Categorizing Econometric Software
Economic software platforms have the ability to combine theories, mathematics, and statistics to manage and process data, test hypotheses, find patterns, visualize findings, and forecast trends. Therefore, they are vital in various organizations, ranging from academic institutions to central banks. But while they may share commonalities, they differ in their mechanisms, as follows:
General Purpose Programming Languages
These platforms are also known as open-source scripting languages or code-based statistical environments. Unlike other platforms that are menu-driven or those that offer graphic user interfaces, platforms in this category excel in offering their users complete control over data management as they allow them to write code. This flexibility has made them the industry go-to for research that requires intensive data manipulation or machine learning integration.
Top among the languages here are:
R
Modern econometrics heavily relies on R because it operates as an open-source object-oriented programming language that allows analysts to write their own code to manipulate data, run regressions, and generate charts. In many ways, it serves as an ecosystem such that its base software can run linear regressions while its packages can handle specific econometric techniques. For example, econometricians often use its Comprehensive R Archive Network packages to run everything from OLS regressions to time series analyses, thanks to the abundant user-contributed libraries of code that have been designed for such analyses.
So, who uses R and where? Given its flexibility, this software is commonly used in a wide range of applications. Not only is it being used in universities to test economic theories, but it is also commonplace in government agencies, tech companies, think tanks, and policy institutes.
Advantages and Disadvantages
R is completely free, which makes it accessible to everyone regardless of their financial capabilities. To add to this, whenever a new econometric method is invented, an R package for this is often published immediately, which eliminates the need for users to wait years for a commercial software update. Add the fact that its scripts are reproducible, and you have a goldmine when it comes to scientific integrity.
While R has many perks, it is subject to some disadvantages, including its steep learning curve, its high RAM use, and the lack of commercial support, which pushes users to rely on community forums for troubleshooting.
Python
Interestingly, Python was created as a general-purpose programming language, and since researchers often leaned on tools like R and Stata, it was not quite popular as a research tool back then. However, due to reasons such as the rise of machine learning and the growing need for scalable web applications, this programming language has become the go-to in modern economic analyses. Why?
By itself, Python was not built for econometrics. However, since it is an interpreted object-oriented language with an ecosystem of data science and statistics libraries, it serves as a foundation on which analysts can import libraries and create a powerful statistical tool. And since this language was built for software engineering, its code tends to be very clean, readable, and easy to integrate into other environments, such as apps and websites.
It thus comes as no surprise that Python is the mainstay in tech companies, government agencies, academic institutions, and other organizations that handle large datasets.
Advantages and Disadvantages
Python is well capable of processing large datasets in the tune of terabytes and is also effective in machine learning integration, and its models can be embedded into other software with ease. Best of all, it wears many hats such that it can do quite a lot, from running regressions to deploying regressions and scraping data from websites, all within the same script.
That is not to say that it does not have drawbacks. Since it is a general-purpose tool, you find that its econometric tools are often split across different packages, unlike tools like R and Stata, which have more comprehensive packages. Economists also find that running simple tests in Python, such as OLS regressions, requires a lot of setup code as compared to the simple commands they would run in other econometric software.
Hybrid Statistical Packages
This category of econometric software combines programs that feature dropdown menus while also enabling researchers to write their own code. It includes the programs that were designed to help analysts access and comfortably use statistical tools even without any advanced programming expertise. While Gretl and SPSS are also popular options in this category, many economists gravitate towards the following:
Stata
Stata’s name tells you everything you need to know about it. After all, it combines statistics and data, hence Stata. But is it as powerful as its users claim it is? Yes. This integrated statistical software comes with everything that researchers need to manipulate, visualize, and report data, thus making it quite vital in quantitative research. Unlike the open-source programming languages we have covered, such as R and Python, Stata comes with the added advantage of blending traditional command line interfaces (CLI) and script files with intuitive graphic user interfaces (GUI).
To understand why Stata is so widely used in quantitative research, it helps to explore its very architecture. To start with, modern Stata versions now use data frames that allow researchers to load, link, and manipulate various datasets at the same time in RAM, which speeds up operations. The Stata ecosystem further simplifies data analysis by offering several tiers, including do-files that allow researchers to write command sequences, ado-files that help users create custom commands, and multi-language integration. The latter integration allows researchers to call libraries from other software, such as Python, directly into their scripts.
Based on these capabilities, Stata augurs well with organizations that use observational and longitudinal data in their empirical research. For example, in a case where sociologists are analyzing data from a national longitudinal survey on the effects of parental income on the likelihood of children getting a degree, they would have data spanning decades, and Stata would offer them the tools necessary to break down the results of the stratified samples, cluster designs, etc.
Advantages and Disadvantages
One aspect that researchers love about Stata is the reproducibility of its scripts. A test that was run a year ago will yield the same results from the same dataset, provided that the analyst uses the same scripts, and this is exactly the kind of integrity that is necessary in research. On top of that, Stata’s commands are well-documented, which allows researchers to understand the formulas in use, which helps them justify their results. And, of course, there is the flexibility. Stata works for everyone, such that beginners can rely on the dropdown menus to build models while experts can write their own commands.
But unlike R and Python, Stata comes at a fee, which may not be affordable to independent researchers or organizations. It is also important to note that because this software works within the RAM, researchers working with large datasets run the risk of system crashes or freezes if they do not have hardware that can handle such a load.
