Bachelor's and master's thesis in criminology - analysing reports and statistics

Reports and statistics in a criminology dissertation should answer a research question rather than simply fill a chapter with numbers. First check what the report's authors counted: offences, proceedings or people. The statistics cover events entered into a register under its rules. Some events are never reported or detected, so the report's total does not describe all crime.

You have a police report, figures for several districts and a topic about safety. Drawing a chart is straightforward; explaining what it shows is harder. Do the figures concern incidents, recorded offences, proceedings, suspects or people with final convictions? Similar labels do not imply the same unit. This guide develops an analysis of property crime, from narrowing the problem to interpreting an increase that cannot automatically be called an increase in all crime.

Formulate a problem that matches the available data

If you have only a few annual datasets, “Crime in Poland” will soon prove too broad. Choose an offence category, area and period, then write down what you want to compare. In the example below, we ask: “How did the number and rate of recorded offences in a selected category change in two districts over a comparable period?” This dissertation describes changes visible in the data. Explaining all their causes or assessing police effectiveness would require additional material.

Preserve the source's category name even in your working question. If the dataset concerns offences confirmed in completed preparatory proceedings, do not shorten that to “all offences committed”. Such a substitution broadens the subject without adding evidence. Explain an indicator in accessible language, but retain its meaning.

Decide what you want the figures to tell you. To study change over time, compare years. To study composition, examine how much of a defined total each category represents. To compare districts, you need data collected on the same basis. Choose one main task and select charts that help answer it. You do not have to use every dataset simply because you downloaded it.

Identify the unit of analysis

Check what one case means in each column. One person may be linked to several offences, and one set of proceedings may cover several acts. Counts of people, offences and proceedings are therefore not interchangeable. A suspect is also not the same as a person with a final conviction. Read the detection rate using the source's definition: it does not tell you what proportion of all offences committed, including those never recorded, was detected.

Statistics Poland's Local Data Bank metadata explain that confirmed offences concern incidents whose criminal character was established in completed preparatory proceedings. They also describe changes in data coverage and the calculation of detection rates. Record the definition and dataset version you use. Do not invent a definition from an abbreviated chart label.

Before calculating anything, describe each column you use: what it counts, its unit, source, area and period. Record offence counts separately from the population by which you will divide them. Do not assign a regional figure to a single city. Check whether the year in the dataset refers to registration of a case or commission of an offence. Without that distinction, you may describe events as belonging to the wrong period.

Follow the path from an incident to a statistical record

For an incident to appear in the statistics, it must be reported or detected and then entered in the appropriate category. The register's rules matter at each stage. Changes in reporting, detection or classification may also change the total in a report. When a figure rises, look beyond the chart to the data notes and documents explaining how the records were collected.

In her article on statistical descriptions of crime, Janina Błachut discusses different perspectives on interpreting statistics and the differences between official data, victimisation surveys and self-report studies. The practical implication is that sources may complement one another without counting the same thing. They cannot simply be combined as equivalent observations.

A report does not tell you exactly how many offences remained outside its register. Do not add a chosen percentage for the “dark figure” to its total. If you want to use an estimate from another study, check the incidents, population and years it covers, together with its uncertainty. An estimate for one type of victimisation cannot simply be applied to other categories or to all crime.

Choose sources according to their role

Alongside the numerical file, keep the notes and literature that will help you understand it. Use a local institution's report to examine its activities, while remembering who is describing them. Look in research articles for findings and explanations you can compare with your own results. A news story may draw attention to a problem without establishing its scale. For each source, note what you will actually use it to establish.

• Main dataset: comparable values addressing your central question.
• Metadata: definitions, coverage, corrections, collection changes and missing-value rules.
• Research literature: concepts, previous findings and possible explanations.
• Contextual sources: documents that help you check whether a proposed explanation fits the area and period you are studying.

Keep the original download, access date and a description of your selection. Record filters and units. If the institution later corrects the data, you can identify why values differ. Do not silently replace data halfway through the dissertation. An update may be appropriate, but it requires recalculation and consistent documentation of the version used.

