What Is EViews Software?
EViews — short for Econometric Views — is a statistical software package developed by IHS Global Inc. It was built specifically to analyse time series, cross-sectional, and panel data in an econometric framework. If you have ever worked with economic, financial, or policy data and needed to model relationships between variables across time, EViews is likely the tool you need.
At its core, EViews is a workfile-based environment. Every project lives inside a workfile that stores your data series, estimated equations, graphs, and output tables in one place. This makes it easy to pick up where you left off and reproduce your results exactly.
Key Insight
EViews works with two types of data files: native EViews workfiles (.wf1) and foreign files such as Excel spreadsheets (.xlsx), CSV files, and SPSS data files. You rarely need to reformat your data before opening it in EViews.
Time series data — the bread and butter of EViews — is any dataset where observations are recorded at regular intervals: annually, quarterly, monthly, or even daily. Because consecutive observations in time series are often correlated with each other, ordinary cross-sectional methods frequently break down. EViews handles this directly through built-in tests, estimators, and diagnostics designed for temporal dependence.
A Brief History
EViews evolved from an earlier DOS-based programme called MicroTSP in the 1980s. By the 1990s it had become a Windows application under the EViews brand, and it has since grown into one of the most widely used econometrics packages in academic research, central banks, government agencies, and financial institutions worldwide. Most undergraduate and postgraduate econometrics courses in Africa, Asia, and Europe teach with EViews because of its intuitive point-and-click interface — yet it also supports a full scripting language for automation.
What Is EViews Used For?
EViews is used across a wide range of quantitative research tasks. Below are the most common applications, matched to the EViews module that handles them.
| Task / Research Question |
EViews Tool |
| Is my time series stationary or does it have a unit root? | Unit Root Tests (ADF, PP, KPSS) |
| What is the relationship between Y and several X variables? | OLS / Least Squares Regression |
| Do variables share a long-run equilibrium? | Johansen Cointegration / ARDL Bounds Test |
| What are the short-run and long-run dynamics together? | ARDL / Error Correction Model (ECM) |
| Does X Granger-cause Y? | Granger Causality Test |
| How do shocks propagate across variables? | VAR / Impulse Response Functions |
| Does volatility cluster in financial returns? | ARCH / GARCH Models |
| Is there heteroscedasticity in my residuals? | White’s Test / ARCH LM Test |
| Do residuals violate normality? | Jarque-Bera Normality Test |
| Is there multicollinearity among regressors? | VIF / Correlation Matrix |
| Panel data across firms, countries, or individuals? | Pooled OLS / Fixed Effects / Random Effects |
| Which panel model is most appropriate? | Hausman Test / Wald Test |
In practice, EViews is used heavily by researchers in economics, finance, development studies, public health, agriculture, and environmental science — any discipline where data is collected repeatedly over time or across units.
How Much Does EViews Cost?
EViews is a commercial software package sold by IHS Markit / S&P Global (the developer, IHS Global Inc., was acquired by IHS Markit, now part of S&P Global). Pricing depends on the edition and licence type:
| Edition |
Typical Use Case |
Approximate Price (USD) |
| EViews Student Version | Course assignments, dissertations | ~$40–$65 (annual) |
| EViews University Single-User | Academic researchers, PhD students | ~$395 (perpetual) |
| EViews Commercial / Professional | Central banks, consultancies, firms | $1,000+ (perpetual) |
| EViews Network Licence | University labs, departments | Varies by seats |
💡 Note for students and researchers in Kenya & Africa: Many universities hold campus-wide or lab licences for EViews. Check with your institution’s IT or library services before purchasing. Tobit Research Consulting Ltd also offers EViews training and licensed workbooks for researchers on the continent.
A fully-functional free trial version is available from the official EViews website (eviews.com) and runs for a limited period. The Student Version is fully functional but imposes a dataset row limit and restricts certain advanced features. For most dissertation and academic journal research, the Student or University edition is more than sufficient.
How to Import Data from Excel to EViews
One of the most common questions from new EViews users is how to get data from an Excel spreadsheet into an EViews workfile. The process is straightforward and takes less than two minutes once you know the steps.
1
Open EViews and go to File → Open → Foreign Data as Workfile
In the EViews menu bar, click File, hover over Open, and select Foreign Data as Workfile. This tells EViews you are importing from a non-native format.
2
Browse to your Excel file and click Open
Navigate to the folder where your .xlsx or .xls file is saved. Select it and click Open. EViews will display an import preview window.
3
Specify the workfile frequency and sample period
In the import dialogue, set the data frequency (Annual, Quarterly, Monthly, etc.) and the start date and end date of your data. This tells EViews how to label each observation in time.