EViews
Econometric Views, shortened as EViews, was initially built with the premise of running time series analyses and macroeconomic forecasts. As such, you find that it differs a lot from traditional spreadsheets and programming languages in that it relies on object-oriented structures. How so?
Well, when you open data in this software, you create a Workfile which follows a specific frequency structure, such as daily or annual. Everything within that Workfile is seen as an object. For example, a column of data is referred to as a series object while a regression model is seen as an equation object. Thanks to this approach, each time you run an analysis, the results show up inside that object. For example, the results of a regression would be saved inside the equation object. Beyond the Workfile approach, EViews is also known for enabling researchers to create formula series, which is pretty much like Microsoft Excel but with the added perk of handling advanced econometrics. What’s more, it allows you to tile the windows side by side such that you can watch various outputs and inputs at the same time, e.g., comparing the raw data to the regression model.
In most cases, you find that EViews dominates industries that deal with time series data, especially regarding the economy, financial markets, and business cycles. These include central banks, financial firms, investment banks, government bureaus, academic institutions, and so on.
Advantages and Disadvantages
EViews is ideal for time series analyses as it is not only able to handle lags but also aspects such as differences, growth rates, and frequency conversions fast and without requiring complex codes. Furthermore, it can generate dynamic or static forecasts in interactive formats that make it easy for researchers to explain their work. And it integrates well with global economic databases, allowing you to access live data and use it in your Workfile. To add to this, just like Stata, EViews is fairly simple to learn.
So, does it have any downsides? Unfortunately, it does. One major setback is its scripting limitations. While it does have a programming command language, researchers find that writing scripts in this software is much harder than it is in programming languages like R and Python. Moreover, this software comes at a high licensing fee and is poorly optimized to handle cross-sectional data.
Matrix Programming Languages
These languages are also referred to as matrix-oriented computational languages. Unlike the menu-driven packages we covered in the past section, these languages are designed to build estimators and models by scripting matrix algebra. The premise here lies in filling the gaps in research that come about when new methodologies or novel computations are not yet available in standard statistical software. Thus, you find that the use of matrix programming languages in econometrics leans on modeling, simulations, and the development of custom algorithms. Ox, RATS, and GNU Octave are some of the examples you might encounter in research. But for the purpose of this guide, we will focus on two of the top options, which are as follows:
MATLAB
While Stata was mainly built for social science data and EViews was geared towards time series analyses, MATLAB, which is short for Matrix Laboratory, was designed for advanced quantitative analysis. It is essentially a high-level programming language that treats every piece of data as a matrix, thus making it very efficient at solving complex structural economic models and dynamic systems. As such, you find that it is often used in organizations where mathematical simulation and system designs are needed before researchers and their teams can build products for the real world.
Take quantitative financial analysis, for example. Since activities such as pricing derivatives and high-frequency trading require intensive calculations, analysts choose MATLAB as it allows them to combine computer science with pure math. Let’s say, for example, that a Wall Street analyst wants to build a model that can run 10,000 Monte Carlo simulations at the same time so as to calculate the risk profile of a stock option. They can turn to MATLAB to develop and test such a model.
Advantages and Disadvantages
Thanks to the design of its core engine, MATLAB treats data as matrices. As a result, it is able to perform complex simulations faster than native programming languages, and it also makes complex computations seem straightforward. It is also worth noting that it comes with a highly interactive environment that enables researchers to engage in all sorts of activities, from visualization to debugging, all without having to write extensive code. Besides, it comes with specialized toolboxes suited for various disciplines that users can use. But the thing that gets most analysts talking is its easy integration with Simulink, which enables them to design, simulate, and generate code for dynamic systems. And while it is a very advanced language, it has professional documentation as well as online help available to help researchers make the most out of its core functions and toolboxes. Besides, it comes with memory-mapping, which enables it to process files that are too large to fit into the computer memory.
But as you may have guessed, it comes with some downsides. First, it requires you to pay for a license. Additionally, while basic scripting is fairly easy to learn, researchers find that mastering the language for advanced tasks is quite difficult.
GAUSS
Did you know that this matrix language was created back in 1984? The idea here was to enable researchers to write mathematical algorithms using syntax that looked like real-world matrix algebra equations. That is why you find that it allows users to write out mathematical equations almost as they look on paper.
But how does it work? Instead of providing you with a dropdown menu, GAUSS allows you to write the code that instructs it on what to do, thereby translating mathematical formulas into computer code. Thanks to this, researchers are able to build new and custom models. Luckily, they do not need to do all the heavy-lifting themselves as this software offers pre-built applications and packages that they can use for tasks such as panel data modeling and time series forecasting.
Given how niche GAUSS is, it is often found in advanced quantitative research settings, such as PhD research and quantitative risk management. For example, a university professor may want to develop a new regression model that is not available in standard statistical software and would thus use GAUSS to write the mathematical foundation from scratch. In the same way, where a financial firm needs to rebalance its asset portfolios based on a changing economic climate, it may turn to GAUSS to run massive estimation models in real time.
Advantages and Disadvantages
The fact that GAUSS allows researchers to write code from textbook theories has to be one of its key selling points. It also helps that this software can handle large amounts of data quite well and is faster than standard programming languages when it comes to heavy-duty and custom operations. Even so, this software lacks the simple dropdown menus that beginners often rely on when analyzing data, and without knowing the exact matrix math behind what you want to do, using the software is not possible. It also comes at a cost and lacks the extensive communities available in programming languages like R and Python.