Lecture 7:

Factor analysis:

• Technique to reduce a large amount of information (contained in a number of original variables into a simple message (fewer variables, factors) with a minimum loss of information
• Done at the start of your empirical analysis, before your regression analysis

Latent variables:

• Often we need to measure things that cannot be measure directly or cannot be observed (e.g. burnout), because it could have multi dimensions. So you can measure faucets. Next we can see whether these facets reflect a single variable.  So are these different facets driven by the same underlying variable

Main use of factor analysis:

• To understand the structure of a set of variables (structure of a latent variable)
• To construct a questionnaire to measure an underlying variable (design a questionnaire to measure the variable)
• To reduce a dataset to manageable size while retaining as much of the original information as possible (e.g. merging two variables)

Step 1:

Calculate the correlation coefficients for each pair of these variables.

Step 2:

Cluster variables, clusters of large correlation coefficients suggest that those variables could be measuring aspects of the same underlying dimension. These underlying dimensions are known as factors (latent variables)

You can also make is graphical:

Imagine two factors as being axes of a graph, (plot these variables). Coordinate of variables indicate the strength of the relationship between the variable and each factor. The coordinates of a variable are known as factor loadings (correlation between a factor and a variable , axes {-1.1}. ideally a variable should have large coordinate for one axis, small coordinate for the other axis. Variable relate to one axis

Mathematical:

• Factor matrix: matrix of factor loadings.

Assumption: the factors represent real-world dimensions, which must be guessed at by inspecting which variable have high loads on the same factors

It is possible to find as many factors as variable, but decide how many factors to keep (extraction) using eigenvalues:

• Indicate the substantive importance of a factor
• Retain only factors with large eigenvalues >1
• Scree plot: graph eigenvalues against the factor with which the eigenvalue is associated
• Cut-off point of inflection (where s-slope of line changes dramatically) à often somewhere around one

Decision is a must theoretical as statistical

Once factors are extracted, we can calculate loadings usually, variables have high loadings on the most important factor and smaller loadings on other factors.

What is factor rotation:

• Visualise factors as axes
• To ensure load maximally to only one factor and have loading of approximately zero on the other factor
• Use orthogonal: independent/uncorrelated factors
• Oblique: allows factors to be correlated
• Choice depends on whether there is a good theoretical reason to suppose that the factors are related or independent

Reliability analysis:

• Extent to which a variable is consistent in what it is intended to measure
• An assessment of the degree of consistency between multiple measures of a variable
• Other things being equal, a person should get the same scores on a questionnaire if he/she completes it at two different points in time
• Use Cronbach’s alpha, required to be above 0.8 or 0.7
• Values substantially lower indicate an unreliable scale.

Exploratory factor analysis:

• Searching for structures among a set of variables
• Data reduction method
• Goal: identify appropriate factors

Confirmatory factor analysis:

• You have some preconceived thoughts about the structure of the data
• Based on theory or previous research
• Goal: assess the degree to which the data meet the expected structure and test hypotheses about the structures of latent variables

Article:

• Research with focus on developed countries, but what about emerging economies
• New theoretical explanations regarding outward FDI and institutional environment
• Relationship between firm-specific ownership advantages and outward FDI
• Factors on which this relationship is contingent
• Focus on China
• Moderation: home industry competition and export intensity positively moderate the positive relationship between a firm’s ownership advantages and international venturing
• Mediation: the extent to which a firm engages in innovations, venturing and strategic renewals mediates the positive relationship between its ownership advantages and intentional venturing
• Sample: representative, Chinese, manufacturing, not-foreign-invested firms, located in Beijing, Shanghai and Guandong
• Source: questionaries to CEOs and/or deputies in 2003 and 2004, two surveys (causal relationship
• Methods: exploratory factor analysis and OLS regression
• Variables: Dependent variables: international venturing, independent variables: technological capabilities, management capabilities, home country network ties. Moderators: home country competition and export intensity. Mediators: corporate entrepreneurship (innovation, venturing, strategic renewal) and control variables