Research Methods Quantitative
Session 1 - Working with Quantitative Data
Understanding Data
- Data is a set of observations of the world around us. Data is:
• Collected intentionally and systematically (random, unsystematic observations lead to
bias)
• Coded clearly, to facilitate understanding and comparison
• Analyzed deductively or inductively to draw conclusions
- Choose a data source based on the goal of your research. In quantitative research, this
means testing a hypothesis relating two constructs / concepts. Observations are rarely
identical to these concepts. We need observations that are strongly related to the
concepts we want to explore.
- Epistemic correlation refers to the strength of the logical connection between the
observed data and the concept it represents. Quantitative data come from observations
that are examples of broader concepts.
- Empirical context is the setting in which our study occurs:
• Relevant contexts are informed by RQ & theory
• The choice of context sets bounds on generalisability of findings
- The Unit of Analysis is the object our data describe - i.e. what the data are about.
• Individual (Entrepreneur, CEO, student)
• Organisation (Firm, Board, Investors)
• Time (quarter, year)
• Others (Country, Industry, Network)
Note: combinations of the above are possible! (Firm-Yr ... panel data)
- Quantitative data has naturally-numerical values, for example:
• Number of patents (#)
• Team Diversity (%, index)
• IPO valuation ($)
- Many qualitative characteristics can be useful for quantitative studies:
• Gender: binary or dummy variable (0,1)
• Education level – ordinal variable (0, 1, 2, ...)
• Industry – categorical variable (no logical order)
- Qualitative observations require codification as number values for use in quantitative
analysis.
Page 1 of 10
, Research Methods Quantitative
Coding Data into Quantitative Values
- Variable Coding:
• Observations: a set of people & firms with unique characteristics.
• Dataset: a table with rows for individuals, and columns defining their qualities.
• Variables: the set of columns / qualities in the dataset.
• Codes: the meaning behind the values taken for each variable in the dataset.
• Codebook: your documentation about the data, including each variable and its
coding.
- Naturally Quantitative Variables include:
• Continuous (Prices, C.A.R.)
• Count (# Acquisitions, # Startups)
• Proportions (%Family Ownership, %Female)
- Qualitative Variables in Disguise include:
• Binary / Dummy (0,1): Yes/No; Present/Absent
• Multinomial (0,1,2,3,...): Race, Industry, Country
• Ordinal (0,1,2,3,...): Education level, S/M/L
- When coding your own data (from web documents, etc.) you should bear in mind to:
1. Simplify → make it so that the values seem intuitive Variable = Female? Men take
value 0, not 1.
2. Document → others must be able to interpret meaning. Create a codebook and
maintain a log file.
3. Preserve Variance → don’t erase the nuance. Group firms into sizes from continuous
vars.
4. Align codes with insight you eventually need → Base this on your hypotheses.
Describing Quantitative Data
- Univariate Statistics describe sets of observations along a specific characteristic
(variable):
• Mean = average
• Mode = most frequent
• Median = middle Value of Ordered List
• Range = Min, Max values (set or difference)
• Standard Deviation = average distance from mean
- Histograms describe the distribution of a variable:
• Normal distribution (bell curve) = most useful!
Page 2 of 10
, Research Methods Quantitative
• Uniform distribution
• Poisson (count) distribution
- Histograms quickly reveal many univariate statistics. For example, normal distribution
reveals the mean (μ) and the standard deviation (s.d.; σ).
- Skewed Distribution is where one tail is longer than another. These distributions are
sometimes called asymmetric or asymmetrical distributions.
• Left (Negative) Skew: long “tail” on left, where Mean < Median.
• Right (Positive) Skew: long “tail” on right, where Mean > Median.
- Variables Vary. Without Variance, there is no covariance.
Without covariance, there is no correlation.
Without correlation, there is no causation.
If your variable only has one value, it explains nothing.
Populations and Samples
- Population refers to the entire set of existing individuals for our unit of analysis, and
within our empirical context. For instance, Census of People and Registry of Firms. Yet, it
is hard to know the entire population.
- Sample is a selection from the population, which makes data collection feasible and
informs statistics interpretation. Good samples represent the population (no bias).