4
Confirm variable names and click Finish
EViews reads the column headers from your Excel sheet as variable names. Review them, make any corrections, and click Finish. Your variables now appear in the workfile window.
5
Save the workfile as a native EViews file
Go to File → Save As and save as an .wf1 file. From now on, open this EViews workfile directly — you do not need to re-import from Excel every time.
Pro Tip
Before importing, make sure your Excel sheet has variable names in the first row and no merged cells, blank rows, or special characters in the headers. EViews variable names must be alphanumeric and cannot start with a number. Use underscores instead of spaces (e.g., stock_price not stock price).
How to Use EViews to Run a Regression
Running a regression in EViews involves specifying an equation object, choosing your estimation method, and interpreting the output table. Here is a step-by-step walkthrough for an Ordinary Least Squares (OLS) regression.
Step-by-Step: OLS Regression in EViews
1
Select your variables in the workfile window
Hold Ctrl and click to select all relevant variables — start with your dependent variable, then select all independent variables.
2
Right-click → Open as Equation
Right-click on the selection and choose Open → as Equation. The Equation Estimation window will open.
3
Confirm the equation specification
EViews auto-fills the equation box in the format:
stock_price interest_rates inflation_rate gdp forex_rate balance_of_trade c
The c at the end represents the constant (intercept). Make sure it is included.
4
Select Method: Least Squares (LS) and click OK
In the Method dropdown, select LS — Least Squares (NLS and ARMA). Click OK and EViews estimates the model.
How to Interpret OLS Regression Results in EViews
The regression output table contains several key statistics. Here is how to read each one:
| Output Item |
What It Means |
Decision Rule |
| Coefficient | The estimated effect of a one-unit change in X on Y, holding others constant | Positive = positive relationship; Negative = inverse |
| Prob. (p-value) | Probability of observing this result if the true coefficient were zero | p < 0.05 → statistically significant at 5% |
| R-squared | Proportion of variance in Y explained by the model | Closer to 1 = better fit (context-dependent) |
| Adjusted R-squared | R-squared penalised for number of predictors | Prefer this over R² when comparing models |
| F-statistic / Prob(F) | Tests whether all coefficients are jointly zero | Prob(F) < 0.05 → model is jointly significant |
| Durbin-Watson stat | Tests for first-order serial correlation in residuals | Close to 2.0 is ideal; <1.5 or >2.5 signals autocorrelation |
| Akaike / Schwarz criterion | Information criteria for model comparison | Lower value = better model (used in lag selection) |
Example Interpretation
If the coefficient on interest_rates is −2.36 with a p-value of 0.037, you would write: “Interest rates have a negative and statistically significant effect on stock price at the 5% level (β = −2.36, p = 0.037 < 0.05). A one-unit increase in interest rates is associated with a 2.36-unit decrease in stock price, all else equal.”
How to Interpret Unit Root Test Results in EViews
Before selecting a time series model, you must determine whether your variables are stationary — meaning their mean and variance do not change over time. A variable with a unit root is non-stationary and will produce spurious regression results if used in levels.
Understanding the Hypothesis
In EViews, the most common unit root test is the Augmented Dickey-Fuller (ADF) test. The hypotheses are:
- H₀ (Null Hypothesis): The data has a unit root (non-stationary)
- H₁ (Alternative Hypothesis): The data has no unit root (stationary)
Decision Rule
Rule of Thumb: If the p-value is greater than 0.05, fail to reject H₀ → the variable has a unit root (non-stationary). If the p-value is less than 0.05, reject H₀ → the variable is stationary.
Running the ADF Test in EViews
1
Double-click the variable in the workfile to open its series window.
2
Click View → Unit Root Test from the menu.
3
Select Augmented Dickey-Fuller as the test type, choose Level, and set the lag selection to Automatic (Schwarz Info Criterion). Click OK.
4
Read the output table. Focus on the ADF test statistic and the Prob.* value (MacKinnon one-sided p-value).
Interpreting the Output: A Worked Example
Say the ADF test for stock_price at level returns:
- ADF t-statistic: −1.025521
- Prob.*: 0.7425
Interpretation: Since 0.7425 > 0.05, we fail to reject H₀. The stock price series has a unit root at level — it is non-stationary. We must difference the series and re-test.
After applying first-difference (1st difference), if the p-value drops below 0.05, the variable is said to be integrated of order one, I(1). If it requires two rounds of differencing, it is I(2). Most macroeconomic variables are I(1).
