top of page

ACADEMIC RESEARCH

See 'Policy Research' page for further links and details.

fdi_edited_edited.jpg

Skills / Tools

'R', GGPLOT, BAYESIAN REGRESSION ANALYSIS

CORPORATE SECULAR STAGNATION (WITH JANGHO YANG)

We detail a secular slowdown in investment rates using a large panel of advanced economy firms from 18 countries between 1994-2017. We test competing explanations for this using a Bayesian ‘mixed effects’ model, with time-varying and country-varying coefficients. The slope of the investment demand curve remains roughly constant, indicating that growing monopoly power has not dulled firms' responsiveness to investment opportunities. Contrary to precautionary savings arguments, advanced economy firms are not meaningfully financially constrained. The corporate sector as a whole is increasingly a net external `releaser' of funds to shareholders, creditors, and bondholders, and this closely tracks declining investment rates.

  • Preliminary findings published with Oxford University's Oxford Martin School Working Paper series. 

inv_edited.jpg

FOREIGN DIRECT INVESTMENT

Writings on foreign direct investment published with Columbia University, UNCTAD, and others.

  • Book chapters for Columbia University's International Investment Yearbook 2019 on emerging China-USA rivalry (here) & 2015 (Oxford University Press) using a Bayesian VAR (here).

  • Columbia University, Columbia FDI Perspectives (peer reviewed), here and here.

  • Published with UNCTAD's Transnational Corporation Journal, here.

  • Using basic network theory, in application to regional integration, here - published with African Development Review.

fincf_edited.jpg

I use R language for data analysis, data visualisation (ggplot), and panel regressions (Bayesian). I work tends to use Bayesian approaches to statistical research problems. I have expertise in Bayesian multilevel / hierarchical modelling which is essentially a data-informed compromise between fixed effects and random effects, which allows for one to model heterogeneity directly rather than hold it constant through cluster-robust standard errors.

bottom of page