- Sampling strategies:
• Random Samples: every individual in population has equal chance of selection (limits
potential bias). This is a statistical ideal, especially for experiments (e.g. Lottery - simple
random sample). Limitations of this strategy include that you may not find data and that
bias is still possible (by random chance).
• Systematic Sample: organise population by group and randomly select within
Page 3 of 10
, Research Methods Quantitative
groups. These groups have categories or ordered rank (i.e. Strata) and they are
geographically proximate (i.e. Cluster). Limitations include that it requires good
population data, where a benefit is that sampling bias is unlikely.
• Convenience Sampling: collecting data through your friends, peers, personal
network, or based on availability. A limitation is that the observed individuals are
unlikely to be representative of all individuals (biased), but the data are more easily
accessed, where the types of bias present can be understood.
- Worst-case scenario in sampling = If you do not know or cannot explain where the data
come from, no one can say whether your findings are useful.
- When working from biased samples, you must think critically: how does bias impact
our results? The assumption of regression is random unbiased sampling, so can we
interpret regressions on biased samples?
• Positive (/Negative) Bias: we believe the values in our data are higher (/lower) than in
the population.
• Conservative Bias: we believe the observed trend is more extreme in the population.
This is often preferred due to the limited chance of Type I error.
Data Collection
- Quantitative studies use many kinds of data:
• Archival data (government and industry databases) often contain financial and
demographic statistics for a variety of empirical contexts and units of analysis. This is
easily available, but often with poor documentation.
• Hermeneutics (analysis of text; content analysis) can be based on language use
frequencies in webpages, annual reports, strategic plans, interviews, etc. This is
accessible, yet mixes qualitative and quantitative approaches.
• Surveys are commonly the source of archival data, and the source for most large new
data collection initiatives. It is more precise, takes a lot of time.
• Other: Field notes, unobtrusive measures (uncommon).
- When collecting data, select a strategy that:
1. Aligns your theoretical constructs with observations.
2. Is feasible (and can be done in time!).
3. Reduces bias, or has a known type of bias you can acknowledge.
- What data might be useful?
• Check the methodologies of the papers you cite, and see what data they use.
• Many databases are online, or will be shared by researchers using them.
Page 4 of 10
Session 1 - Working with Quantitative Data
Understanding Data
- Data is a set of observations of the world around us. Data is:
• Collected intentionally and systematically (random, unsystematic observations lead to
bias)
• Coded clearly, to facilitate understanding and comparison
• Analyzed deductively or inductively to draw conclusions
- Choose a data source based on the goal of your research. In quantitative research, this
means testing a hypothesis relating two constructs / concepts. Observations are rarely
identical to these concepts. We need observations that are strongly related to the
concepts we want to explore.
- Epistemic correlation refers to the strength of the logical connection between the
observed data and the concept it represents. Quantitative data come from observations
that are examples of broader concepts.
- Empirical context is the setting in which our study occurs:
• Relevant contexts are informed by RQ & theory
• The choice of context sets bounds on generalisability of findings
- The Unit of Analysis is the object our data describe - i.e. what the data are about.
• Individual (Entrepreneur, CEO, student)
• Organisation (Firm, Board, Investors)
• Time (quarter, year)
• Others (Country, Industry, Network)
Note: combinations of the above are possible! (Firm-Yr ... panel data)
- Quantitative data has naturally-numerical values, for example:
• Number of patents (#)
• Team Diversity (%, index)
• IPO valuation ($)
- Many qualitative characteristics can be useful for quantitative studies:
• Gender: binary or dummy variable (0,1)
• Education level – ordinal variable (0, 1, 2, ...)
• Industry – categorical variable (no logical order)
- Qualitative observations require codification as number values for use in quantitative
analysis.
Page 1 of 10
, Research Methods Quantitative
Coding Data into Quantitative Values
- Variable Coding:
• Observations: a set of people & firms with unique characteristics.
• Dataset: a table with rows for individuals, and columns defining their qualities.
• Variables: the set of columns / qualities in the dataset.
• Codes: the meaning behind the values taken for each variable in the dataset.
• Codebook: your documentation about the data, including each variable and its
coding.
- Naturally Quantitative Variables include:
• Continuous (Prices, C.A.R.)