Other Unit Root Tests in EViews
Besides ADF, EViews offers several alternative tests, each with different assumptions:
- Phillips-Perron (PP) test — Non-parametric correction for serial correlation; useful when residuals are correlated
- ADF-GLS test — Detrended version of ADF; more powerful in small samples
- KPSS test — Reverses the hypothesis: H₀ is stationarity. Useful for cross-checking ADF results
- NGP test — Modification of PP; based on detrended data from ADF-GLS
Best Practice
Always run at least two unit root tests (e.g., ADF and KPSS) and confirm they agree. ADF and KPSS use opposite null hypotheses, so if both reject their respective nulls, there is an ambiguity worth investigating.
How to Interpret ARDL Results in EViews
The Autoregressive Distributed Lag (ARDL) model is one of the most popular approaches in applied econometrics because it handles mixed orders of integration — it works when some variables are I(0) and others are I(1), without requiring all variables to be stationary at the same order.
When to Use the ARDL Model
Model Selection Based on Unit Root Results
All I(0)
All variables stationary at level → use Simple OLS / Linear Regression
Mix of I(0) & I(1)
Variables stationary at level and 1st difference → use ARDL Bounds Test
All I(1) — Cointegrated
All non-stationary but cointegrated → use Johansen / ECM / VECM
All I(1) — Not Cointegrated
All non-stationary, no cointegration → use Unrestricted VAR
Mix of I(1) & I(2)
Variables at 1st and 2nd difference → use Toda-Yamamoto (1995)
What the ARDL Model Estimates
The ARDL model simultaneously estimates short-run dynamics (how variables adjust in the near term) and the long-run equilibrium relationship (the stable relationship between variables over time). When a cointegrating vector is found via the bounds test, the ARDL model is reparameterised into an Error Correction Term (ECT).
Key Elements of ARDL Output to Interpret
| Output Component |
What to Look For |
How to Interpret |
| Bounds F-test | F-statistic vs. I(0) and I(1) critical bounds | F > upper I(1) bound → cointegration exists; F < lower I(0) bound → no cointegration |
| Long-run coefficients | Coefficient and p-value for each variable | Shows permanent effect; p < 0.05 = significant long-run relationship |
| Short-run coefficients | Differenced variables (ΔX) | Shows immediate/transitory effects each period |
| ECT (Error Correction Term) | Must be negative and significant | Speed of adjustment back to equilibrium; e.g., ECT = −0.45 means 45% of disequilibrium corrects each period |
| CUSUM test plot | Blue line within red boundaries | Confirms structural stability of the ARDL model |
Critical Check
The Error Correction Term (ECT or CointEq) must carry a negative sign and be statistically significant (p < 0.05). A positive ECT or an insignificant one means the model does not exhibit a valid error-correction mechanism — cointegration may not be present even if the bounds test suggested it.
How to Interpret GARCH Results in EViews
The Generalised Autoregressive Conditional Heteroscedasticity (GARCH) model is used when the variance of your residuals is not constant over time — a phenomenon called volatility clustering that is extremely common in financial returns, exchange rates, and commodity prices.
Before Running GARCH: The ARCH LM Test
You should confirm the need for a GARCH model by first running an ARCH LM test on the residuals of a preliminary regression. In EViews:
1
Estimate an OLS regression first. In the equation output window, go to View → Residual Diagnostics → Heteroscedasticity Tests.
2
Select ARCH from the test type list. The ARCH LM test output includes an F-statistic and a Chi-square statistic. If either p-value is below 0.05, ARCH effects are present and GARCH is warranted.
Running a GARCH(1,1) Model in EViews
1
In the equation estimation window, set the Method to ARCH — Autoregressive Conditional Heteroscedasticity.
2
In the ARCH specification, set ARCH order = 1 and GARCH order = 1 for a standard GARCH(1,1).
3
Click OK. EViews uses maximum likelihood estimation to fit the model. The output contains both a mean equation and a variance equation.
Interpreting the GARCH Output
| Parameter |
Symbol |
Interpretation |
| Constant (Variance eq.) | ω (omega) | Long-run average variance; must be positive and significant |
| ARCH term | α (alpha) | Effect of last period’s squared shock on current volatility; measures reaction to market news |
| GARCH term | β (beta) | Effect of last period’s variance on current variance; measures persistence of volatility |
| α + β | Persistence | Must be < 1 for stationarity. Values close to 1 indicate highly persistent (slow-decaying) volatility |
Example Interpretation
If α = 0.18 (p < 0.05) and β = 0.75 (p < 0.05), you would write: “The ARCH coefficient (α = 0.18) is positive and significant, indicating that past shocks significantly influence current conditional volatility. The GARCH coefficient (β = 0.75) is also positive and significant, confirming strong volatility persistence. The sum α + β = 0.93 < 1, confirming covariance stationarity of the variance process.”