Check comparability across years and areas

Two columns with the same name may contain data collected differently. Check which offences were included, which institutions reported them, the area covered and the cut-off date. Read any notes about changes to definitions or legal classifications during your period. You need the documentation for this particular series. A change affecting one category does not necessarily apply to every figure in the report.

If district boundaries or recording rules changed, do not automatically join all years into a single trend. You can choose a shorter comparable period, discuss the periods separately or mark a break on the chart. Explain the decision in your methodology. Readers should know whether they are seeing a change in offence counts or a change in how offences were counted. Do not hide the difference to make a chart look smoother.

Do not add an annual police total to a court total until you know what each covers. The same case may appear in both institutions' records, while the datasets may concern different stages and periods. Their sum therefore need not represent distinct incidents. Following cases through the system requires a way to link them. Similar annual totals are not enough.

Calculate counts, rates and changes consistently

In the example below, district A has 120 recorded offences in a category and 60,000 residents; district B has 180 offences and 120,000 residents. The rates per 100,000 residents are 200 and 150 respectively. B has more recorded offences but a lower population-based rate. Both descriptions are correct and answer different questions.

Calculate the rate by dividing recorded offences by the population for the same area and period, then multiplying by the chosen base, here 100,000. This helps compare districts with different populations. It does not let you read an individual resident's risk of victimisation from the chart. Incidents may involve visitors or repeatedly affect the same people, and the meaning of the rate also depends on the offence category.

An increase from 100 to 120 is a relative change of 20% and an absolute change of 20. If a category's share rises from 10% to 15%, the difference is 5 percentage points. With a small starting value, a large percentage may concern only a few cases, so present counts as well. When the starting value is zero, the ordinary percentage-change formula does not apply.

Distinguish zero, missing data and non-comparability

An empty cell does not necessarily mean that nothing was recorded. It may indicate missing reporting, unavailable information, an inapplicable category or another condition defined in the legend. Preserve original codes and explain how you handled them. Replacing every empty cell with zero can create false declines and subsequent apparent increases.

If you cannot establish what a symbol means, do not use it in a calculation requiring a known value. Explain the omission. A chart should show the actual scope of knowledge: retain a gap instead of connecting it with a line implying observations. Give the unit, period, source and details of your own calculations for each figure.

Consider competing explanations for an increase

If the register contains more cases, examine several possible explanations: more incidents, more reporting, greater detection or different recording rules. Consider a campaign encouraging residents to report offences. An announcement does not establish that it took place. A report tells you how the organiser describes the activities; suitable records of those activities may corroborate them. Even then, you do not yet know whether the campaign caused the later increase in reporting.

A victimisation survey may contribute if its scope fits your question. Examine sampling, question wording, reference period and the definition of experience. A study of adults does not automatically describe children. Lifetime experience is not directly comparable with annual registrations. Do not subtract values measured in different units.

Also check when the offences occurred and when the cases were registered or completed. A year's figures may include earlier incidents. The sequence “first a campaign, then more reports” is not enough to establish why reporting rose. If the sources cannot resolve that question, discuss the possible explanations and identify the missing information. This makes clear what your dissertation has established and what remains unanswered.

Explain the limits of knowledge about unrecorded crime

In “Czy «ciemna liczba przestępstw» istnieje?”, Błachut shows that measurement depends on what we understand by crime and the assumptions we adopt. Name precisely what your dissertation examines: registry entries, participants' accounts of experiences or an estimate from a particular study. The phrase “actual crime” cannot replace an explanation of where a figure comes from and what it covers.

An unknown number does not become known because the literature gives it a name. You can explain why some incidents remain outside registers and discuss research on that problem. Each time, identify the people and incidents those studies covered. Do not extend their findings to the whole population without justification. Reasons for non-reporting also require suitable evidence; they cannot be inferred from a gap in the statistics.