• Count (# Acquisitions, # Startups)
• Proportions (%Family Ownership, %Female)
- Qualitative Variables in Disguise include:
• Binary / Dummy (0,1): Yes/No; Present/Absent
• Multinomial (0,1,2,3,...): Race, Industry, Country
• Ordinal (0,1,2,3,...): Education level, S/M/L
- When coding your own data (from web documents, etc.) you should bear in mind to:
1. Simplify → make it so that the values seem intuitive Variable = Female? Men take
value 0, not 1.
2. Document → others must be able to interpret meaning. Create a codebook and
maintain a log file.
3. Preserve Variance → don’t erase the nuance. Group firms into sizes from continuous
vars.
4. Align codes with insight you eventually need → Base this on your hypotheses.
Describing Quantitative Data
- Univariate Statistics describe sets of observations along a specific characteristic
(variable):
• Mean = average
• Mode = most frequent
• Median = middle Value of Ordered List
• Range = Min, Max values (set or difference)
• Standard Deviation = average distance from mean
- Histograms describe the distribution of a variable:
• Normal distribution (bell curve) = most useful!
Page 2 of 10
, Research Methods Quantitative
• Uniform distribution
• Poisson (count) distribution
- Histograms quickly reveal many univariate statistics. For example, normal distribution
reveals the mean (μ) and the standard deviation (s.d.; σ).
- Skewed Distribution is where one tail is longer than another. These distributions are
sometimes called asymmetric or asymmetrical distributions.
• Left (Negative) Skew: long “tail” on left, where Mean < Median.
• Right (Positive) Skew: long “tail” on right, where Mean > Median.
- Variables Vary. Without Variance, there is no covariance.
Without covariance, there is no correlation.
Without correlation, there is no causation.
If your variable only has one value, it explains nothing.
Populations and Samples
- Population refers to the entire set of existing individuals for our unit of analysis, and
within our empirical context. For instance, Census of People and Registry of Firms. Yet, it
is hard to know the entire population.
- Sample is a selection from the population, which makes data collection feasible and
informs statistics interpretation. Good samples represent the population (no bias).
- Sampling strategies:
• Random Samples: every individual in population has equal chance of selection (limits
potential bias). This is a statistical ideal, especially for experiments (e.g. Lottery - simple
random sample). Limitations of this strategy include that you may not find data and that
bias is still possible (by random chance).
• Systematic Sample: organise population by group and randomly select within
Page 3 of 10
, Research Methods Quantitative
groups. These groups have categories or ordered rank (i.e. Strata) and they are
geographically proximate (i.e. Cluster). Limitations include that it requires good
population data, where a benefit is that sampling bias is unlikely.
• Convenience Sampling: collecting data through your friends, peers, personal
network, or based on availability. A limitation is that the observed individuals are
unlikely to be representative of all individuals (biased), but the data are more easily
accessed, where the types of bias present can be understood.
- Worst-case scenario in sampling = If you do not know or cannot explain where the data
come from, no one can say whether your findings are useful.
- When working from biased samples, you must think critically: how does bias impact
our results? The assumption of regression is random unbiased sampling, so can we
interpret regressions on biased samples?
• Positive (/Negative) Bias: we believe the values in our data are higher (/lower) than in
the population.
• Conservative Bias: we believe the observed trend is more extreme in the population.
This is often preferred due to the limited chance of Type I error.
Data Collection
- Quantitative studies use many kinds of data:
• Archival data (government and industry databases) often contain financial and
demographic statistics for a variety of empirical contexts and units of analysis. This is
easily available, but often with poor documentation.
• Hermeneutics (analysis of text; content analysis) can be based on language use
frequencies in webpages, annual reports, strategic plans, interviews, etc. This is
accessible, yet mixes qualitative and quantitative approaches.
• Surveys are commonly the source of archival data, and the source for most large new
data collection initiatives. It is more precise, takes a lot of time.
• Other: Field notes, unobtrusive measures (uncommon).
- When collecting data, select a strategy that:
1. Aligns your theoretical constructs with observations.
2. Is feasible (and can be done in time!).
3. Reduces bias, or has a known type of bias you can acknowledge.
- What data might be useful?
• Check the methodologies of the papers you cite, and see what data they use.
• Many databases are online, or will be shared by researchers using them.
Page 4 of 10
