Forecast accuracy vs. forecast value when modeling the Dollar-Pound Sterling since the 19th century

Findings

This study estimate purchasing power parity (PPP) models of the US Dollar-Pound Sterling based on data from 1861-2014, after which the statistical accuracy and economic value of these forecasts are compared. This study finds that 1) the estimated model beats a random walk model in-sample and 2) that both the accuracy and value of out-of-sample forecasts fall as the forecast horizon increases.

Model developed

The developed models predict estimated future changes in the USD-GBP 1, 4, 8, 12 and 16 years ahead based on gaps between the spot and PPP-implied exchange rate. Table 1 and diagram 1 below show that all five models are statistically significant and have intuitive coefficients: an increase in the fundamental exchange rate (ft) over the spot exchange rate (st) leads to a future increase in the spot exchange rate across all forecasting horizons.

Table 1: model findings

Diagram 1: coefficients and forecast errors across different forecast horizons

To ensure congruency all five models were estimated using heteroskedasticity and autocorrelation-consistent standard errors. Except for the model forecasting four years ahead, then the parameter stability of other forecast horizons was deemed sufficiently stable to use for forecasting.

Forecast accuracy and value

The statistical accuracy and economic value of the estimated models were analyzed across the different forecasting horizons. The economic value of forecasts was examined using the directional accuracy and time value-adjusted root mean squared forecast error (RMSFE), and a stylized finding is that both the statistical accuracy and economic value of the estimated models decrease as the forecasting horizon increases. 

Read the full paper here: http://bit.ly/2qYoxdl

Capital mobility and the Feldstein-Horioka Puzzle

I recently completed an empirical research paper on the Feldstein-Horioka Puzzle (“FHP”) for my class Quantitative Global Economics and wanted to share my findings. One of my main findings can be seen in the diagram above. A higher savings coefficient indicates low capital mobility (0 is perfect capital mobility and 1 is full capital immobility).

The FHP is the empirical finding by Feldstein and Horioka (1980) that international capital mobility, when measured as the cross-country correlation between savings and investment rates, is rather low, despite economic theory arguing that capital is highly mobile.

My paper carries out two analyses: the original FHP is examined using the same methodology as the original paper but using different sample periods and country groups. Secondly, I examined whether the quality of economic institutions is correlated with capital mobility when measured as the difference between national savings and investment rates, something that previous research has not analyzed.

Findings

The first analysis of this paper used the original OECD country group that Feldstein and Horioka used and found that international capital mobility has increased from the 1970s until 2009, after which capital mobility has decreased. Using the same method but 193 countries it was found that capital mobility increased until 1999, after which fell.

The second part of this paper found that the only significant determinant of capital mobility during 1995-2014 was the openness of trade regimes, and that capital mobility is not significantly correlated with time after controlling for economic institutions. It should be noted, however, that the findings were not robust when analyzed separately for the first and last ten years of the sample period.

You can read the full paper here: https://goo.gl/HsVDcm 

Determinants of Chinese local government bond spreads

How come that Chinese local government bond yields are different from Chinese central government bond yields - despite that the central government guarantees local government debt?
My hypothesis is that provincial economic conditions should not matter for local government bond spreads given that the central government guarantees local debt. To test my hypothesis I analyzed whether provincial economic conditions can explain the central-local spreads of 2,680 Chinese local government bonds issued during 2015-2017 - bonds equal to 21% of Chinese Q3 2017 GDP.

My findings show that some indicators of provincial economic conditions explain local government bond spreads, although this finding is against my hypothesis. You can read the whole paper here: https://goo.gl/5SY5d1.

The below photo is from University of Pennsylvania's Symposium on Contemporary China, where I had the opportunity to present, and get feedback on, my research.