limitation is the measurement of innovation linkage. In this paper, we used geographical distance, economic distance and inter-city co-patents to depict the innovation connection between cities, while given the fact that not only the number of innovation interactions but also position in innovation networks may also influence cities’ technological innovation capacity. In the future, diversified network topology measurements can be applied to solve this issue. The other is about data limitation. More granular data for driving factors, e.g., industrial transfer and transformation, may help us better reveal inner mechanisms for how innovation diffusion effect and innovation polarization effect happen within UAs. We await these limitations can be addressed in future research.

    • Tables A1−A7 could be found in the corresponding article at http://egeoscien.neigae.ac.cn/article/2022/6.

参考文献 (71)
补充材料:
caojinxian-supplementary.pdf

目录

    /

    返回文章
    返回