Analysing grants for LGBTQI+ organisations

The end of June was a time of pride. 🏳️‍🌈

To help mark it, I analysed 360Giving data to understand grants made to LGBTQI+ causes.  It was also a good excuse to use some of the new principles I learned from the book Data Feminism, by Catherine D’Ignazio and Lauren F. Klein, and see if I could apply intersectional data analysis to 360Giving data.

Whilst it was a longer analysis than I expected, I wanted to document the approach and share my learning with others who might be interested in looking at data in this way.  

My main learnings were:

1. 360Giving data is getting better in quality, but there’s more we can do to enable more powerful analysis, like having more elaborated and well thought through grant descriptions. 
2. We can build on an existing body of knowledge in what terminology might be relevant for different intersections. 
3. We need to share more of our analysis methodology with others and keep improving it. No methodology is perfect and reviews by funders and beneficiaries can help us to reduce biases in the data analysis. 

What is intersectionality?

Coined by civil rights advocate Kimberlé Crenshaw, intersectionality is a theoretical framework for understanding how aspects of a person’s social and political identities (eg gender, race, class, sexuality, ability, etc.) might combine to create unique modes of discrimination and privilege. Intersectionality identifies advantages and disadvantages that are felt by people due to a combination of factors. 

This is why in this analysis I looked not only at the LGBTQ+ community but also considered ability, minorities and age groups.  

Getting the data

Some of our previous data analyses were done through searching and downloading data through our free search-engine for grants data, GrantNav. However, last year we created a datastore to allow data to be easily queried, both for software and analysts. If you are a developer or an analyst, you don’t have to download the data in bulk from GrantNav anymore – you can simply query it through Google Colab notebooks (and if you are a coder or analyst and want access to our data, send us this form).

A nice thing about Google Colab is that it allows us to share our codes and methods with others, so they don’t have to do all the work from scratch and we can get better code ideas as well. It’s win-win. Here is the Colab Notebook that David Kane and I built together for this analysis. 

For this analysis, I decided to focus on language and exploratory methods for a couple of reasons:

1. This analysis is done around a theme, and searching for words in grants can yield more relevant results (than their geography, for example). 
2. We don’t know who all the LGBTQ+ organisations in the UK are, and we assume that some of them are “below the radar” as well, so word analysis can help us expose relevant grants. 

I then searched for the following keywords from GLAAD Glossary of terms in the “Grant Title”, “Grant Description” and “Grant Programme” fields for all grants between the years 2017-2019 – “Gay”, “lesbian”, “bi-sexual”,”homosexual”, “transgender”, “queer”, “homophobia”, “intersex” and “LGBT*” (the * is a wildcard that can identify multiple endings such as LGBTQ or LGBT+). I omitted the word “pride” because the results were associated with many grants that were not LGBTQ related. The term “non-binary” never appeared by itself in the theme of sexual orientation, however, it comes up own its own in grants that are about biology and science, which are not relevant to this analysis, so I omitted that as well. The data was for the time period between June 2017 and June 2020.

The limitation of this approach is that we might have missed some organisations dedicated to working with the LGBTQI+ community where these terms might not be included in the grant description as it is a given. In an ideal world, we would also be able to search based on organisations’ charitable objects and activities to identify organisations too, but this data is not easily accessible at the current time.

Examining the data

After querying the data, I returned 639 results. I then tried to understand how many fields I could work with. I noticed that 72% of the grants had planned dates fields (which allow us to understand how long grants programmes are). 88% had data in the grant programme title field of the dataset (in which we could find indicative names for LGBTQI+ programmes). 

I also wanted to check that the grants in the dataset were relevant, so I ran a keyword analysis to identify the words that were repeated in the grants and see if our search word returned relevant grants. 

Below is a list of the top 20 most common words and their frequency. 

I also calculated some quick statistical figures for the time period of 2017-2019 to give us a baseline for grants trends in this topic.

The average is the mean, which is calculated by dividing the sum of the values in the set by their number.

The Median is the middle number in a sorted, ascending or descending, list of numbers

Cleaning the data

360Insights showed that almost 20% of the grants between 2017-2019 were given to  unregistered organisations. This made me wonder whether these were really small community groups, or whether the data needed to be cleaned and some organisational identifiers were wrong (you can read more on identifiers in 360Giving data here). 

After examining this for 3 hours, including looking up the charity name in FindThatCharity and searching it on Google to see what type of organisation it was, I found out that 83 grants – which accounts to 14% of all grants – were given to organisations not registered with the Charity Commission or Companies House – initiatives like Pride in different cities, sports groups or productions

I also found that six grants were given to International NGOs, an entity we can’t tag yet on 360Insights.

