Session 8 c diewert discussion of feenstra inklaar and timmer

of 18
35 views
PDF
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Document Description
1. Discussion by Erwin Diewert of: “Penn World Tables 8.0: A User Guide” Robert C. Feenstra (University of California and NBER) Robert Inklaar (University of…
Document Share
Document Transcript
  • 1. Discussion by Erwin Diewert of: “Penn World Tables 8.0: A User Guide” Robert C. Feenstra (University of California and NBER) Robert Inklaar (University of Groningen) Marcel Timmer (University of Groningen) 33rd IARIW General Conference, August 24-30, 2014, Rotterdam, The Netherlands 1
  • 2. Overview  The main purpose of the paper is to provide readers of a non-technical overview of the latest version of the Penn World Tables: PWT: Version 8.  What are the new concepts?  How was the data set constructed (roughly)?  Which version of the Tables should be used in various specific research contexts?  What are the limitations of the measures?  This is a very exciting paper! I think that the authors do an excellent job of answering the above questions!  75000 visitors to the website since its launch in July 2013!  75 citations to the main paper since the launch! Definitely a best selling paper! 2
  • 3. Three Major Changes to the PWT • For their output oriented measures of real GDP RGDPo that are comparable across countries, they introduce PPPs for exports and imports. This measure of real GDP is useful for comparing productivity levels across countries. • PWT:8 now interpolates real GDP levels between benchmark PPP data instead of using the most recent PPP data to determine relative GDP levels across countries and then using national growth rates of real GDP to extrapolate backwards into the indefinite past. • PWT:8 also provides some measures of country capital stocks (in comparable units across countries) and labour input by country so that the Penn World Tables can provide measures of country productivity levels across countries. • All of these innovations are very useful. 3
  • 4. The China-US Comparison 2005 China United States China/US A: GDP per capita in national currency 14565 42330 in US dollars, converted with exchange rate 1777 42330 4% in US dollars, GDPe 5342 43209 12% in US dollars, GDPo 5270 42330 12% B: productivity GDPo per worker (US$) 8967 87691 10% Tangible capital stock per worker (US$) 29245 262546 11% Human capital per worker (index) 2.33 3.64 64% Total factor productivity (index) 0.29 1.00 29% 4
  • 5. Discussion of Expenditure Side Real GDP, RGDPe (1)  RGDPe is basically a variant of the World Bank’s ICP estimates of Real GDP by country, that compares real C+I+G across countries in comparable units and adds the countries trade balance deflated by its exchange rate with a numeraire country to form expenditure side real GDP.  The resulting real GDP estimates are divided by the country’s population and the resulting real per capita GDP levels are (approximately) comparable across countries and give some indication of relative living standards across countries.  Historically, the Penn World Tables have used the latest available World Bank ICP GDP relative levels (the last two being 2005 and 2011) and then national real GDP growth rates are used to extrapolate the latest (comparable across space) ICP real GDP levels back through time. 5
  • 6. Discussion of Expenditure Side Real GDP, RGDPe (2)  PWT:8 makes use of all available World Bank PPP exercises (1970, 1975, 1980, 1985, 1996, 2005) as well as OECD annual benchmark estimates of member country relative real GDP levels and interpolates between successive benchmark estimates instead of just using one benchmark and extrapolating backwards forever using national real GDP growth rates.  I believe that this change in methodology is a huge improvement over the earlier Penn Tables. The problem was that each new benchmark radically changed relative GDP levels backwards for 40 years or so. The new methodology will change relative GDP levels but only between successive benchmark years. [Unless there are historical GDP revisions, in which case there will be further changes].  Basic inputs into the construction of the tables: PPPs for benchmark years, nominal and real national GDP series. 6
  • 7. Discussion of Expenditure Side Real GDP, RGDPe (3)  Feenstra, Inklaar and Timmer (FIT) made a few adjustments to PWT:7 and World Bank ICP methodology:  (1) Instead of using the PWT:7 method or the World Bank method for aggregating C+G+I (basically the GEKS method for making international comparisons with some bells and whistles to account for the problems of maintaining within region parities), PWT:8 used the GEKS method to aggregate C+G and then used the Geary-Khamis method to aggregate C+G with I. Typically this will not make much difference.  (2) FIT made consistent productivity adjustments for the Health, Education and General Government sectors (the World Bank made some adjustments for some regions but not for others—had to respect regional preferences).  (3) Following Deaton and Heston (2013), Chinese consumption prices for 2005 were adjusted down by 20%.  The following chart compares WB and FIT parities for 2005. 7
  • 8. Discussion of Expenditure Side Real GDP, RGDPe (4) 8
  • 9. The Index Number Formula and PPP Reliability (1)  FIT note that (following Deaton and Heston 2010) it is hard to compare wildly different countries.  Why? Because they can be consuming totally different goods and services. How can we compare the incomparable?  Thus price comparisons over time (within a country) are likely to be much more accurate than price comparisons over space (PPP comparisons). Thus the relative real GDP estimates generated by PPPs are not as reliable as the time series estimates.  Even with perfectly accurate PPPs at the basic heading level, there will be problems in comparing real GDP over space due to big differences in relative prices across countries; superlative indexes will no longer closely approximate each other. And non-superlative multilateral indexes (like GK) will be even more different than multilateral indexes base on bilateral superlative indexes (like GEKS). 9
