Trouble with Critical Bounds in ARDL Model

For econometric discussions not necessarily related to EViews.

Moderators: EViews Gareth, EViews Moderator

Posts: 1
Joined: Wed May 02, 2018 9:45 pm

Trouble with Critical Bounds in ARDL Model

Postby SANhedrin » Thu May 03, 2018 9:38 pm

Dear Colleagues

I am currently running an ARDL model where I look at the sources of economic growth. My research is estimating a linearised version of the Cobb-Douglas production function with an "unpacked" TFP variable. The variables I have included in my model as part of the TFP variable are:

Real Government Revenue,
Real Government Expenditure,
Real Foreign Direct Investment,
CPI Inflation,
Openness (import + export) ratio of GDP and
Real Official Development Assistance (ODA).

In addition to the above variables, I have three other variables in my model, namely:

Real Capital Stock,
Real Gross Domestic Production (GDP) and
Total Employment

When I run my ARDL model, my K=8. After I estimate the model and run the long-run results. The attached table appears.

What could be the reason that the critical bounds from Narayan are all -1?

Note: When I drop one of the variables, the Narayan critical bounds improve to those that are more familiar.

Is there a maximum number of variables that can be included in the ARDL?
F-test.JPG (29.81 KiB) Viewed 672 times

EViews Mirza
Posts: 46
Joined: Sat Apr 22, 2017 8:23 pm

Re: Trouble with Critical Bounds in ARDL Model

Postby EViews Mirza » Fri May 04, 2018 9:24 am

Hi SANhedrin,

I just looked a bit more carefully into your request. The reason that -1 appears is because the Narayan (2005) Applied Economics paper, "The saving and investment nexus for China: evidence from cointegration tests", from which the Narayan tables are obtained, does not support k>7. Hence, the output. I suppose what we can do is make that -1 into something more user readable, but apart from that, not much we can do until someone publishes more accommodating tables. For the sake of research, stick to the asymptotic critical values. They are not wrong, and you probably won't make outrageously unreliable inference using them, despite your low sample size.

I hope this helps.

Return to “Econometric Discussions”

Who is online

Users browsing this forum: No registered users and 5 guests