A sum α + β close to but below 1 (e.g., 0.93–0.98) is typical for financial return series and indicates long memory in volatility — shocks take a long time to die out. A sum equal to 1 is called an IGARCH (Integrated GARCH) process.
EViews Model Selection: A Practical Summary
One of the most common questions among EViews users is: “Which model should I use for my data?” The answer almost always begins with the unit root test. Here is a consolidated decision framework drawn directly from the EViews Tutorial Guide.
Time Series Model Selection Flowchart
Step 1
Run unit root tests (ADF/PP/KPSS) on all variables. Determine order of integration: I(0) or I(1).
If All I(0)
→ Use OLS / Simple Linear Regression. Check residuals for heteroscedasticity (White/ARCH test).
If Mix I(0)&I(1)
→ Use ARDL Bounds Test. If cointegration confirmed, report long-run & ECT. If not, use ARDL without ECM.
If All I(1)
→ Run Johansen cointegration test. If cointegrated → VECM. If not → Unrestricted VAR.
Volatility
→ Run ARCH LM test. If ARCH effects present → use GARCH / EGARCH / TGARCH model.
Panel Data
→ Test: Pooled OLS vs Fixed vs Random. Use Hausman test to choose between Fixed and Random effects.
Frequently Asked Questions About EViews
What is EViews software used for in research?
EViews is used to estimate, test, and forecast econometric models. Researchers use it for time series analysis (unit root tests, VAR, ARDL, GARCH), panel data regression, cointegration testing, Granger causality, and diagnostic tests such as normality, heteroscedasticity, and serial correlation tests. It is especially popular in economics, finance, and development research.
Is EViews better than Stata or SPSS?
EViews, Stata, and SPSS serve overlapping but distinct niches. EViews is the strongest choice for time series and macroeconomic data. Stata is broader and better for panel micro-econometrics and causal inference designs (IV, DiD, RDD). SPSS is commonly used in social science survey analysis. For financial econometrics with ARCH/GARCH models and impulse responses, EViews has no peer in ease-of-use.
How do I know which unit root test to trust?
Use at least two tests. Run ADF and KPSS together — they have opposite null hypotheses, so they serve as a robustness check. If both agree (ADF rejects non-stationarity AND KPSS fails to reject stationarity), you can be confident the series is stationary. If they conflict, check whether there is a structural break in your data and consider the Zivot-Andrews test.
What does it mean when the ECT in ARDL is positive?
A positive Error Correction Term (ECT) means the model is diverging from the long-run equilibrium rather than correcting toward it. This is economically implausible and statistically invalid. A properly specified ARDL model with a valid cointegrating relationship must have a negative and significant ECT. A positive ECT usually signals model misspecification, incorrect variable ordering, or the absence of a true cointegrating relationship.
Can EViews run GARCH for non-financial data?
Yes. While GARCH is most commonly applied to financial returns, it is appropriate for any series showing time-varying volatility — commodity prices, inflation rates, exchange rates, and even agricultural output in volatile economies. The key criterion is whether your ARCH LM test detects significant ARCH effects in the residuals.
Is EViews available for Mac?
EViews is natively a Windows application. Mac users can run EViews using Boot Camp, Parallels Desktop, or a virtual machine running Windows. As of the most recent major release, there is no native Mac version. Many university computer labs and remote desktop environments provide access to EViews for Mac users.
How do I fix the problem of autocorrelation detected in EViews?
If the Durbin-Watson statistic or the Breusch-Godfrey Serial Correlation LM test detects autocorrelation, the standard solutions include: (1) adding lag terms of the dependent variable or independent variables to the model, (2) using Newey-West HAC standard errors which are robust to autocorrelation, or (3) reconsidering your model specification — autocorrelation is often a symptom of an omitted variable or incorrect functional form.
Final Thoughts
EViews remains one of the most powerful and accessible tools for quantitative researchers working with time series and panel data. Whether you are running a simple OLS regression, testing for cointegration with ARDL, or modelling financial volatility with GARCH, EViews provides intuitive menus, rigorous diagnostics, and publication-quality output.
The key to getting reliable results is always the same: begin with unit root tests, let the stationarity properties of your data guide your model choice, and run thorough residual diagnostics before drawing conclusions.
Based on the EViews Tutorial Guide by Tobit Research Consulting Ltd · tobitresearch.com