Analysing the data

What did the data tell us?

To make this an intersectional analysis, I wanted to know which different groups within the LGBTQI+ community received funding. Since the 360Giving Data Standard doesn’t have a beneficiary or recipient group type, I used keyword analysis again to extract the data, this time on titles and descriptions of grants. This is why writing good descriptions of grants can help in understanding the funding type – the more holistic it is, the better word analysis we can do. 

I then used the Google Sheets QUERY function to extract the relevant grants. In simple terms, this allowed me to search better for text without leaving Google Sheets. If you want to know more about it, QUERY function uses SQL like language to extract data from any spreadsheet, which actually makes your Google Sheet into a tiny, yet powerful, SQL server, with no need to install an actual server.

So I made the decision to make four subgroups: Ethnicity and race, youth, older people and people with a disability. For Ethnicity and race I used words from the ONS guidance. For youth I used the phrases “youth” and “young people”. For older people I used “elderly” and “older”. Lastly, for Disability I used the UN International Classification of Impairments, Disabilities, and Handicaps, out of all of it,  the words “disable”, “disability”, “deaf” or “autism” yielded grants. 

I didn’t cover mental health, migrants or refugee communities in this analysis, but they can be added if needed, using existing vocabularies (which are linked above). 

Amount of LGBTQI+  Grants by Population Group

As you can see, grants to support young people account for 23% of the total sample, grants to racial and ethnic minority populations are 6%, older people are 4% and people with disabilities is less than 1%.  Other grants are of a more general nature.

Average grant size by sub group through time

We can see that the average grant awarded per grant is higher than all grants average in 2017 for youth and minorities groups, but it drops drastically in 2018 and 2019. The only group that has an increase is the disability group, but there are no grants in this group for 2018 and the overall grant amount for it is small to begin with, making it an outlier. 

Why is there a drop in the average amount awarded? Can it be because of the grant duration period? 

Grant Duration by sub group

Examining grant duration by month does not help us to understand the drop in funding. I looked at the median point of grants duration (meaning the half point). All of our sub groups grant duration is higher than the total amount of grant duration which is a year. Youth is the only one that shows a declining trend in grant duration from 24 month in 2017 to 21 months in 2018 and finally 12 months in 2019.
Minorities have a jump in grant duration to 36 months in 2018, which can explain why they have less grants in 2019 (since the next renewal can be predicted to be in 2021). Older people shows a stable pattern of 17.5 month, then 12 months and then back to 17.5. Disability is a big outlier again with 36 months and 30 months of grant duration period, 3 times more than total grants.

Why is there a drop in grants average? Here are some avenues we still need to explore –

• Did all publishers that publish in 2017 published in 2019? How is this impacted by the length of grants and the relatively small sample?
• Did all publishers use the descriptions correctly? Did we miss any words or phrases?
• Did some funders change funding priorities and stop funding LGBTQI+ causes?
• Is there a change in funding within the LGBTQI+ groups, for example, increased funding for trans people but less funding to lesbian causes?

It should also be noted that the volume of grants is relatively small and some changes will be normal statistical variations.

What more can we do?

This analysis was an experiment in looking at intersectional data, on a theme that is already an intersection in itself (sexual orientation). While we did divide grants into subgroups, we didn’t check which grants were overlapping and appearing in more than one subgroup. Sometimes when grantees write lists of beneficiaries, they list more than one subgroup – meaning the grant is not dedicated to only one group. We also didn’t do any geographical analysis, which might have shown trends in funding in some parts of the country in comparison to other parts.
In addition, we used words in our analysis from taxonomies known to us or from our own knowledge from previous analysis.

If you think we used the wrong words or can add more words to the search, you can go directly to our working dataset on Google Sheets and add your own words. You can also book an Office Hour with us if you want to speak about this more.
It is also important to remember that this research was not comprehensive, but experimental, and it had its limitations.

To sum up, when it comes to text analysis It’s all about the quality of grant descriptions and titles. If you are describing in the grant description the activity and the different populations that will receive and benefit from the grant, it allows us (and you) to do better text analysis based on this field. If your grant descriptions are short and do not describe different populations and activities, it will be hard to find relevant grants information about those populations, especially since those are currently not recorded as part of the 360Giving Data Standard.

If you want to learn more about data feminism, read the book by Catherine D’Ignazio and Lauren F. Klein, which has been the inspiration to this blog post.