Manuela Zind, As I understood, your data is based on a questionnaire or survey.

Principal component analysis (PCA) could be helpful for basic understanding of your data, and correlations/clustering based on the distribution of variables along the PC1 and PC2 axis.

You could consider each individual farmer/candidate as separate experimental units (rows of the data table). And arrange their answers to each question (columns of the data table).

And check whether PC1 and PC2 are represent majority of the variability of the data set ( > 60-70%), before marking your conclusions.

Dear Manuela Zind, I understood that you have only categorical variables and that you want to do an exploratory analysis of your data. Therefore, you will need to deal with unsupervised machine learning techniques. Basically, your possibilities are: cluster analysis, factor analysis, correspondence analysis. On categorical variables you cannot apply clustering or factor analysis. These two techniques are applied to numeric variables and if they are applied to your database (via arbitrary transformations of categorical variables into numeric ones) you will be making the arbitrary weighting error.

The way to analyze the relationship between categories (without arbitrary weighting) of different categorical variables is through simple (for two variables) and multiple (more than two variables) correspondence analysis.

The output of this technique is a numeric variable (without arbitrary weighting). This output can, on its turn, be used in other non-supevised machine learning techniques that evaluate numerical variables (e.g., clustering and PCA).

I suggest you seek help to apply a multiple correspondence analysis to your data. If you see no patterns, your output can be applied to a PCA (to group variables) and/or clustering (to group observations). Hope this helps. Good luck with your data!

Article Multiple Correspondence Analysis