You should look at two possibilities:

(1) Problems of endogeneity in the econometric model (two-causality, unobserved heterogeneity...) or problems of another nature in it.

(2) Problems of the theory. Maybe there can be explanations for your result.

Trade openness could either have a positive or negative impact on the economy. The stage (productive or consuming) of the economy matters. Besides, the proxy for trade openness equally matters.

Why would urbanization have a negative effect on economic growth? I would think that economic growth is associated to urbanization. As has been said, there might be a problem of endogeneity and theory.

Everything said about export should be greater than import to create a trade surplus contributory to foreign exchange reserves, a higher import connotes robust economic activity which reflects positive economic growth reflecting lower unemployment, maybe a trade-off to approximately higher inflation.

Hi Welde,

Hope you are doing well.

If you are expecting a positive relationship and you end up getting a negative relationship, the outcome may be context driven. That is, though theory may postulate a particular relationship, the situation pertaining to an economy may not support that and hence the sign may change. In this case, you have to read wide and justify the alternative sign you have obtained especially if it is significant.

Good luck and stay safe.

I am with Cecilia: Why expect a negative sign for urbanization? But I think, that also the effect of trade openness is not clear. Mostly, trade openness is measured by the percentage of trade in GDP, i.e. (X+M)/GDP. By definition GDP=DD+X-M with DD as domestic demand. Therefore, one may likely have different signs of the relation, depending on growth of the GDP components and if a country has a trade surplus or not. Unfortunately, the question does not mention how urbanization and trade openness are measured in this regression.

It is expected for null hypothesis; that urbanization should have positive significance impact on economic growth except, if urbanization is regressed in relation to security issue that we may expect to have negative impact on the result. Trade Openness could have either positive or negative impact on economic growth, it depends if the country concern is a producing or importing country which will have great impact on value and strength of the currency of that country. Time and season could also alter the decision of the target respondents during data collection.

It is expected for null hypothesis; that urbanization should have positive significance impact on economic growth except, if urbanization is regressed in relation to security issue that we may expect to have negative impact on the result. Trade Openness could have either positive or negative impact on economic growth, it depends if the country concern is a producing or importing country which will have great impact on value and strength of the currency of that country. Time and season could also alter the decision of the target respondents during data collection.

It is expected for null hypothesis; that urbanization should have positive significance impact on economic growth except, if urbanization is regressed in relation to security issue that we may expect to have negative impact on the result. Trade Openness could have either positive or negative impact on economic growth, it depends if the country concern is a producing or importing country which will have great impact on value and strength of the currency of that country. Time and season could also alter the decision of the target respondents during data collection.

First, be sure of a reliable data source. You cannot judge from mere description, the softwares runs what you input into it.

Also be careful of problems of endogenity.

See enclosed a small simple time series simulation model that shows that the relation between trade openness and growth is far from significance and its sign changes from + to - (and reverse). If one has a panel for several countries, one must also take into account that trade as percent of GDP is, in general, higher the smaller a country.

In most cases, it is not feasible to relate a (what I would call) status variable (e.g. absolute values or percentage points for periods or points of time) to percentage changes. At least, one must have good reasons to do that.

Many others have illustrated why the sign differs from what you expected. Perhaps this (partial) summary will help:

• Incorrect data
• Incorrect theory
• The model suffers from specification error (e.g., underspecification = too few variables or leaving out important variables, overspecification = too many variables and also multicollinearity, endogeneity, incorrect functional form such as linear when the relationship is nonlinear, and others including using panel data when it is contraindicated
• Sampling issues (i.e., too few observations, bias and especially selection bias, non-random sampling and others)
• Software error (unlikely)
• The data, model, and relationship (i.e., the sign on the variable) are all correct, which is the same as an incorrect conjecture
• Others exist and remember this is only a partial list

I hope this helps in addition to some great answers that were previously posted:@ José-Ignacio Antón @ Emmanuel Ikpe Michael @ Anton Rainer @ Kehinde Mary Bello @ Olanrewaju Solomon Olatunji @ Samuel Tawiah Baidoo @ Cecilia Garavito

Sometimes using too many variables in the model can give unpredicted results for some of them. Maybe different choice of variables solves this problem.

Agreed with the answer of José-Ignacio Antón.