  • 10. The Index Number Formula and PPP Reliability (2)  Thus FIT note that the differences between the GK and GEKS PPP for a single country can easily be as big as 50% and thus they note that “due care is needed when comparing living standards between rich and poor countries”.  A problem with the GEKS multilateral method is that it may be too democratic; i.e., every country plays a role as a numeraire country, where its price (or quantity) structure is compared in a bilateral fashion with the prices (or quantities) of every other country in the comparison. Once all of these “star” parities have been constructed, GEKS simply takes the equally weighted geometric mean of these star parities as the final parities.  I think that a better way of linking countries is by spatial linking or similarity linking; see the many papers by Robert Hill but in particular his chapter 9 in D.S. Prasada Rao (2009), Purchasing Power Parities of Currencies (and look at my Chapter 8 in the book as well!). 10
  • 11. The Problem with an Additive Method (like GK) 11 A B C q2 q1
  • 12. The Index Number Formula and PPP Reliability (4)  FIT also note that it is hard to justify the assumption of homothetic (and identical) preferences across countries, which is the usual justification for the use of the Fisher index. • But Törnqvist-Theil bilateral price and quantity indexes can be justified for non-homothetic (but identical) preferences (this was already done in my 1976 Journal of Econometrics article on superlative indexes) and for non-homothetic and even non-identical preferences; see Caves, Christensen and Diewert (1982), Econometrica 50 (1982), pages 1409-1411 and Diewert (2009), “Cost of Living Indexes and Exact Index Numbers”, Chapter 8 in Quantifying Consumer Preferences, edited by Daniel Slottje in the Contributions to Economic Analysis Series, United Kingdom: Emerald Group Publishing, pp. 207-246. • We turn now to a discussion of RGDPo 12
  • 13. Discussion of Output Side Real GDP, RGDP0 (1) • The second major change that FIT implemented was to introduce output oriented measures of Real GDP that could be compared across time and space, RGDPo . • Essentially, they constructed PPPs for exports and imports (using detailed unit value statistics for X and M) which is appropriate if we want to apply a production theory framework to model the relative outputs of each economy in the comparison project. • Thus this is all good as far as I am concerned! • These production theory real output measures can be matched up with the corresponding real input measures for at least a number of countries (EUKLEMS and World KLEMS provide suitable estimates for many countries) and then we can compare productivity levels of countries compared to a world frontier and study convergence etc. ,13
  • 14. Discussion of Output Side Real GDP, RGDP0 (2)  On the previous slide, I have already mentioned the third major innovation that FIT have introduced into the Penn World Tables; namely information on capital stocks and labour input for selected countries.  But the capital stock estimates are incomplete. These estimates are of stocks and the list of stocks is incomplete.  In particular, there are no estimates for the stocks of land, inventories, subsoil natural resources and R&D at present.  Presumably as more countries develop their balance sheet estimates, these gaps will be filled in.  The labour estimates are also a bit crude; no adjustment for hours worked. But it is a start!  The following slide shows the output differences for 2005. 14
  • 15. Final Expenditure Side vs Output Side Real GDP 15
  • 16. Possible Directions for Change (1) (1) For RGDPe , it seems to me that we should move to net product from gross product. Nobody can consume depreciation but net additions to the capital stock are definitely contributions to welfare. If PWT were to move to the net framework, it would set a wonderful precedent! Moreover, net investment is a lot smaller than gross investment and so it would not matter so much what index number formula is used to aggregate C+G with I. (2) For both GDP concepts, I am assuming that the interpolation method is the same. The authors suggest a form of linear interpolation between benchmarks. I have suggested an alternative interpolation scheme to the authors which I think is “better”; see Diewert (2014), “Alternative Measures of OECD Output Growth and Inflation” UBC discussion paper 14-01. Maybe they could consider it. 16
  • 17. Possible Directions for Change (2) For RGDPo , I would like to suggest the following: • Indirect taxes should be removed from consumption prices in order to better justify the economic approach to the output quantity index and to productivity analysis. (This follows the productivity literature started by Jorgenson and Griliches, 1968, 1972). • The authors should consider using the approach outlined by Caves, Christensen and Diewert EJ (1982) and Diewert and Morrison EJ (1986), which justifies the use of Törnvist-Theil indexes in a production theory context. • Try to fill in some of the gaps in the capital stock estimates. It should be possible to extend the KLEMS data to include inventory stocks and for the few countries that have annual balance sheets, land estimates should be available. Land is a tremendously important missing input in the KLEMS framework. 17
  • 18. Possible Directions for Change (3) • As the World KLEMs database improves, the primary input data base for PWT can also improve. I think that the PWT should not put out poor quality data on primary inputs. If they do, people will use it and come up with all sorts of problematic empirical results. So restrict the primary input base to countries where you are more comfortable about the data quality. • The application of traditional production theory to the Health, Education and General Government sectors is problematic. Also, traditional production theory does not apply to the financial sector (and we do not have a consensus on how to measure outputs and inputs in this sector). Perhaps PWT should leave out these problematic sectors until measurement improves in these hard to measure sectors. (Advertising, R&D and Business Consulting also tough). Conclusion: a great paper and a great leap forward! 18
  • Search Related
    We Need Your Support
    Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

    Thanks to everyone for your continued support.

    No, Thanks
    SAVE OUR EARTH

    We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

    More details...

    Sign Now!

    We are very appreciated for your Prompt Action!

    x