Show your analytical contribution when using existing data

Working with existing data need not mean summarising a report. You choose the question, justify the selection of comparable datasets, calculate measures and examine what the comparison shows. Make those steps visible in the chapter. Identify figures taken from a source and figures you calculated yourself. Readers should be able to follow your reasoning and distinguish the report author's commentary from your interpretation.

Return to the two districts. Show offence counts first, then population-based rates, followed by changes in both over time. If district B has more offences but a lower rate, explain the difference in population size. Do not omit a measure to make the comparison look more persuasive. Try answering your main question in one sentence and check whether the calculations support it. Keep other charts where they help readers understand that answer.

Do not compare every possible pair of variables in the hope of finding an interesting relationship. With many comparisons, particularly across only a few years, an apparent pattern is easy to find. First explain why you expect a relationship and whether your data can address it. Even if two values rise together, one need not cause the other. When studying change over time, also account for the possibility that observations from successive years are related.

Before calculating shares, check whether the categories belong to the same total. They may overlap or omit some cases. Establish the denominator before deciding which values can be added. Do not remove a category simply to make percentages total one hundred. If you combine report entries into your own groups, record exactly which entries belong to each. Readers can then trace your summary back to the source figures.

Organise the chapter around an answer, not successive reports

In the methodology, explain why you selected these sources, how you checked comparability and what you calculated. In the results, answer the question about counts, rates or category shares. Then discuss possible explanations and compare the findings with the literature. Make sure readers can distinguish a reported figure from your calculation and from a suggested cause of change. They may appear together, but they should not sound like a single established fact.

For support with data, see “Statistical analysis for bachelor’s and master’s theses”. For developing summaries and their interpretation, see “Analysis and presentation of research results for your thesis”. The subject-specific page is Criminology. For our discussion, have your data files, explanatory notes, research question and supervisor's comments ready. We can then identify whether you need help comparing figures, choosing a rate or writing the interpretation.

On your final read, pause at the words “increase”, “threat” and “effectiveness”. Check the category, period and basis of comparison in each such sentence. Be careful not to infer individuals' characteristics from differences between districts. A correlation between district-level measures does not explain why a particular person committed an offence. Such questions require evidence about people, rather than only totals for geographical areas.

Keep the calculations so that someone can follow the steps from the downloaded file to the finished chart. Separate source data, the organised dataset and results. Record which areas were combined and which entries were omitted, with the reasons. After correcting a numerical error, reread the commentary: the old explanation may no longer fit the new chart. Finally, check that repeated values use consistent units and decimal places throughout.

If you need a model draft of a section based on these data, we can first agree which comparisons to include, the years covered and how to explain your calculations. There is no need to add a survey simply to make the study look more complete. It is worth considering when it could answer an important question the reports leave unresolved. It would, however, need a separate design and actual data collection.

If you also need an outline and a larger piece of writing, Bachelor’s and master’s thesis writing can cover a model bachelor’s or master’s thesis, or a selected criminology chapter. Send your topic and available sources so that we can establish which question they can answer and what is still missing. I will confirm the price and deadline before starting.

Literature and data documentation

• Błachut, J. (1999-2000). Problemy związane ze statystycznym opisem przestępczości w oparciu o dane statystyk policyjnych. Archiwum Kryminologii, 25, 123-140. DOI: 10.7420/AK1999-2000D.
• Błachut, J. (2007-2008). Czy „ciemna liczba przestępstw” istnieje? Archiwum Kryminologii, 29-30, 75-80. DOI: 10.7420/AK2007-2008F.
• Główny Urząd Statystyczny. Bank Danych Lokalnych: metadata for group G371, Przestępstwa stwierdzone przez Policję w zakończonych postępowaniach przygotowawczych. Group description displayed on subgroup P4633 page. Metadata and definitions. Accessed 4 November 2026.

Research articles support critical interpretation; the definitions for a particular series must be checked in its current documentation. An older publication does not establish that recording rules remain unchanged. Your dissertation bibliography should also include every report actually used, with its title, institutional author and data period.